A CROSS-SCALE EXAMINATION OF HOW KNOWLEDGE AND THE PHYSICAL ENVIRONMENT INFLUENCE THE USE OF BEST MANAGEMENT PRACTICES IN US COMMODITY AGRICULTURE By Riva Caroline Hodges Denny A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Sociology—Doctor of Philosophy 2018 ABSTRACT A CROSS-SCALE EXAMINATION OF HOW KNOWLEDGE AND THE PHYSICAL ENVIRONMENT INFLUENCE THE USE OF BEST MANAGEMENT PRACTICES IN US COMMODITY AGRICULTURE By Riva Caroline Hodges Denny While agriculture is of the utmost importance to human societies for its production of food and fiber (and increasingly fuel), it is also a source of environmental harm as a producer of nonpoint-source pollution to the air and water. Even though efforts to control agricultural nonpoint-source pollution have focused on encouraging farmers to voluntarily use practices that will better keep soil and nutrients on farms, agriculture remain a substantial source of water pollution in the US. This has led to continued efforts to better understand why farmers choose to use these best management practices (BMPs) so that incentives can be improved. Therefore, this dissertation focuses on farmer decision-making related to the use of soil, water, and nutrient conserving BMPs in row-crop agriculture in the US Midwest, with particular interest in the importance of the roles of knowledge and the biophysical environment. Conceptually, my dissertation is informed and organized by a theory of action, in which purposive individual actions cumulatively create system-level effects. This is done by analyzing data at the individual, county, and state scales as well as across-scales in three empirical chapters. The first analysis uses qualitative interview data from 2014 with 151 Midwestern corn farmers in three states to explore the range and variation of the tillage practices in use, and to consider the reasons farmers give for their tillage decisions. The results show that farmers commonly use more than one style of tillage on their operation and base their tillage decisions on biophysical conditions, compatibility with other practices, and the relative savings and costs of different tillage styles. The second analysis uses 2017 farmer survey data from four states (n=814) to compare the relative importance of a wide range of predictors of BMP use from the literature on the use of five BMPs: pre-sidedress nitrate tests (PSNTs), nutrient maps, variable rate nitrogen (N), variable rate phosphorus (P) or potassium (K), and cover crops. It also explores the use of practice knowledge as a predictor of BMP adoption that has been underutilized in previous research. The results show little overlap in the predictors of the five practices, even between those with similar characteristics. Two predictors in particular—the use of independent crop consultants as an information source and a general measure of practice knowledge—were found to be the most important influences across all five models. The third analysis considers the cumulative decisions of multiple farmers by using county and state data for the North Central region (1,042 counties) to examine the effects of conservation tillage, no-till and cover crops on county-level fertilizer use. The results show no significant total effects on fertilizer use by no-till, conservation tillage, or cover crops, though there is the potential for cover crops to increase fertilizer use. The importance of the biophysical environment is a key finding of all three analyses, and different biophysical conditions can increase and decrease the use of BMPs. The differences in effects of both social and biophysical factors across BMPs has important implications for future research and suggests the need for specific measures of practices that include temporal and spatial variation, and the careful consideration of practices individually, rather than as groups of ostensibly related practices. Knowledge also shows promise as a variable of importance in practice use—both through knowledge of how to use or implement a practice, and through knowledge of environmental-systems that shape how farmers perceive and interpret their biophysical context. Copyright by RIVA CAROLINE HODGES DENNY 2018 Dedicated to my parents as a very small repayment for all their love and support throughout the years. And to JD, whose support, patience, and humor have kept things real during this journey. You are appreciated. v ACKNOWLEDGEMENTS To my dissertation committee, Sandra Marquart-Pyatt (chair), Diana Stuart (former chair), Aaron McCright, Stephen Gasteyer, and Phil Robertson, my most heartfelt thanks for their feedback and guidance during my dissertation process. Extra special thanks to Diana for her early guidance and continued support in research and academic writing. Super special thanks to Sandy for being such an incredible advisor and collaborator, and for kicking my butt when I needed it; I cannot adequately express how important your mentorship and support has been to me these last three years. I also want to thank past committee members Phil Howard and Larry Busch for their time and contributions. Also, my sincere thanks to the farmers who took the time to sit for an interview or who filled out one of our many surveys—without them this research could not have happened. I also want to acknowledge and thank the MSU Environmental Science and Policy Program, the MSU Ecological Food and Farming Systems graduate specialization, the MSU W.K. Kellogg Biological Station Long Term Ecological Research program, and the MSU Graduate School for the financial support, large and small, that they have given me over the years, and which significantly facilitated my graduate experience and research. Last but not least, I want to thank the other graduate students who have mentored, challenged, supported, and collaborated with me (not to mention laughed and cried with me) and who made the years rich, even when they were stressful. You know who you are. vi TABLE OF CONTENTS LIST OF TABLES .......................................................................................................................... x LIST OF FIGURES ..................................................................................................................... xiii KEY TO ABBREVIATIONS ....................................................................................................... xv CHAPTER 1: INTRODUCTION ................................................................................................... 1 WHY AGRICULTURE? ............................................................................................................ 3 WAYS TO UNDERSTAND FARMER-DECISION MAKING ............................................... 4 Adoption-Diffusion Model ..................................................................................................... 5 Social Psychological Theories .............................................................................................. 10 OVERVIEW OF EMPIRCAL ANALYSES ............................................................................ 15 REFERENCES ............................................................................................................................. 17 CHAPTER 2: MOVING BEYOND BASIC TILLAGE CATEGORIES IN MIDWEST ROW- CROP AGRICULTURE: AN EXPLORATORY ANALYSIS OF TILLAGE PRACTICES AND REASONS .................................................................................................................................... 22 BACKGROUND ON TILLAGE.............................................................................................. 24 LITERATURE REVIEW ......................................................................................................... 27 METHODS ............................................................................................................................... 32 Overview ............................................................................................................................... 32 Data Collection ..................................................................................................................... 33 Analysis................................................................................................................................. 35 RESULTS ................................................................................................................................. 37 Primary Tillage Tools ........................................................................................................... 39 Secondary Tillage Tools ....................................................................................................... 40 Types of Tillage .................................................................................................................... 41 Tillage Decision Factors ....................................................................................................... 47 DISCUSSION ........................................................................................................................... 54 Variation in Tillage Practices and Meanings ........................................................................ 54 Variation in Tillage Reasons ................................................................................................. 56 Implications and Directions for Future Research ................................................................. 58 CONCLUSION ......................................................................................................................... 59 REFERENCES ............................................................................................................................. 61 CHAPTER 3: DRIVERS OF THE USE OF FIVE BMPS IN MIDWEST ROW-CROP AGRICULTURE .......................................................................................................................... 67 APPROACHES TO UNDERSTANDING BMP ADOPTION/USE ....................................... 68 Social-Psychological Theories .............................................................................................. 68 Adoption-Diffusion Model ................................................................................................... 69 The Potential Importance of Knowledge .............................................................................. 70 PREVIOUS RESEARCH ON DRIVERS OF BMP USE ........................................................ 76 Connecting Variables and Theories ...................................................................................... 77 vii METHODS AND ANALYSIS ................................................................................................. 77 Sample................................................................................................................................... 78 Variables ............................................................................................................................... 79 Analysis................................................................................................................................. 87 FINDINGS ................................................................................................................................ 88 General Approach to Results ................................................................................................ 88 Results of PSNT Model ........................................................................................................ 96 Results of Nutrient Map Model ............................................................................................ 97 Results of Variable Rate N Model ........................................................................................ 99 Results of Variable Rate P or K Model .............................................................................. 100 Results of Cover Crop Model ............................................................................................. 102 DISCUSSION ......................................................................................................................... 106 Independent Crop Consultants ............................................................................................ 106 Knowledge .......................................................................................................................... 107 Corn Yield ........................................................................................................................... 109 Similarities Between Practices ............................................................................................ 110 CONCLUSION ....................................................................................................................... 113 APPENDICES ............................................................................................................................ 117 APPENDIX A: Data Cleaning and Variable Calculation ....................................................... 118 APPENDIX B: Modeling Decisions ....................................................................................... 124 APPENDIX C: Additional Results ......................................................................................... 133 APPENDIX D: Mplus Code ................................................................................................... 150 REFERENCES ........................................................................................................................... 152 CHAPTER 4: QUESTIONING THE FERTILIZER SAVINGS INCENTIVE FOR BMP ADOPTION IN THE NORTH CENTRAL REGION ................................................................ 157 BACKGROUND .................................................................................................................... 159 Fertilizer and BMPs ............................................................................................................ 159 Coupled Natural and Human Systems ................................................................................ 163 RESEARCH PROBLEM AND QUESTIONS ....................................................................... 165 METHODS AND ANALYSIS ............................................................................................... 167 Data and sample .................................................................................................................. 167 Variables ............................................................................................................................. 168 Analysis............................................................................................................................... 172 RESULTS AND DISCUSSION ............................................................................................. 173 CONCLUSION ....................................................................................................................... 180 APPENDICES ............................................................................................................................ 185 APPENDIX A: Variable Calculations .................................................................................... 186 APPENDIX B: Centering Analysis Variables ........................................................................ 190 APPENDIX C: Extra Sample Descriptives ............................................................................ 192 REFERENCES ........................................................................................................................... 194 CHAPTER 5: CONCLUSION ................................................................................................... 203 MEASURES AND RESEARCH METHODS ....................................................................... 203 KNOWLEDGE AND THE BIOPYISCAL ENVIRONMENT .............................................. 204 PLANS FOR FUTURE RESEARCH ..................................................................................... 212 Tillage ................................................................................................................................. 212 viii BMP Adoption .................................................................................................................... 213 Fertilizer Use ....................................................................................................................... 214 REFERENCES ........................................................................................................................... 215 ix LIST OF TABLES Table 2.1. Sample recruitment sources by state. ........................................................................... 34 Table 2.2. Mean and median farm size in acres per operation by state in 2012. .......................... 35 Table 2.3. Cross-tab comparison of instances of tillage tool/type overlap, as percent of 151 and number of respondents. ................................................................................................................. 38 Table 2.4. Summary of tillage tool use and tillage type by state, as percent and number of respondents. .................................................................................................................................. 39 Table 2.5. No-till sub-categories by state and for total sample, as percent and number of respondents. .................................................................................................................................. 44 Table 2.6. Comparison of tillage tools/types used for conservation (or minimum) tillage and conventional tillage, as percent and number of respondents. ....................................................... 46 Table 2.7. Examples of quotes pertaining to the biophysical conditions that influence farmers’ tillage decisions. ............................................................................................................................ 48 Table 2.8. Examples of quotes pertaining to the crop rotation and residue management aspects of operational needs and practice compatibility features that influence farmers’ tillage decisions. . 50 Table 2.9. Examples of quotes pertaining to the fertilizer application aspects of operational needs and practice compatibility features that influence farmers’ tillage decisions. .............................. 52 Table 2.10. Examples of quotes pertaining to the operational savings and costs that influence farmers’ tillage decisions. ............................................................................................................. 53 Table 3.1. Overview of conservation/best management practice adoption meta-analyses. .......... 75 Table 3.2. Comparison between final sample (n=814) and 2012 Census of Agriculture on key demographic variables by state. .................................................................................................... 79 Table 3.3 Combined categories used in Prokopy et al. (2008) and Baumgart-Getz, Prokopy, and Floress (2012). .............................................................................................................................. 81 Table 3.4. Variables used in analysis, their definition and scale, grouped by author-defined categories, and using sub-categories from the literature. All variables are for 2016. ................... 83 Table 3.5. Analysis descriptives (n = 814 unless otherwise noted). ............................................. 89 Table 3.6. Select probit regression coefficients and standard errors for PSNT and nutrient maps (n=814). ......................................................................................................................................... 91 x Table 3.7. Select probit regression coefficients and standard errors for variable rate N and variable rate P or K (n=814). ........................................................................................................ 92 Table 3.8. Select probit regression coefficients and standard errors for cover crops (n=814). .... 93 Table 3.9. Sign of significant variables in analyses. ..................................................................... 94 Table 3.10. Total effects for significant variables on PSNT use in terms of probabilities with all other variables at average. ............................................................................................................. 97 Table 3.11. Total effects for significant variables on nutrient map use in terms of probabilities with all other variables at average. ............................................................................................... 98 Table 3.12. Total effects for significant variables on variable rate N use in terms of probabilities with all other variables at average. ............................................................................................. 100 Table 3.13. Total effects for significant variables on variable rate P or K use in terms of probabilities with all other variables at average. ........................................................................ 101 Table 3.14. Total effects for significant variables on cover crop use in terms of probabilities with all other variables at average....................................................................................................... 103 Table 3.15. Skewness and kurtosis for endogenous variables. ................................................... 126 Table 3.16. ICC comparison by sample and estimator. .............................................................. 129 Table 3.17. Wald Chi-square for joint test of logistic regression for state differences in practice use using final sample (n=814). .................................................................................................. 130 Table 3.18. Wald Chi-square for joint test of logistic regression for group-by-covariate interactions for select variables using final sample (n=814). ..................................................... 132 Table 3.19. Tolerance and variance inflation factor (VIF) for exogenous analysis variables (n=814). ....................................................................................................................................... 133 Table 3.20. Select Pearson correlation coefficients for exogenous variables where correlation coefficient is < -0.5 or > 0.5 (n = 814). ....................................................................................... 135 Table 3.21 Average and range of measurement model results across the five practice analyses. ..................................................................................................................................................... 136 Table 3.22. Probit regression coefficients and standard errors for PSNT and nutrient maps (n=814). ....................................................................................................................................... 137 Table 3.23. Probit regression coefficients and standard errors for variable rate N and variable rate P or K (n=814). ........................................................................................................................... 139 Table 3.24 Probit regression coefficients and standard errors for cover crops (n=814). ............ 141 xi Table 3.25. Standardized regression coefficients and standard errors for PSNT and nutrient maps (n=814). ....................................................................................................................................... 144 Table 3.26. Standardized regression coefficients and standard errors for variable rate N and variable rate P or K (n=814). ...................................................................................................... 146 Table 3.27. Standardized regression coefficients and standard errors for cover crops (n=814). 148 Table 4.1. Uncentered descriptive statistics for analysis variables (n=1042 unless otherwise noted). ......................................................................................................................................... 174 Table 4.2. Direct effect coefficients and standard errors from multilevel structural equation model with random intercepts for 12 states, 2012 (n=1042). ..................................................... 175 Table 4.3. Standardized, county-level, total effects on fertilizer use per acre. ........................... 179 Table 4.4. Counties excluded from analysis due to low amount of agriculture or withheld data. ..................................................................................................................................................... 192 Table 4.5. State-level averages for analysis variables. ............................................................... 193 xii LIST OF FIGURES Figure 1.1. Adopter categorization on the basis of an adoption frequency distribution (Rogers 2003:351). ....................................................................................................................................... 7 Figure 1.2. Conceptual diagram of the Theory of Reasoned Action and the Theory of Planned Behavior based on figures in Kaiser, Wölfing, and Fuhrer (1999:3) and Madden, Ellen, and Ajzen (1992:4). ............................................................................................................................. 11 Figure 1.3. Conceptual diagram of the Reasoned Action Approach based on figure in Fishbein and Ajzen (2015:22). .................................................................................................................... 12 Figure 2.1. Tillage definitions used by Vitale et al. (2011:252). .................................................. 29 Figure 2.2. Tillage definitions used by Andrews et al. (2013:504). ............................................. 29 Figure 3.1. Kaiser and Fuhrer’s (2003:606) conceptual model of how know knowledge types influence each other and ultimately ecological behavior. ............................................................. 72 Figure 3.2. Combining perceived practice characteristics from the adoption-diffusion model with the Reasoned Action Approach by Reimer, Weinkauf, et al. (2012). .......................................... 73 Figure 3.3. Probability of practice use by a typical farmer. .......................................................... 96 Figure 3.4. Graphical depiction of the predicted probabilities of select variables on cover crop use across a range of values. ....................................................................................................... 105 Figure 3.5. Graphical depiction of the probability of using BMPs at different frequencies of use of independent crop consultants as an information source, with all other variables at the average response level. ............................................................................................................................. 107 Figure 3.6. Graphical depiction of the probability of using BMPs at the five reported levels of knowledge, with all other variables at the average response level. ............................................ 108 Figure 3.7. Graphical depiction of the probability of using BMPs at the mean average corn yield, and at one and two standard deviations below and above the mean. .......................................... 110 Figure 3.8. Practice use by state for final sample. ...................................................................... 130 Figure 4.1. Diagram of feedback of CHN feedback in an agricultural system from Stuart et al. (2015:573). .................................................................................................................................. 164 Figure 4.2. Conceptual Multilevel SEM diagram of fertilizer use. ............................................ 171 Figure 4.3. Yield response graph (Vanotti and Bundy 1994:184). ............................................. 181 xiii Figure 5.1. Conceptual diagrams from Rosa (1998:35) on left and Carolan (2004:515) on right. ..................................................................................................................................................... 207 Figure 5.2. Simple example of realist (part A.) versus critical realist (part B.) approach to CNH feedback model. .......................................................................................................................... 211 xiv KEY TO ABBREVIATIONS AgCensus Census of Agriculture ANOVA Analysis of Variance BMP CFI Best Management Practice Comparative Fit Index CHANS Couple Human and Natural Systems CNH CRP CSP Couple Natural and Human Conservation Reserve Program Conservation Stewardship Program CTIC Conservation Technology Information Center EQIP IA ICC IN K MI Environmental Quality Incentives Program Iowa Intraclass Correlation Coefficient Indiana Potassium Michigan MIMIC Multiple Indicators and Multiple Causes MLR N Robust Maximum Likelihood Nitrogen NASS National Agricultural Statistics Service NRCS Natural Resource Conservation Service OLS Ordinary Least Squares xv P PBC Phosphorus Perceived Behavioral Control PSNT Pre-Sidedress Nitrate Test RAA Reasoned Action Approach RMSEA Root Mean Square Error of Approximation SEM Structural Equation Modeling SEMLV Structural Equation Modeling with Latent Variables SES TLI TPB TRA Social-Ecological System Tucker-Lewis Index Theory of Planned Behavior Theory of Reasoned Action USDA United States Department of Agriculture WLSMV Weighted Least Squares Means and Variance xvi CHAPTER 1: INTRODUCTION While agriculture is of the utmost importance to human societies for its production of food and fiber (and increasingly fuel), it is also a source of environmental harm as a producer of non-point-source pollution to the air and water (Aneja, Schlesinger, and Erisman 2009; Carpenter et al. 1998; Drinkwater and Snapp 2007; Lal 1993; Robertson et al. 2013; Robertson and Vitousek 2009; Uri and Lewis 1998). Nutrients, nitrogen and phosphorus in particular, are vital to crop production but also create great environmental costs when they leave the farm and enter surrounding waterways where they contribute to eutrophication, algal blooms and hypoxia (“dead zones”) (Blesh and Drinkwater 2013; Drinkwater and Snapp 2007; Hamilton 2015; Robertson and Vitousek 2009). Similarly, soil that washes off fields can impair surrounding water bodies through suspended soil particles that cloud the water and sediment accumulation at the bottom, and soil that is blown away by wind can impair air quality (Uri and Lewis 1998). Concern over soil erosion in the US goes back to at least the Dust Bowl of the 1930s (Halcrow, Heady, and Cotner 1982), while concern over the off-farm effects of nutrient loss have only been of widespread concern since the 1970s (Drinkwater and Snapp 2007). Efforts to control both issues have focused on encouraging farmers to voluntarily use practices that will better maintain soil and nutrients on farms and thus benefit both farmers and society. While many farmers have changed their production practices in a variety of ways in response to these efforts, both nutrient loss and soil erosion from agriculture remain a source of pollution for water bodies in the US (EPA 2017). This has led to continued efforts to better understand why farmers choose to use these conservation or best management practices (BMPs) so that they can be better encouraged to use more of them on more of their land. 1 State-led efforts to reduce both soil and nutrient loss in the US have historically taken a voluntary approach through a number of federally supported conservation programs. The largest and oldest US conservation program is the Conservation Reserve Program (CRP), which pays farmers to either remove from cultivation or not start farming sensitive land. Working-land programs such as the Environmental Quality Incentives Program (EQIP) and the Conservation Stewardship Program (CSP) provide cost-share and resources for farmers with more specific management goals related to a range of topics (Reimer and Prokopy 2014). Incentives for program participation have not typically exceeded making money available to farmers for setting aside land or to help in adopting a practice, and thus have been of limited effectiveness (Ribaudo 2015). The one exception is the Conservation Compliance provision, initiated in the 1985 Farm Bill, which requires that steps be taken to reduce erosion on land classed as highly erodible for the farmer to be eligible for commodity program payments (Arbuckle 2013). With the end of the commodity programs in the 2014 Farm Bill, Conservation Compliance for highly erodible land and wetlands is now required for eligibility for federal subsidies for crop insurance premiums among other things (NRCS n.d.). My dissertation focuses on farmer decision-making related to the use of soil, water, and nutrient conserving BMPs in row-crop agriculture in the US Midwest. In it, I seek to better identify the variables that drive farmers’ decisions to use or not use BMPs, with a particular interest on how the drivers may differ between practices, and to take a step towards considering how the use of BMPs influence other production decisions—in this case the degree of fertilizer use. The BMPs of interest in the analyses that follow are related to fertilizer rate and application methods, tillage practices, and the use of winter cover crops. These are all in-field practices that 2 can have a direct impact on fertilizer use, fertilizer use efficiency, and ultimately nutrient loss from agriculture. In my analyses I engage with and draw on the two most common conceptual approaches used in the literature to understand farmer decisions and behavior: the Theory of Planned Behavior/the Reasoned Action Approach (TPB/RAA), and the Diffusion of Innovations, or adoption-diffusion, model. I also specifically consider the role of knowledge and especially the biophysical environment in farmer decision-making, which represent two domains that, while not wholly absent from the approaches just mentioned, are extremely underutilized in the literature. In my analyses, I specifically work to integrate social and biophysical elements in the same analyses in a way that is rarely seen especially in the sociological literature. Conceptually, my dissertation is informed and organized by a theory of action, in which purposive individual actions all together create system-level effects. This is done by analyzing data at the individual scale as well as at the county and state scale, as appropriate, across my three empirical chapters. WHY AGRICULTURE? Agricultural systems sit right at the interface between society and the biophysical environment—they are rooted (literally) in chemical and ecological processes that exist independently of humans, and at the same time are highly “socialized” systems created by and for human purposes. As such agricultural systems are shaped by both “nature” and society, thus to understand the relationship between agriculture and society the biophysical world must also be considered (Carolan 2005; Dunlap and Martin 1983). The “recognition of the fact that physical environments can influence (and in turn be influenced by) human societies and behavior” (Dunlap and Catton 1979:244) is the founding 3 principle of environmental sociology. However, just how these reciprocal influences should be conceptualized and operationalized in research has often been a source of disagreement within the discipline—as an example, the social constructionist vs realist debate is/was at its core a disagreement over this question (see York and Dunlap 2011). The tendency thus far has been to take a biophysical problem as the context of a study, with the biophysical problem as an abstract outcome of the human activities being studied, but not to include biophysical problems or conditions as an influence on the human/social actions. This has resulted in research that retains social-social relationships as its primary interest. Additionally, in spite of agriculture being one of the most immediate and tangible ways that societies interact with their biophysical environment (Dunlap and Martin 1983), agriculture has not generally been a topic of much interest to environmental sociology. Agriculture has been a common subject of rural sociologists, particularly those who study the sociology of agriculture though they have not always given as much attention to the biophysical context of agriculture as some have thought they should (Dunlap and Martin 1983). Natural resource sociologists and social scientists have perhaps given agriculture and its issues related to sustainability and resource conservation the most attention (Field, Luloff, and Krannich 2002), though largely in an applied way and, like environmental sociology, including environmental problems as the justification for a study but not incorporating them into the research question or design. WAYS TO UNDERSTAND FARMER-DECISION MAKING Two primary conceptual approaches have been used to explain farmers’ adoption of BMPs: the adoption-diffusion model, and social-psychological theories in the Theory of Planned Behavior tradition. While both approaches have been used with apparent success to better 4 understand the factors influencing farmers’ decisions to use or not use BMPs, they are both, to different degrees, limited in how they account for the external social and biophysical factors that are the context in which farmers’ make their management decisions. They both also have other limitations, which I discuss below. Adoption-Diffusion Model The adoption-diffusion model describes the process through which a new idea, technology, or practice (i.e., innovation) spreads through a social system, and is ultimately adopted (or not) by individuals over time (Rogers 2003). The study of innovation diffusion was originally developed independently in several social science fields in the first half of the twentieth century, including by rural sociologists to understand the spread and adoption of new agricultural technologies (Rogers 1958; Ruttan 1996; Ryan 1948; Ryan and Gross 1943). These traditions merged somewhat in the 1960s, through work by Katz and colleagues (Katz 1963; Katz, Levin, and Hamilton 1963), and that of Everett Rogers (Rogers 2003). Rogers, though not the creator of adoption-diffusion research1, is perhaps the single person most associated with it, through his work in rural sociology and latter in communications (Rogers 2003). Rogers (2003:5) describes diffusion as “the process in which an innovation is communicated through certain channels over time among the members of a social system.” It is a macro process (as opposed to the micro-level adoption decision) that is heavily dependent on communication, and where the newness of the innovation and its effects creates uncertainty about the innovation (Rogers 2003). Over time the innovation is communicated to more people and uncertainty is reduced through greater information, sometimes resulting in the full adoption of the innovation in the population (Rogers 2003). The result of diffusion (and the adoption 1 Rogers calls his model the Diffusion of Innovation Model. 5 process that results) makes it a form of social change, in that it can alter the structure and function of the social system (Rogers 2003). As diffusion happens, individuals (or other decision-making entities) gain information about the innovation and ultimately decide to use it or not use it. The “innovation-decision process,” as described by Rogers (2003:14), “is essentially an information-seeking and information-processing activity in which an individual is motivated to reduce uncertainty about the advantages and disadvantages of the innovation.” Rogers (2003) breaks the innovation- decision process into five stages (Rogers 2003): 1) Knowledge of the innovation—existence of the innovation and how it works. 2) Persuasion of the usefulness/benefit or drawbacks of the innovation—formation of positive or negative attitudes about the innovation. 3) Decision to adopt or reject the innovation based on activities that provide additional information, such as using it on a trial basis. 4) Implementation of the innovation. 5) Confirmation, or reassessment, of the decision made based on additional information. Rogers also describes five perceived attributes of innovations that influence how quickly the diffusion and adoption processes takes for a given innovation (Rogers 2003). These attributes are: 1) Relative advantage of the innovation over current options on any of a number of dimensions (e.g., economic, social, convenience, etc.) 2) Compatibility of the innovation with existing systems, beliefs and values. 3) Complexity of the innovation as it pertains to understanding how it works and how to use it. 4) Trialability of the innovation—being able to test it on a limited basis. 5) Observability of the innovation being in use and of its benefits. 6 Low levels of perceived complexity and high levels of the other attributes serve to accelerate the adoption-diffusion process. The time it takes an individual to complete the innovation-decision process and start using an innovation is considered a measure of their “innovativeness,” and has been used to descriptively categorize individuals (Rogers 2003). If innovativeness is considered to be a personality trait it is expected to generally follow the normal distribution in a given population, and as such, adoption of an innovation also follows a generally normal distribution (Rogers 2003). This results in the “S-shaped curve of adoption” when the cumulative frequency of adopters over time is graphed, or a bell-shaped curve results if frequency of new adoption is graphed (Rogers 2003:272). Normality assumptions are based on full adoption, which often does not happen, so index measures of innovativeness are often used to avoid the technical difficulties of having non-exhaustive categories (Rogers 2003; Rollins 1993). Figure 1.1. Adopter categorization on the basis of an adoption frequency distribution (Rogers 2003:351). 7 Rogers (2003) suggests making use of the normal properties of adoption rate by using standard deviation units from the mean to create five categories of innovation adopters: innovators, early adopters, early majority, late majority, and laggards, with early majority and late majority being the categories one standard deviation less than the mean and greater than the mean respectively (see Figure 1.1). Use of the adoption-diffusion approach in rural sociology peaked in the late 1950s, but had declined dramatically by the early 1970s (Ruttan 1996). Ruttan (1996) suggests this decline can be explained by a shift in sociological perspectives of modernity and technology, in which new technologies could decrease autonomy and increase inequalities rather than being a liberator, coupled with a period of agricultural production surpluses that reduced concern with agricultural technology adoption. Other criticisms of the approach seem to be based more on how it has been applied, than critiques of the approach itself, such as a lack of engagement with sociological theory, and pro-innovation biases (Ruttan 1996). While the adoption-diffusion approach has been out of general favor in rural sociology for nearly 50 years, it has not disappeared entirely, and aspects of it, such as the qualities of innovations that influence adoption, are still commonly used and referenced (e.g., Reimer, Weinkauf, and Prokopy 2012; Weber and McCann 2015). The adoption-diffusion model has certainly been considered as an explanation for conservation practices or BMPs, though not recently. The model has been challenged in its applicability to environmentally oriented practices (Pampel and van Es 1977), but has also been found to still be useful for understanding the adoption of these practices with appropriate adjustments for the purpose of the innovation being considered (Taylor and Miller 1978). The most recent form of the adoption-diffusion model (Rogers 2003) certainly is flexible enough to 8 accommodate both economically and environmentally oriented practices if the purpose of the practice is accounted for and appropriate variables are used. For example, the relative advantage of a practice like vegetative buffers might reasonably be expected to lie more in the social realm (such as the farmers’ personal identity, how they see themselves in their community, social and personal norms and values etc.) than a practice like precision fertilizer application that would have direct economic implications in the costs of adoption and the potential savings in more efficient fertilizer use. However, the adoption-diffusion model does have its flaws. A major challenge to fully utilizing the adoption-diffusion model for understanding BMP adoption is that the model is not easily operationalized empirically. The multi-scale nature of the model is one complicating factor, as the information networks and the diffusion process are conceptualized as macro or meso scale processes that are measured over time, while the five stages of the adoption decision take place at the individual level. This is further complicated by most farmer-level survey data being cross-sectional, and thus unsuitable for a true measure of adoption. In addition, there are the characteristics of the innovation that do not exist at either level, being properties of the innovation itself. It is important to note that the characteristics of a practice may have a subjective component at the individual level, especially for the relative advantage, compatibility, and complexity attributes. Two obvious things that the adoption-diffusion model is missing are an explicit inclusion of the biophysical and the political-economic contexts in which the innovation is being considered. The biophysical context is especially important when considering agricultural practices since the outcome of their use (and its success or failure to achieve the expected or desired result) will depend at minimum on environmental conditions like the soil and climate in 9 which it is being used. The political-economic context also potentially influences several aspects of the adoption-diffusion model, such as information sources, and the relative advantage of the innovation. The rise of large, integrated agricultural input suppliers that provide crop consulting services, and decreases in funding to agricultural extension has shifted to whom farmers turn for information away from extension and towards business affiliated consultants (Stuart et al. 2018), though the implications of this shift for what information farmers are receiving is not yet well understood. Similarly, the relative advantage of a practice would be expected to be influenced by factors such as crop prices, the availability of cost-sharing programs, and policies such as the Conservation Compliance provision discussed above. Social Psychological Theories With the adoption-diffusion model out of favor for studying agricultural decision- making, the TPB family of theories from social psychology has become a very common conceptual approach for understanding farmer adoption of conservation practices. This family consists of the theory of reasoned action (TRA), the theory of planned behavior (TPB), and the reasoned action approach (RAA) (Ajzen 1991; Fishbein and Ajzen 2015). The use of the TPB for understanding agricultural decisions started as early as the mid-1990s (Lynne et al. 1995) currently quite popular (Arbuckle and Roesch-McNally 2015; Borges, Tauer, and Lansink 2016; Van Hulst and Posthumus 2016; Price and Leviston 2014; Reimer et al. 2012; Werner et al. 2017). The three theories in this family are related and began with the theory of reasoned action (TRA). In the TRA attitude towards the behavior and subjective norms—the perceived expectations of others–drive behavioral intention, in turn drives behavior (see Figure 1.2). The use of behavioral intention as a mediator between attitudes, norms and behavior is a 10 distinguishing feature the TRA and the theories derived from it. The assumption is that intention leads to action, though there is evidence that this is not always the case in critical ways (Niles, Brown, and Dynes 2016). The theory of planned behavior (TPB) is an extension of the TRA that adds perceived behavioral control (PBC) as a predictor of behavioral intention and actual behavior (Madden et al. 1992 highlight the differences). Fishbein and Ajzen (2015:64) define PBC “as people’s perceptions of the degree to which they are capable of, or have control over, performing a given behavior.” This addition incorporates a contextual element to the theory and improves its predictive power, especially for behaviors that have greater actual or perceived external constraints (Ajzen 1991; Madden et al. 1992). Solid line = typical Theory of Reasoned Action (TRA) Solid line + dotted line = TRA with additional elements included by Kaiser, Wolfing, and Fuhrer (1999) Solid line + dashed line = Theory of Planned Behavior Figure 1.2. Conceptual diagram of the Theory of Reasoned Action and the Theory of Planned Behavior based on figures in Kaiser, Wölfing, and Fuhrer (1999:3) and Madden, Ellen, and Ajzen (1992:4). 11 Figure 1.3. Conceptual diagram of the Reasoned Action Approach based on figure in Fishbein and Ajzen (2015:22). The reasoned action approach (RAA) (see Figure 1.3) is the most recent incantation of this family of theories and extends the TPB by adding “individual,” “social,” and “information” background factors, along with a new tier of belief variables and an actual control variable (Fishbein and Ajzen 2015). The belief components underlie the attitude, subjective norms and perceived behavioral control variables, and typically measures of the different types of beliefs that are used to measure attitudes, subjective norms and perceived behavioral control using factor analysis (Fishbein and Ajzen 2015). Fishbein and Ajzen (2015:96, 223) formally describe beliefs “as the subjective probability that an object has a certain attribute” and more casually as “the information they [the person] have about a behavior.” These beliefs can be formed through a range of experiences, education, and information sources, and can be short or long lasting, with 12 the only conceptual requirement being that actions follow reasonably from beliefs, no matter how acurate or rational the beliefs may be (Fishbein and Ajzen 2015)2. Similar to the adoption-diffusion model, I see notable conceptual limitations in the TPB/RAA in that there is no conceptual location for political-economic or biophysical contexts3 that may influence the decision process. Behavioral beliefs and PBC broadly might be able to accommodate some aspects of these contexts, such as beliefs in practice attributes and suitability, and adoption costs, but a closer look reveals these measures to be incompatible with the attitudes-measured-by-beliefs approach as discussed below, and with the theoretical conceptualization of PBC (Fishbein and Ajzen 2015), as whether costs constrain behavior is a complex and context specific condition that is heavily dependent on trade-offs and what alternatives are available. At best, the TPB/RAA offers indirect ways of incorporating political- economic or biophysical contexts that are entirely dependent on the individual’s perceptions of them. Of greater concern, is that the TPB/RAA does not account for the nature or characteristics of the behavior under investigation. More importantly, because the TPB/RAA was intended to predict the voluntary behavior of individuals in their everyday life, it may not be appropriate for studying farmers’ decisions about their farming practices. The key distinction is that farming is a business, making it fundamentally different from use of public transportation 2 “It is important to note that, within our reasoned action framework, we do not assume that people are rational but only that their actions follow reasonably from their beliefs. Given that beliefs are often based on information provided by others and on fallible inference processes, behavioral, normative, and control beliefs need not be veridical. They can be inaccurate, biased to conform with preconceptions or motives, or they may represent rationalizations, wishful thinking, or other irrational processes. Nevertheless, the beliefs people hold constitute the information they have about a behavior, and because they naively assume that their beliefs are valid . . . , they act upon them.” (Fishbein and Ajzen 2015:223) 3 The “background factors” as described by Fishbein and Ajzen (2015:24) are all specific to the individual and could include: “age, gender, ethnicity, socioeconomic status, education, nationality, religious affiliation, personality, mood, emotion, general attitudes and values, intelligence, group membership, past experiences, exposure to information, social support, and coping skills.” Fishbein and Ajzen suggest that the selection of background factors can be guided by domain-specific theories. 13 (Heath and Gifford 2002), hotel choice (Han, Hsu, and Sheu 2010), hunting (Hrubes, Ajzen, and Daigle 2001), and the use of dual-flush toilets to save water (Lam 2006) as a few examples. For farmers to remain farmers they must remain profitable, so any decisions about the methods and tools they will use to farm must meet this basic standard. This is not to say that all or even most farmers are profit-maximizers, but their farming decisions have to at least not cost them too much money compared to what the decision will make them. This frequently places farmers in a situation where they could implement a given practice, but they have to decide if they should implement it, thereby greatly complicating the PBC portion of the TPB/RAA, and potentially curtailing the role of attitudes and norms in their decisions. I also see empirical challenges with operationalizing the TPB/RAA, though these challenges are different than the ones for the adoption-diffusion model as the TPB/RAA already has a path model that is easily estimated through structural equation modeling (SEM). The main issue involves the potential conflict between the conceptualization and the common operationalization of beliefs and attitudes. Conceptually in the RAA beliefs are separate from and predict attitudes, but empirically it is most common to use beliefs as measures of attitudes in a latent variable. The measurement model of a latent variable uses the observed attitudes as endogenous variables that are predicted by the latent construct. In other words, the attitude predicts the beliefs, which is the opposite direction of the relationship as depicted in the RAA. This results in a model in which the belief variables have no direct or indirect effect on the behavior. It is possible to operationalize the model differently, as Borges, Tauer, and Lansink (2016) have done using a multi indicators and multiple causes (MIMIC) model4, which maintains the directions of the causal relationships in the TPB/RAA, but this is not a common method. 4 See Kline (2012) for more on MIMIC models. 14 OVERVIEW OF EMPIRCAL ANALYSES This dissertation includes three empirical analyses. In the first analysis I use qualitative data and grounded theory to investigate farmers’ tillage practices and the factors that influence their tillage choices. My primary purpose is to examine what tillage practices Midwestern corn farmers are using and why. The results that emerge on the ways farmers are tilling challenges the commonly used tillage categories, as I find that farmers’ tillage practices are not easily assigned to the usual three or four categories. The results of the reasons farmers gave for why they till as they do calls into question the social-psychological theories that are commonly used in studies of farmer decision-making, as I find that the biophysical context, the needs of the operation, and practice compatibility were much more heavily referenced by farmers as reasons for their tillage choices than were attitudes or values related to the environment or the tillage practices themselves. The second analysis is a quantitative analysis using structural equation modeling with latent variables that takes a broad approach to investigating the drivers of the use of five BMPs: pre-sidedress nitrate tests (PSNTs), nutrient maps, variable rate nitrogen (N), variable rate phosphorus (P) or potassium (K), and cover crops. The goal is to identify the common drivers of BMP use and assess their relative importance. It also explores the use of practice knowledge as a predictor of BMP use that has been underutilized in previous research. The results of this analysis show that these five practices have few drivers in common, even between practices that were expected to be similar. The relative importance of the significant variables suggests that the adoption-diffusion model may provide more predictive of farmers’ use of these BMPs than the social-psychological theories, as information sources were found to be more important to practice use than attitudes. 15 The third analysis uses county and state data in a multilevel structural equation model to examine the scaled-up effects of many individual decisions relating to the rate of nitrogen fertilizer used in the presence of cover crops, conservation tillage and no-till. The retention of nutrients by these in-field practices means that they have the potential to reduce the amount of fertilizer that a farmer needs to apply to get a good crop, thereby saving farmers money and reducing non-point-source nutrient pollution from agriculture. However, it is not clear to just what extent farmers actually do reduce their fertilizer use when they adopt any of these practices. If farmers are not experiencing tangible savings from the use of these practices then it is likely that the present incentive system for their use is insufficient for encouraging their adoption. The analysis does not show any of the three practices to have a significant total effect on fertilizer use, but it does suggest that cover crops may have the potential to increase fertilizer use. Not only does this analysis scale up from the individual to better assess the cumulative effects of many individual actions at a scale where the effects are more visible, but it also includes a mixture of ecological and economic and non-economic social variables in a single model, which is novel in sociological studies. 16 REFERENCES 17 REFERENCES Ajzen, Icek. 1991. “The Theory of Planned Behavior.” Orgnizational Behavior and Human Decision Processes 50:179–211. Aneja, Viney P., William H. Schlesinger, and Jan Willem Erisman. 2009. “Effects of Agriculture upon the Air Quality and Climate: Research, Policy, and Regulations.” Environmental Science & Technology 43(12):4234. Arbuckle, J. G. 2013. “Farmer Support for Extending Conservation Compliance beyond Soil Erosion: Evidence from Iowa.” Journal of Soil and Water Conservation 68(2):99–109. Arbuckle, J. G. and G. Roesch-McNally. 2015. “Cover Crop Adoption in Iowa: The Role of Perceived Practice Characteristics.” Journal of Soil and Water Conservation 70(6):418–29. Blesh, J. and L. E. Drinkwater. 2013. “The Impact of Nitrogen Source and Crop Rotation on Nitrogen Mass Balances in the Mississippi River Basin.” Ecological Applications 23(5):1017–35. Borges, João Augusto Rossi, Loren Willian Tauer, and Alfons G. J. M. Oude Lansink. 2016. “Using the Theory of Planned Behavior to Identify Key Beliefs Underlying Brazilian Cattle Farmers’ Intention to Use Improved Natural Grassland: A MIMIC Modelling Approach.” Land Use Policy 55:193–203. Carolan, Michael S. 2005. “Society, Biology, and Ecology: Bringing Nature Back Into Sociology’s Disciplinary Narrative Through Critical Realism.” Organization & Environment 18(4):393–421. Carpenter, S. R., N. F. Caraco, D. L. Correll, R. W. Howarth, A. N. Sharpley, and V. H. Smith. 1998. “Nonpoint Pollution of Surface Waters with Phosphorus and Nitrogen.” Ecological Applications 8(3):559–68. Drinkwater, L. E. and S. S. Snapp. 2007. “Nutrients in Agroecosystems: Rethinking the Management Paradigm.” Advances in Agronomy 92:163–86. Dunlap, Riley E. and William R. Catton. 1979. “Environmental Sociology.” Annual Review of Sociology 5:243–73. Dunlap, Riley E. and Kenneth E. Martin. 1983. “Bringing Environment into the Study of Agriculture: Observations and Suggestions Regarding the Sociology of Agriculture.” Rural Sociolgy 48(2):201–18. EPA. 2017. National Water Quality Inventory: Report to Congress. EPA 841-R-16-011. U.S. Environmental Protection Agency. Retrieved April 22, 2018 (https://www.epa.gov/sites/production/files/2017- 12/documents/305brtc_finalowow_08302017.pdf). 18 Field, Donald R., A. E. Luloff, and Richard S. Krannich. 2002. “Revisiting the Origins of and Distinctions Between Natural Resource Sociology and Environmental Sociology.” Society & Natural Resources 15(3):213–27. Fishbein, Martin and Icek Ajzen. 2015. Predicting and Changing Behavior: The Reasoned Action Approach. New York, NY: Routledge. Halcrow, Harold G., Earl O. Heady, and Melvin L. Cotner, eds. 1982. Soil Conservation Policies Institutions and Incentives. Ankeny, IA: Soil Conservation Society of America. Hamilton, Stephen K. 2015. “Water Quality and Movement in Agricultural Landscapes.” Pp. 275–309 in The Ecology of Agricultural Landscapes: Long-Term Research on the Path to Sustainability. New York: Oxford University Press. Han, Heesup, Li-Tzang (Jane) Hsu, and Chwen Sheu. 2010. “Application of the Theory of Planned Behavior to Green Hotel Choice: Testing the Effect of Environmental Friendly Activities.” Tourism Management 31(3):325–34. Heath, Yuko and Robert Gifford. 2002. “Extending the Theory of Planned Behavior: Predicting the Use of Public Transportation.” Journal of Applied Social Psychology 32(10):2154–89. Hrubes, Daniel, Icek Ajzen, and John Daigle. 2001. “Predicting Hunting Intentions and Behavior: An Application of the Theory of Planned Behavior.” Leisure Sciences 23:165–78. Van Hulst, Freddy J. and Helena Posthumus. 2016. “Understanding (Non-) Adoption of Conservation Agriculture in Kenya Using the Reasoned Action Approach.” Land Use Policy 56:303–14. Kaiser, F. G., S. Wölfing, and U. Fuhrer. 1999. “Environmental Attitude and Ecological Behavior.” Journal of Environmental Psychology 19:1–19. Katz, Elihu. 1963. “Communication Research and the Image of Society Convergence of Two Traditions.” American Journal of Sociology 65(5):435–40. Katz, Elihu, Martin L. Levin, and Herbert Hamilton. 1963. “Traditions of Research on the Diffusion of Innovation.” American Sociological Review 28(2):237–52. Kline, Rex B. 2012. “Assumptions in Structural Equation Modeling.” Pp. 111–25 in Handbook of Structural Equation Modeling, edited by R. H. Hoyle. New York: The Guilford Press. Lal, Rattan. 1993. “Tillage Effects on Soil Degradation, Soil Resilience, Soil Quality, and Sustainability.” Soil and Tillage Research 27(1–4):1–8. Lam, San-Pui. 2006. “Predicting Intention to Save Water: Theory of Planned Behavior, Response Efficacy, Vulnerability, and Perceived Efficiency of Alternative Solutions.” Journal of Applied Social Psychology 36(11):2803–24. Lynne, Gary D., C. Franklin Casey, Alan Hodges, and Mohammed Rahmani. 1995. 19 “Conservation Technology Adoption Decisions and the Theory of Planned Behavior.” Journal of Economic Psychology 16(4):581–98. Madden, Thomas J., Pamela Scholder Ellen, and Icek Ajzen. 1992. “A Comparison of the Theory of Planned Behavior and the Theory of Reasoned Action.” Personality and Social Psychology Bulletin 18(1):3–9. Niles, Meredith T., Margaret Brown, and Robyn Dynes. 2016. “Farmer’s Intended and Actual Adoption of Climate Change Mitigation and Adaptation Strategies.” Climatic Change 135(2):277–95. NRCS. n.d. “Conservation Compliance.” Natural Resources Conservation Service, U.S. Department of Agriculture. Retrieved April 22, 2018 (https://www.nrcs.usda.gov/wps/portal/nrcs/detailfull/national/programs/farmbill/?cid=stelp rdb1257899). Pampel, Fred and J. C. van Es. 1977. “Environmental Quality and Issues of Adoption Research.” Rural Sociolgy 42(1):57–71. Price, Jennifer C. and Zoe Leviston. 2014. “Predicting Pro-Environmental Agricultural Practices: The Social, Psychological and Contextual Influences on Land Management.” Journal of Rural Studies 34(0):65–78. Reimer, Adam P. and Linda S. Prokopy. 2014. “Farmer Participation in US Farm Bill Conservation Programs.” Environmental Management 53(2):318–32. Reimer, Adam P., Denise K. Weinkauf, and Linda S. Prokopy. 2012. “The Influence of Perceptions of Practice Characteristics: An Examination of Agricultural Best Management Practice Adoption in Two Indiana Watersheds.” Journal of Rural Studies 28(1):118–28. Ribaudo, Marc. 2015. “The Limits of Voluntary Conservation Programs.” Choices: The Magazine of Food, Farm & Resource Issues 30(2):1–5. Robertson, G. Philip Tom W. Bruulsema, Ron J. Gehl, David Kanter, Denise L. Mauzerall, C. Alan Rotz, and Candiss O. Williams. 2013. “Nitrogen–climate Interactions in US Agriculture.” Biogeochemistry 114(1–3):41–70. Robertson, G. Philip and Peter M. Vitousek. 2009. “Nitrogen in Agriculture: Balancing the Cost of an Essential Resource.” Annual Review of Environment and Resources 34(1):97–125. Rogers, Everett M. 1958. “Categorizing the Adopters of Agricultural Practices.” Rural Sociology 23(4):346–54. Rogers, Everett M. 2003. Diffusion of Innovations. Fifth. New York, NY: Free Press. Rollins, Timothy. 1993. “Using The Innovation Adoption Diffusion Model To Target Educational Programming.” Journal of Agricultural Education 34(4):46–54. 20 Ruttan, Vernon W. 1996. “What Happened to Technology Adoption-Diffusion Research?” Sociologia Ruralis 36(1):51–73. Ryan, Bryce. 1948. “A Study in Technological Diffusion.” Rural Sociology 13(3):273–85. Ryan, Bryce and Neal C. Gross. 1943. “The Diffusion of Hybrid Seed Corn in Two Iowa Communities.” Rural Sociology 8(1):15–24. Stuart, D., R. C. H. Denny, M. Houser, A. P. Reimer, and S. Marquart-Pyatt. 2018. “Farmer Selection of Sources of Information for Nitrogen Management in the US Midwest: Implications for Environmental Programs.” Land Use Policy 70:289–97. Taylor, David L. and William L. Miller. 1978. “The Adoption Process and Environmental Innovations: A Case Study of a Government Project.” Rural Sociology 43(4):634--648. Uri, Noel D. and James A. Lewis. 1998. “The Dynamics of Soil Erosion in US Agriculture.” Science of the Total Environment 218(1):45–58. Weber, Catharine and Laura McCann. 2015. “Adoption of Nitrogen-Efficient Technologies by U.S. Corn Farmers.” Journal of Environment Quality 44(2):391–401. Werner, Magdalena, Erwin Wauters, Jo Bijttebier, Horst-Henning Steinmann, Greet Ruysschaert, and Andrea Knierim. 2017. “Farm Level Implementation of Soil Conservation Measures: Farmers’ Beliefs and Intentions.” Renewable Agriculture and Food Systems 1– 14. York, Richard and Riley E. Dunlap. 2011. “Environmental Sociology.” Pp. 504–21 in The Wiley- Blackwell Companion to Sociology. Hoboken, NJ: John Wiley & Sons. 21 CHAPTER 2: MOVING BEYOND BASIC TILLAGE CATEGORIES IN MIDWEST ROW-CROP AGRICULTURE: AN EXPLORATORY ANALYSIS OF TILLAGE PRACTICES AND REASONS Extensive research and outreach efforts have focused on encouraging farmers to use more conservation-oriented practices to reduce the environmental impacts of agriculture (Aneja, Schlesinger, and Erisman 2009; Carpenter et al. 1998). Therefore, there is also great interest in better understanding why farmers choose to use or not use these practices and what incentives can encourage their voluntary use (Baumgart-Getz, Prokopy, and Floress 2012; Knowler and Bradshaw 2007; Liu, Bruins, and Heberling 2018; Prokopy et al. 2008; Wauters and Mathijs 2014). The use of “conservation” tillage has long been a practice of interest in efforts to reduce soil erosion in US agriculture (Allmaras and Dowdy 1985; Halcrow, Heady, and Cotner 1982). However, the sociology of agriculture and the agriculturally oriented natural resource literature have long tended to oversimplify the ways that tillage is used and measured, particularly by ignoring the spatial and temporal aspects of tillage decisions (Reimer et al. 2014), though soil scientists are aware of the importance of crop rotation and soil type/qualities in the use of specific tillage practices (e.g., Reicosky and Allmaras 2003; Lal, Reicosky, and Hanson 2007). The over simplification of farmers’ tillage practices and decision-making process related to tillage has been exacerbated by the common division of tillage practices into three categories: “conventional” tillage, “conservation” or “reduced” tillage, and “no-till” and the use of binary measures of practice use/disuse, without consideration of duration, frequency, or area of use (Reimer et al. 2014). Furthermore, the use of these overly general categories of tillage mean that these binary measures may not be as accurate as desired. Reimer et al. (2014:57A) point out this problem: 22 Conservation tillage, for example, can mean continuous no-till to one farmer, strip-till after soybeans (Glycine max [L.] Merr.) but not after corn (Zea mays L.) to another, and use of a chisel plow rather than moldboard plow to another. Such variance in meanings and behavior can lead to substantial measurement error associated with discrete, binary measures of adoption. Categories are important and useful research tools, but too much aggregation can disguise important differences. Soil scientists Reicosky and Allmaras (2003) point out that terms like “conventional,” “conservation” and “reduced” are all broad and relative, and often have different regional meanings. Additionally, as tillage practices have shifted in a more conservation direction over time (Lal et al. 2007), and with the advent of new tillage technology, the meanings of these terms will also have inevitably changed (Reicosky and Allmaras 2003)5. My primary purpose for this exploratory analysis using qualitative interview data is to examine what tillage practices Midwestern corn farmers are using and why. The results that emerge on the ways farmers are tilling challenges the commonly used tillage categories, as I find that farmers’ tillage practices are not easily assigned to the usual three or four categories. The results of the reasons farmers gave for why they till as they do calls into question the social- psych theories that are commonly used in studies of farmer decision-making, as I find that the biophysical context, the needs of the operation, and practice compatibility were much more heavily referenced by farmers as reasons for their tillage choices than were attitudes or values related to the environment or the tillage practices themselves. 5 “As new technology develops and new tillage and planting equipment become available, we should discontinue the use of the vague and nondescript terms of “conventional” and “conservation” terminology and provide explicit descriptions of equipment for tillage, residue management and planting. … An accurate description requires listing all operations in the system and should be the prime consideration in discussing various conservation tillage systems” (Reicosky and Allmaras 2003:77). 23 BACKGROUND ON TILLAGE Tillage has long been an important agricultural practice. Reicosky and Allmaras (2003:76) describe tillage as “the mechanical manipulation of the soil and plant residue to prepare a seedbed where crop seeds are planted… [it] fragments soil, enhances the release of soil nutrients for crop growth, kills weeds, and modifies the circulation of water and air within the soil.” Early wooden hand tillage implements facilitated the rise of settled agriculture and their improvement into what we would recognize today as a “plow” that could be pulled by draft animals around 4000 to 6000 BC increased food production and supported the rise of urban- oriented civilizations in Mesopotamia (Lal et al. 2007). Further improvements over time included the use of metal plow shares and shaping the share to invert the soil, rather than just break and loosen it, the hallmark of the modern “moldboard” plow (Lal et al. 2007). As a result, the most common tillage practice in the US for several hundred years was moldboard plowing followed by “secondary” tillage, which breaks up the soil more finely, and levels and smooths the soil in preparation for planting (Belknap and Saupe 1988; Bultena and Hoiberg 1983; Reicosky and Allmaras 2003). While soil erosion from plowing has been a concern for hundreds if not thousands of years, it took on new urgency in the US following the Dust Bowl of the 1930s and resulted in policies and programs to address it, such as the establishment of the Soil Erosion Service in 1933, and the passage of the Soil Conservation Act in 1935 (Lal et al. 2007; Rasmussen 1982). Addressing soil erosion meant addressing tillage practice, and led to the challenging of moldboard plowing and of plowing at all in favor of “conservation” tillage and “no-till” practices during the 1940s-1960s (Lal et al. 2007). New technologies facilitated this challenge of plowing, especially new herbicides that could replace tillage for weed management, and the development 24 of implements for planting into untilled ground (Allmaras et al. 1998; Coughenour 2003; Lal et al. 2007; Reicosky and Allmaras 2003). However, the use of moldboard plows nationally did not start to decrease until the 1970s (Bultena and Hoiberg 1983; Crosson 1981; Reicosky and Allmaras 2003). In the 1980s, continued concern over high rates of soil erosion (Crosson 1984) and relatively low rates of adoption of non-moldboard tillage led to a large body of research on the adoption process of “conservation tillage” by rural sociologists (Bultena and Hoiberg 1983; Dillman and Carlson 1982; Korsching et al. 1983; Napier, Thraen, and Gore 1984; Nowak 1987; Nowak and Korsching 1985; Warriner and Moul 1992) and agricultural economists (Belknap and Saupe 1988; Ervin and Ervin 1982; Gould, Saupe, and Richard 1989; Lee and Stewart 1983; Norris and Batie 1987; Rahm and Huffman 1984; Shortle and Miranowski 1986) during this decade. Research during the 1980s mostly focused on the adoption of “conservation tillage” which was most commonly defined as any tillage system that did not involve the use of a moldboard plow (e.g., Belknap and Saupe 1988; Bultena and Hoiberg 1983). More broadly, “conservation tillage” has been defined as “a tillage system in which either crop residues are retained on or near the surface, or soil surface roughness is maintained, or both, to control soil erosion and to achieve good soil-water relations” (Allmaras and Dowdy 1985:197-198). However, a 30 percent remaining soil cover standard has been associated with “conservation” tillage since the 1980s (Allmaras and Dowdy 1985), and continues to be used by the Conservation Technology Information Center (CTIC, formerly the Conservation Tillage Information Center), with “reduced” tillage being defined as retaining 15-30 percent soil cover. These percentage soil cover standards have been regularly used in social research on tillage practice adoption (e.g., Reimer, Weinkauf, and Prokopy 2012; Vitale et al. 2011; Caswell et al. 25 2001). The retention of 30 percent or more residue can be achieved by a wide range of tillage methods from “one fall chisel with straight shanks, a shallow disking in the spring, a field cultivation and planting” to a full no-till system (NRCS 1992). How “conservation tillage” is defined and operationalized is important because different environmental implications that result from the use of the range of practices that are often included in the “conservation tillage” category. For example, Kumar et al. (2014) compare soil organic carbon levels in drained and undrained field under no-till and chisel plowing on continuous corn in Ohio, and find significantly more SOC in the no-till plots compared to the chisel plowed plots after 18 years. Similarly, López-Fando and Pardo (2012:33) compare total nitrogen content and nitrogen stratification ratios between “no-tillage (NT), minimum tillage (18–22 cm depth) with chisel plow (MT), and conventional tillage (25–30 cm depth) with mouldboard plow (CT)” over 19 years of a chickpea-barley rotation in central Spain. They find significant differences between the no-tillage, minimum tillage and conventional tillage systems, with no-tillage having higher total nitrogen and nitrogen stratification than minimum tillage which in turn was higher than conventional tillage, and they conclude that these measures indicate better soil properties particularly under no-tillage as compared to minimum tillage and conventional tillage. Greater soil fertility, in the form of nitrogen and carbon content, resulting from tillage practices on the no-till end of the spectrum have the potential to help mitigate global climate change by storing carbon in the soil (Robertson et al. 2014; Robertson, Paul, and Harwood 2000; Uri and Bloodwirth 2000). 26 LITERATURE REVIEW Many of the 1980s studies on the adoption of “conservation” tillage as well as more recent ones have been reviewed in meta-analyses of the predictors of “best management” or “conservation practices” not limited to tillage (Baumgart-Getz et al. 2012; Knowler and Bradshaw 2007; Liu et al. 2018; Prokopy et al. 2008; Wauters and Mathijs 2014). However, these analyses find mixed or inconclusive results of predictors, even when potentially focused on tillage practices. For instance, Prokopy et al. (2008) include sub-analyses based on practice type, including one for “soil management” which is presumably some form of “conservation tillage,” though they do not define what they include in this category. They find that, only acres farmed is significant more often than it is not significant as a predictor of “soil management” practices. One reason for this lack of consistency may be the measurement error pointed out by Reimer et al. (2014). Recent quantitative studies (Andrews et al. 2013; Ryan, Erickson, and De Young 2003; Vitale et al. 2011) on factors related the adoption/use of “conservation” tillage and no-till in the US have acknowledged the potential for multiple meanings of tillage practice terms, and some have made efforts to limit measurement error by providing respondents with definitions of the response categories. However, these studies, while recognizing that farmers use tillage in complex ways, limited their analyses to overly simple categories of tillage use (Andrews et al. 2013; Ryan et al. 2003; Vitale et al. 2011). These quantitative studies featured beliefs and attitudes but did not use any of the theories discussed in the introduction (Chapter 1). Ryan, Erickson, and De Young (2003) considered the likelihood of farmers in a Michigan watershed adopting one of several specific practices, including no-till, along streams and drains in the future. They also considered farmers motivations for adopting these practices and 27 considered six attitude and value-based variables (three of which are latent variables), though they did not draw explicitly on any specific theoretical approach. Twenty-seven percent of their respondents reported using “no-till farming to minimize soil erosion” and farmers rated no-till as the practice category they were most likely to adopt in the future (average score of 3.6 out of 5)6. However, the authors noted that “farmers may conceptualize no-till farming as a range of conservation tillage practices, such as minimum tillage” (Ryan, Erickson, and De Young 2003:28). They found that full time farmers were more likely to adopt no-till than part time farmers, and that attachment to their farm, intrinsic environmental motives, and the visual recognition of no-till were important factors in the reported likelihood of future no-till adoption. Vitale et al. (2011) used a survey to study the adoption of three categories of tillage practices, conventional, reduced, and conservation tillage, among Oklahoma farmers. They took an atheoretical econometric approach that focused on the individual decision process and used three categories of “perception” variables related to conservation tillage: economic, agronomic, and other. The authors provided farmers with the CTIC definitions of each practice (see Figure 2.1). In their descriptives they differentiated between farmers who reported using only one type of tillage and those who reported using the type on at least 50 percent of their operation and one or more other tillage types on their remaining acres; 36.8 percent of their respondents fell into one of these groups or did not use any one type of tillage on a majority of their acres (calculated from authors descriptive table). However, in their regression models the authors only used the typical categories. Their results highlight the importance of farmer age, farm size, crop rotation, presence of livestock, perceptions of the benefits and costs of different tillage practices, and knowledge of conservation tillage practices in explaining farmers’ dominant tillage type. 6 The highest rated individual practice was “maintain a grassy buffer strip/filter strip” with average rating of 3.86, however the authors used factor analysis to group similar practices into categories. No-till was used as an individual item in their analysis as it did not fit into any calculated factor. 28 Figure 2.1. Tillage definitions used by Vitale et al. (2011:252). Figure 2.2. Tillage definitions used by Andrews et al. (2013:504). Andrews et al. (2013) asked a national sample of farmers about their tillage practices in 2009 and 2010 on three crops (see Figure 2.2) and reasons for using them as part of a survey on framing effects of conservation tillage on farmers’ interest in using conservation tillage and in receiving more information about it. They considered four frames: a basic rationale that focuses on soil quality benefits and erosion reduction (the control frame), a profit frame that focuses on the money saving potential of conservation tillage, a carbon offset frame that focuses on the 29 potential for getting carbon offset payments for using conservation tillage, and a payment for environmental benefits frame that focuses on the potential for payments for using conservation tillage based on the off-farm environmental benefits. How the benefits of a practice is presented or “framed” is expected to influence farmers’ attitudes towards the practice and in turn influence their likelihood of using it. Nearly 50 percent of their respondents reported using a combination of tillage types (calculated from authors’ descriptives table). However, the authors combined their respondents into three categories: “conservation tillers” who only used conservation tillage and/or no-till, “combined tillers” who used all three types or a combination of no-till and conventional tillage, and “conventional tillers” who only use conventional tillage or a combination of conventional tillage and conservation tillage. They found that concerns over water quality, soil erosion, improving soil productivity, and lowering costs were important factors for the “conservation tillers” group who were using conservation tillage and/or no-till. In contrast, conventional tillers were highly concerned with high yields, soil compaction/drainage, and earlier plating windows. For the framing treatments, the authors found no significant differences between treatments in the whole sample; they did find that the profit frame significantly reduced “conventional tillers” interest in conservation tillage compared to the basic rationale. The authors hypothesized that this effect could be due to the profit frame highlighting financial considerations that these farmers do not believe can be achieved with conservation tillage, possibly due to local soil qualities and climatic conditions that make conservation tillage practices less advantageous. Of particular interest to the present study are the small number of papers that use qualitative methods to better understand farmers’ use of different tillage practices. Two recent 30 qualitative studies have been conducted in the US on conservation practices that included tillage (Reimer, Thompson, and Prokopy 2012; Reimer, Weinkauf, et al. 2012). Reimer, Thompson, et al. (2012) use 32 semi-structured farmer interviews in an Indiana watershed to examine farmers awareness of environmental issues and attitudes towards conservation practices, as well as their current awareness and use of conservation practices. They defined “conservation tillage” as the use of 100 percent no-till practices on corn, since “nearly all” farmers interviewed were using no-till for their soybeans; 56 percent of their sample was found to practice “conservation tillage” as they defined it. They identified three attitude dimensions in their interview data: “farm as business,” “[concern for] off-farm environment,” and “stewardship.” They found that the farm-as-business attitude was related with low use of conservation practices, while off-farm environmental concern and stewardship attitudes were associated with use of more conservation practices. Reimer, Thompson, et al. (2012) noted that the majority of farmers interviewed in this study were in a middle adopting group, adopting a few conservation practices but not expressing strong off-farm environmental, stewardship, or farm as business attitudes…. This group of farmers talked most extensively about the on-farm benefits and was aware of the indirect financial benefits of erosion control and soil nutrients that certain conservation practices brought. This group was most likely to adopt conservation tillage and grassed waterways where needed. (P. 36) Reimer, Weinkauf, et al. (2012) use the same 32 interviews used in Reimer, Thompson, et al. (2012), as well as an additional 13 farmer interviews in a different Indiana watershed. They use the interviews to examine farmers’ views on the “acceptability characteristics for each practice” (Reimer, Weinkauf, et al. 2012:121), drawing on the diffusion of innovations theory’s five characteristics of innovations that relate to adoption or rejection of a practice along with the reasoned action approach. The authors recognize 8 types of tillage practices among their 31 interview participants, including crop rotation-based differences, though their analysis just refers to “conservation tillage.” About 66 percent of their sample reported using either no-till for soybeans and a different type of tillage for corn, or a combination of tillage types “based on sensitivity of the ground or compatibility with other practices” (Reimer, Weinkauf, et al. 2012:123). They found that soil conservation and input savings were the most important attributes reported that motivated farmers to use conservation tillage, while decreased yield, lack of perceived need, and uncertainty of effectiveness were the most important characteristics that deterred farmers from using conservation tillage. METHODS Overview Interviews with corn farmers were conducted in 2014 as part of a National Science Foundation funded Couple Human and Natural Systems (CHANS) project examining nitrogen efficiency in Midwest corn production. A semi-structured interview guide was used that included questions on general farm characteristics and practices used, sources of information used, influence of crop insurance, conservation programs, private companies and other farmers on farming practices, and views on the environment and climate change. The interview guide had a particular interest in the use of nitrogen fertilizer on corn, and many of the questions were asked in this context. Qualitative methods are particularly well suited to investigating exploratory research questions and gaining a detailed understanding of reasons and motivations. While theories and conceptual frameworks can be very useful research tools, there can also be benefit in collecting farmers’ responses using open-ended questions. As such, qualitative methods are a good choice 32 for approaching my research questions, which are: “what tillage practices do farmers use?” and “why do farmers till the way they do?” Tillage has not been a heavy focus of recent social science research, and the advance of tillage technology and demonstration of mixed tillage use in the literature suggests an exploratory analysis of tillage practices would be worthwhile. Why farmers till the way they do is also a topic that would clearly benefit from an exploratory analysis where farmers’ explanations for their actions could emerge. If the qualitative findings do not support the findings of the previous studies, or only do so in part, then the exploratory analysis will offer suggestions for additional angles of investigation, which can lead to further research questions, hypotheses and theory development. Data Collection The farmer interviews were conducted by three individuals between May and December 2014 in Iowa (IA), Indiana (IN) and Michigan (MI). Interviews lasted between 20 minutes and 3 hours, with a typical interview lasting about 45 minutes. A total of 154 interviews were conducted, with 53 in IA, 51in IN, and 50 in MI7. A small number (fewer than five total) of interviews were conducted by phone and/or not audio recorded; the rest were conducted in- person and were audio recorded with participant permission and subsequently transcribed verbatim. The interviews that were not recorded had notes written up immediately afterwards. The only requirement for inclusion in the sample was that the farmer be growing at least 100 acres of non-organic corn. The final sample of 152 farmers covered 27 percent of IA counties, 25 percent of IN counties, and 20 percent of MI counties8. 7 Two Michigan participants did not meet the sample selection criteria so 48 Michigan interviews are used in the following analysis. 8 Not all MI counties have extensive corn production, especially those in the northern part of the state, which is why the percent of counties covered in MI is lower than in the other states. 33 Farmer recruitment was done primarily through referrals from respective state University Extension personnel and through snowball sampling, though the portions varied by state. Table 2.1 shows the differences in participant recruitment methods by state. In IA, the extension source was through an extension-run program rather than direct referral by extension agents as was the case in IN and MI. In MI the participants recruited through “relevant organizations and programs” were all contacts made through the list of the Michigan Agriculture Environmental Assurance Program participants, while in IA it was through an email to the listserv of Practical Farmers of Iowa. While snowball sampling can help achieve a more diverse sample, the heavy use of extension in the initial participant recruitment is likely to have skewed the sample towards farmers who extension personnel see as being exemplary farmers. This bias may have made the sample more environmentally aware and operationally efficient than is typical of the general farming population. All participants were white men ranging in age from 20s-70s. Female and racial minority farmers were not deliberately excluded from the sample, but due to the focus of the study no special effort was made to include them. Table 2.1. Sample recruitment sources by state. Number of interviews Extension Snowball sampling Other County Soil and Water Conservation District Field days and relevant events Michigan 50† % (#) 64.0 (32) 24.0 (12) 12.0 (6) 0.0 (0) 0.0 (0) Iowa 53 % (#) 22.6 (12) 45.3 (24) 32.1 (17) 22.6 (12) 3.8 (2) Indiana 51 % (#) 58.8 (30) 33.3 (17) 7.8 (4) 0.0 (0) 0.0 (0) Relevant organizations and programs 7.8 (4) † Two Michigan participants did not meet the sample selection criteria so 48 Michigan interviews are used in the following analysis. 12.0 (6) 5.7 (3) 34 The sample is also skewed towards larger farms. Farm sizes in the sample ranged from 170 to 5,000 acres in IA, with an average size of 1,236 acres (average for the state was 3479), from 200 to 9,000 acres in IN, with an average size of 2,216 acres (average for the state was 253), and from 200 to 4,500 acres in MI, with an average farm size of 1,529 acres (average for the state was 193). The larger average farm size in the sample is likely the result of the 100-acre- minimum-corn selection criterion. This requirement automatically restricts the sample to farms greater than the median in MI and IN, and likely functionally does in IA as well (Table 2.2). Since this data is for total operation size, this means that the sample selection was restricted to the largest fifty-percent of farms. It is also plausible that larger farms are more likely to be growing corn. Table 2.2. Mean and median farm size in acres per operation by state in 2012. Mean farm size Median farm size Source: NASS Census of Agriculture. Analysis Michigan Iowa Indiana 191 60 345 136 251 57 The analysis process used was based on Grounded Theory (Glaser and Strauss 1967) in the vein of Glaser’s approach (Glaser 1978, 1992; Walker and Myrick 2006). Grounded theory is the process of developing a theoretical explanation of a social action, interaction or process that is grounded in the analysis of qualitative data (Creswell 2007). It is an inductive, rather than a deductive, process. Briefly, Glaser’s approach involves two stages: “Substantive coding … is concerned with producing categories and their properties. Theoretical coding occurs at the 9 Average farm sizes are for all farms (not corn specific) for the three states are for 2014 from a USDA survey. Data was collected from NASS. 35 conceptual level, weaving the substantive codes together into a hypothesis and theory” (Walker and Myrick 2006:550). The initial substantive coding was done according to the questions on the interview guide. The responses analyzed in this paper were to the question: “How do you till and how does it affect your fertilizer decisions?” The question was alternatively asked in two parts: “How do you till?” and then after their response a follow up: “How does that affect your fertilizer decisions?” The analysis presented here began with coding the tillage question code by tillage tool. Given the large volume of data, even when only considering the tillage question code, the word search query in NVivo 11 (QSR) was used assist in locating all relevant passages. The terms till/tilling/tillage, plow/plowing and work/working were searched, and the relevant instances of them coded according to the tool/type mentioned. Each tool/type code was then searched for specifically to ensure that any instances that appeared without a form of till, plow or work were located. Each tool/type code was then further coded to identify farmers who reported current use of the tool/type and mentions of past use, reasons for use, observations of use by others etc. The tool/type use codes were then further coded by sub-type, season of use, method of use, and/or extent of use as was appropriate for the tool/type. For example, use of harrows was coded for disk harrow use vs. non-disk harrow use, then disk harrow use was coded for use as primary tillage vs secondary tillage, and each of those was further coded for season of use, and method/purpose of use—e.g., based on rotation, or used on “as needed” basis. The content coded to non-use tool/tillage codes as well as the tool/type specific codes related to method/purpose of the specific tool/tillage type were collected and substantively coded for reasons for tillage use, and then to pursue a more detailed set of reasons for each primary 36 reason code. Finally, theoretical coding was used to group these reasons into the explanatory categories discussed in the results section. Quantified results of tillage tool/type were based on the results of using the matrix query tool in NVivo 11. All tillage/type codes were expended as needed so that they overlapped, allowing for the use of multiple tools/types by a single farmer to be identified through matrix queries. RESULTS The first part of this section presents and briefly discusses the results of the substantive coding of tillage tools and tillage types, primarily using the quantified results in Tables 2.3 and 2.4. This portion of the analysis addresses my first research question: How are farmers tilling? My second research question is addressed by the theoretical codes that are discussed in the section on tillage decision factors. As it is not the purpose of this article to describe in detail the characteristics of the specific tillage tools and types, but rather to explore their overlap in use and the reasons for their use, only brief descriptions will be provided as needed. Readers are particularly referred to two publications: Conservation Tillage Systems and Management (MWPS 2000), and Tillage (John Deere 2007) for detailed descriptions, and images of the structure, performance, and options of the tillage tools and types discussed. 37 Table 2.3. Cross-tab comparison of instances of tillage tool/type overlap, as percent of 151 and number of respondents. Mold- board plow % (#) Chisel plow % (#) Ripper Cultivator Harrow finisher Soil Vertical tillage Mulch Ridge till till Strip till No-till Cons. tillage Conv. tillage % (#) % (#) % (#) % (#) % (#) % (#) % (#) % (#) % (#) % (#) % (#) Primary tillage tools Moldboard plow Chisel plow Ripper Secondary tillage tools Cultivator Harrow Soil finisher Tillage types Vertical tillage Mulch till Ridge till Strip till No-till Cons. tillage Conv. tillage 2.0 (3) 0.7 (1) 0.0 (0) 0.7 (1) 0.0 (0) 0.0 (0) 0.7 (1) 0.0 (0) 0.7 (1) 0.0 (0) 0.7 (1) 0.0 (0) 0.0 (0) 0.7 (1) 19.2 (29) 3.3 (5) 5.3 (8) 4.0 (6) 3.3 (5) 2.0 (3) 0.0 (0) 0.0 (0) 0.0 (0) 12.6 (19) 4.6 (7) 1.3 (2) 0.0 (0) 3.3 (5) 19.2 (29) 7.9 (12) 4.0 (6) 0.7 (1) 4.0 (6) 0.7 (1) 0.0 (0) 2.6 (4) 6.6 (10) 2.0 (3) 4.0 (6) 0.7 (1) 5.3 (8) 7.9 (12) 17.9 (27) 4.0 (6) 1.3 (2) 2.0 (3) 0.7 (1) 0.7 (1) 0.7 (1) 7.3 (11) 2.0 (3) 3.3 (5) 0.0 (0) 4.0 (6) 4.0 (6) 4.0 (6) 12.6 (19) 0.0 (0) 3.3 (5) 0.7 (1) 0.0 (0) 0.0 (0) 6.6 (10) 1.3 (2) 1.3 (2) 0.0 (0) 3.3 (5) 0.7 (1) 1.3 (2) 0.0 (0) 5.3 (8) 0.0 (0) 0.0 (0) 0.0 (0) 0.0 (0) 4.0 (6) 1.3 (2) 0.0 (0) 0.7 (1) 2.0 (3) 4.0 (6) 2.0 (3) 3.3 (5) 0.0 (0) 15.2 (23) 0.0 (0) 0.0 (0) 0.7 (7) 6.6 (10) 2.6 (4) 0.0 (0) 0.0 (0) 0.0 (0) 0.7 (1) 0.7 (1) 0.7 (1) 0.0 (0) 0.0 (0) 1.3 (2) 0.0 (0) 0.0 (0) 0.7 (1) 0.0 (0) 0.7 (1) 0.7 (1) 0.0 (0) 0.0 (0) 0.7 (1) 0.0 (0) 0.0 (0) 0.0 (0) 0.0 (0) 0.7 (1) 0.0 (0) 0.0 (0) 0.0 (0) 0.0 (0) 0.0 (0) 0.0 (0) 2.6 (4) 0.7 (1) 0.0 (0) 0.0 (0) 0.7 (1) 0.0 (0) 0.0 (0) 11.9 (18) 5.3 (8) 0.7 (1) 2.0 (3) 0.7 (1) 12.6 (19) 6.6 (10) 7.3 (11) 6.6 (10) 4.0 (6) 6.6 (10) 0.7 (1) 0.0 (0) 5.3 (8) 64.9 (98) 8.6 (13) 3.3 (5) 0.0 (0) 4.6 (7) 2.0 (3) 2.0 (3) 1.3 (2) 1.3 (2) 2.6 (4) 0.0 (0) 0.0 (0) 0.7 (1) 8.6 (13) 13.9 (21) 1.3 (2) 0.0 (0) 1.3 (2) 4.0 (6) 3.3 (5) 1.3 (2) 0.0 (0) 0.0 (0) 0.7 (1) 0.0 (0) 2.0 (3) 3.3 (5) 1.3 (2) 8.6 (13) 38 Table 2.4. Summary of tillage tool use and tillage type by state, as percent and number of respondents. Number of interviews Primary tillage tools Moldboard plow Chisel plow Ripper Secondary tillage tools Cultivator Harrow Soil finisher Tillage types Vertical tillage Mulch till Ridge till Strip till No-till Conservation tillage Conventional tillage Primary Tillage Tools Michigan 48 % (#) Iowa 53 % (#) Indiana 51 % (#) 2.1 (1) 29.8 (14) 27.7 (13) 14.9 (7) 21.3 (10) 4.3 (2) 17.0 (8) 2.1 (1) 0.0 (0) 8.5 (4) 57.4 (27) 14.9 (7) 8.5 (4) 1.9 (1) 15.1 (8) 17.0 (9) 24.5 (13) 7.5 (4) 9.4 (5) 11.3 (6) 1.9 (1) 1.9 (1) 18.9 (10) 71.7 (38) 9.4 (5) 9.4 (5) 2.0 (1) 13.7 (7) 13.7 (7) 13.7 (7) 9.8 (5) 2.0 (1) 17.6 (9) 0.0 (0) 0.0 (0) 5.9 (3) 64.7 (33) 17.6 (9) 7.8 (4) Primary tillage displaces and shatters the soil as it reduces soil strength and tends to bury and mix plant residues and fertilizers within the tilled zone. Primary tillage is more aggressive, deeper and leaves a rough surface relative to secondary tillage operations. Primary tillage tools are the moldboard plow, chisel plow and various types of combination disc-chisel-subsoil tools designed to disturb the soil to greater depths. However, depending on the features, options and operation of that implement, results of primary tillage will be different as soil conditions change. (Reicosky and Allmaras 2003:78-79) As discussed above, the moldboard plow was long the go-to tillage implement of US farmers, but in the Corn Belt at least, this is no longer the case. Just three farmers, one in each state, mentioned that they use moldboard plows (Table 2.4) in some context. The IA and MI farmers indicated that they used it on a regular basis. All three farmers reported using other tillage types at other times or on other parts of their operation (Table 2.3). 39 Chisel plows and rippers are breaking and mixing tools10, rather than an inverting type and appear to have replaced the moldboard plow for typical primary tillage in the Midwest. Overall, almost 20 percent of the farmers interviewed said they used chisel plows in some capacity, and almost 20 percent said they used rippers in some capacity (Table 2.3). MI farmers reported higher use of both chisel plows and rippers (30 percent and 28 percent respectively) compared to IA (15 percent and 17 percent respectively) and IN (14 percent for both types of implements) (Table 2.4). Almost 13 percent said they used chisel plows and no-till, and almost 7 percent reported use of rippers and no-till (Table 2.3). This overlap was often based on their crop rotation. Given the role of chisel plows and rippers as primary tillage tools, the overlap between their use and secondary tillage tools is as expected (Table 2.3). Secondary Tillage Tools Secondary tillage varies widely with type and number of operations, penetration is nearly always shallower than with primary tillage tools. Secondary tillage provides additional soil breakup, levels and firms the soil, closes some air pockets and kills some weeds. Tillage equipment often associated with secondary tillage includes disk harrows, field cultivators, spring-tooth harrows, levelers, packers, and other types of finishing equipment. Interpreting tillage system results may be confusing or meaningless when little information about secondary tillage operations is available. (Reicosky and Allmaras 2003:79) The following tools, field cultivators, harrows, and soil finishers, are considered secondary tillage implements and would typically be used after one of the primary tillage tools discussed above. However, some farmers described using them without first using primary tillage as part of a low-tillage system. These tools were sometimes used sequentially or as 10 The main difference between these tools is the depth they penetrate the soil. Rippers (also called “subsoilers”) run deeper (12-22 inches) than chisel plows (6-12 inches) and can break up compaction below the depth of other tillage tools (MWPS 2000). Both chisel plows and rippers can be designed to provide greater or lesser amounts of soil surface disturbance. Rippers were often mentioned in the form of disk rippers (8 percent of total sample). 40 substitutes for each other, as shown in the overlap of use, especially between field cultivators and harrows (Table 2.3). Overall, field cultivators were used the most (18 percent), followed by harrows11 (13 percent) and soil finishers (5 percent) (Table 2.3). Field cultivators appear to be most popular in IA (25 percent) and similarly less popular in MI (15 percent) and IN (14 percent), while harrows are more commonly used in MI (21 percent) than in IN (9 percent) or IA (8 percent) (Table 2.4). Soil finishers were more popular in IA (9 percent) than in MI (4 percent) or IN (2 percent) (Table 2.4). Types of Tillage The following terms refer to tillage types or styles based on function, rather than specific tools, though some like no-till, strip till and ridge tillage do need tools with specific capabilities. While there is certainly plenty of room for variation within the tillage tools discussed above (different names for the same tool, the use of different options on a tool that make it perform quite differently, etc.) there is even more room for vagueness when it comes to tillage styles, since they are not bound to a particular tillage tool even by name12. Vertical and mulch tillage. Vertical tillage appears to be a broad term without a firm definition (Kanicki 2014). It generally involves the use of an implement with vertical blades intended to cut up residue with minimal soil disturbance, though some have shallowly curved discs or wavy coulters that provide some soil lifting (Kanicki 2014). Mention of vertical tillage use was fairly even across the three states at 15 percent, 11 percent and 18 percent for MI, IA 11 Almost 79 percent of harrow use was of a disk harrow, primarily as secondary tillage. When non-disk harrows were discussed they tended to be used in a first field pass context. The use of disk harrows was quite evenly split between use as primary and secondary tillage (10 farmers and 8 farmers respectively). As primary tillage, there were an even number of reports of fall vs. spring use (4 farmers each). The fall use was typically for residue management and the spring use was for planting preparation. 12 The one exception to this might be ridge tillage, which, while not tied to a specific tool per se does require a tool with quite a specific function. 41 and IN respectively (Table 2.4). Of the 23 total farmers considered to use this method, 18 labeled their practice/implement as vertical tillage, and 5 were considered to use it based on the tool they mentioned using—4 of the 5 said they used a “turbo till,” which is a trademarked name for an implement made by Great Plains that is billed as a vertical tillage tool. More mentions were made of using it in the spring, but fall use was also reported. There appears to be some overlap in tillage strategy with the harrows and cultivators in terms of timing and purpose of vertical tillage use, as well as with actual implement use. Mulch tillage is potentially an even broader term than vertical tillage, being generally defined as any conservation tillage system that is not no-till, strip till or ridge till, either with the requirement of 30% soil surface cover remaining (Caswell et al. 2001; John Deere 2007) or without (MWPS 2000; NRCS 2012b). Only two farmers described their tillage practices as “mulch tillage,” one in IA and one in MI (Table 2.4). The IA farmer described it as “a conventional mulch tillage thing, we field cultivate in the spring and do a disk ripping in the fall.” The MI farmer simply said that he does “mulch tillage on the soybeans to prepare for corn.” Ridge till. Ridge tillage is a specific practice where ridges are formed in the field and the crop is planted into the ridge for multiple years in a row while leaving the residue in the valleys between the rows, and is considered a residue-retaining type of tillage (Caswell et al. 2001; Reicosky and Allmaras 2003). Three farmers reported using ridge tillage in the past, but only one IA farmer reported currently using the practice (Table 2.4). Strip till. Strip till (also sometimes called “zone till”) is often considered a form of no-till (CTIC 2002; NRCS 2012a) but is considered separately in this analysis as most farmers distinguished between strip till and no-till, as can be seen by the small overlap in Table 2.3. Strip 42 till involves only tilling in strips across the field shortly before harvest, leaving the remaining portion of the field undisturbed (Reicosky and Allmaras 2003). The width of the allowed strip ranges up to 25 percent (Reicosky and Allmaras 2003) or 30 percent (CTIC 2002). Overall, 11 percent of our respondents used strip till in some capacity13 (Table 2.3). Nineteen percent of IA farmers mentioned using strip till compared to 6 percent of the MI and IN farmers (Table 2.4). Five percent of the sample noted using both strip till and no-till (Table 2.3). Some farmers blended the strip till/no-till distinction, either by using strip till and no-till at different points in their crop rotation, or by considering them as the same thing. “I've been no till since 1994. Right now, I’m really happy with my system. Its taking some time and haven’t been happy with it every year, this year I can be happy with how things have turned out…I work a 1/3 of the field. They’re roughly 10 inch wide strips the whole way across the field and I plant right into those strips.” IA07 No-till. As previously mentioned, no-till can include strip tilling practices, but a more narrow concept of it is that the only soil disturbance is the opening of a narrow slot by equipment on the front of the planter or seed drill (MWPS 2000; NRCS 2012a). No-till has overlap with all other types of tillage except ridge till, which is incompatible by definition (Table 2.3). No-till of some version was reportedly used by 65 percent of the farmers interviewed, making it the most- used practice (Table 2.3). It was also the most used by state, with 72 percent of IA farmers and 57 percent and 65 percent of MI and IN farmers respectively using it (Table 2.4). Because of the variation described in the use of no-till by participants, however, no-till has been further divided into 4 categories: all no-till, mostly or nearly no-till, some or part no-till, and no-till in rotation. The “all no-till” category is the use of 100 percent, or total no-till, or using no tillage. Farmers were still included in this group even if they mentioned specific exceptions in which 13 There was a mix of reported fall and spring strip till. 43 they tilled in some way. These exceptions were related to smoothing the ground after installing tile drainage or manure spreading, or if they started farming new land that was not ready to be no-tilled. The “mostly or nearly no-till” category is farmers who did not claim to be completely no-till but indicated that they mostly were. This included all no-till on most but not all of their acres, or all no-till most years but not always. The “some or part no-till” category is farmers who indicated that they do no-till some years, or on some parts of their operation, but not as consistently as the previous group. The “no-till in rotation” category is farmers who indicated that they used no-till according to their crop rotation. Most commonly this was no-tilling soybeans into corn stubble, but some also reported no-tiling corn into soybean stubble or as part of a more varied crop rotation. The least common use of rotation-based no-till was between corn crops. Table 2.5. No-till sub-categories by state and for total sample, as percent and number of respondents. Number of interviews All no-till Mostly or nearly no-till All no-till in rotation No-till corn into bean stubble No-till soy into corn stubble No-till in other rotation Some or part no-till No-till is strip till No-till is vertical till Michigan 47 % (#) 14.9 (7) 6.4 (3) 17.0 (8) 2.1 (1) 14.9 (7) 10.6 (5) 19.1 (9) 4.3 (2) 2.1 (1) Iowa 53 % (#) 28.3 (15) 1.9 (1) 24.5 (13) 15.1 (8) 11.3 (6) 1.9 (1) 9.4 (5) 3.8 (2) 1.9 (1) Indiana 51 % (#) 21.6 (11) 9.8 (5) 19.6 (10) 0.0 (0) 19.6 (10) 0.0 (0) 13.7 (7) 0.0 (0) 3.9 (2) Total 151 % (#) 21.9 (33) 6.0 (9) 20.5 (31) 6.0 (9) 15.2 (23) 4.0 (6) 13.9 (21) 2.6 (4) 2.6 (4) Table 2.5 breaks down these four categories by state, and it also indicates the fuzziness between no-till and strip till and between no-till and vertical till. The use of total no-till is lower in MI than in IA and IN, and the use of no-till by crop rotation is higher in IA than in the other two states. The location of no-till in the rotation has some interesting state-level patterns. All of 44 the IN farmers who indicated using no-till in their rotation mentioned doing so between corn and soybeans, where the soybeans were no-tilled into the corn stubble. MI farmers mostly no-tilled soybeans into corn stubble as well, but there were several instances of using it in a non-corn-soy rotation. In contrast, IA farmers reported no-tilling corn into soybean stubble slightly more often than no-tilling soybeans into corn stubble. The overlap between no-till and strip till and vertical till is relatively small. Some of this overlap likely comes some from the mixed definitions of no-till, and some from farmers who are transitioning between methods. “I’m a not till farmer… on my first year corn I’m not doing any tillage at all. On my continuous corn, I’d guess you’d say I’m a vertical tiller, only one year. I just bought that machine.” IA28 “I no till everything. As much as I can with exception if we tile something or do something different or alter something, but other than that we no till or coulter till. Maybe its more of a strip till, cause we have three coulters in front of our corn planter to till a little bit of a strip. Its not necessarily slot planting on the corn planter and the bean planter is a more of no till thing and the wheat is actually a no till. Strip till and no till. However you want to.” MI-40 “Conservation” tillage. As mentioned previously, “conservation tillage” can include a wide variety of practices, so it is of interest to see what farmers consider it to be. Twenty-one farmers (7 MI, 5 IA, 9 IN) identified themselves as using either “conservation” or “minimum” tillage on their operation, in whole or in part (Table 2.6)14. The “conservation” tillage term was used less by farmers (4 farmers) than “minimum” tillage (18 farmers). There was no dramatic difference between states in the use of these terms, so they are considered together. Table 2.6 shows the overlap between farmers who identified themselves as using conservation or minimum tillage and their use of different implements and tillage types. Chisel plows and no-till had the most overlap. None of the farmers who referred to their tillage as either “minimum,” or 14 The coding for this topic was quite conservative and was only used for farmers who used the specific terms— references to “reducing” or “minimizing” tillage were not included. 45 “conservation,” mentioned being completely or nearly no-till; they did report using no-till in their rotation, or on a portion of a crop or on a portion of their fields. Table 2.6. Comparison of tillage tools/types used for conservation (or minimum) tillage and conventional tillage, as percent and number of respondents. Number using Michigan Iowa Indiana Moldboard Chisel plow Ripper Cultivator as primary tillage Harrow as primary tillage Finisher as primary tillage Vertical tillage Ridge till Strip till No-till All no-till Mostly or nearly no-till All no-till in rotation Some or part no-till No-till is strip till No-till is vertical till Conservation tillage 21 % (#) 33.3 (7) 23.8 (5) 42.9 (9) 0.0 (0) 33.3 (7) 14.3 (3) 9.5 (2) 0.0 (0) 9.5 (2) 19.0 (4) 0.0 (0) 4.8 (1) 61.9 (13) 0.0 (0) 0.0 (0) 23.8 (5) 28.6 (6) 0.0 (0) 9.5 (2) Conventional tillage 13 % (#) 30.8 (4) 38.5 (5) 30.8 (4) 0.0 (0) 15.4 (2) 46.2 (6) 15.4 (2) 0.0 (0) 0.0 (0) 0.0 (0) 0.0 (0) 23.1 (3) 38.5 (5) 0.0 (0) 0.0 (0) 23.1 (3) 15.4 (2) 0.0 (0) 0.0 (0) “Conventional” tillage. Thirteen farmers, nearly evenly split by state, reported using what they called “conventional” tillage on all or part of their operation. One additional IA farmer referred to “full” tillage. Most of these farmers didn’t elaborate on their tillage methods, but two said they chisel plowed, six said used a ripper, three used strip till and five used no-till (Table 2.6). 46 Tillage Decision Factors While it is important to know what farmers are doing, it is also important to know why they are doing it. Three broad categories of influences on tillage practices emerged from the interviews: biophysical conditions, operational needs and compatibility, and operational savings and costs. For general reference, biophysical conditions were mentioned by about a third of the sample, topics related to operational needs and compatibility were also mentioned by about a third of the sample, and operational savings and costs were mentioned by about 15 percent of the sample. A handful of farmers mentioned personal values and preferences in relation to their use of no-till, but not frequently enough to be considered a major theme. Examples can be seen in Table 2.7 in the “Less tillage to improve soil” section and in Table 2.10 in the “Yield costs” section. Biophysical conditions. Biophysical conditions centering around soil were by far the most mentioned influence on tillage decisions. However, there was not consistency in how some of these conditions influenced tillage decisions, such that apparently similar conditions led some farmers to till more and others to till less. Reducing soil erosion was consistently cited as a reason for less tillage—either no-till or strip till. Soil quality and texture influenced farmers to either till more or till less. Farmers felt like less tillage improved their soil quality, though none specifically gave this as a reason for why they reduced their tillage to begin with but rather reported it as a result. The texture of the soil was a primary reason given for use of more tillage or not using no-till at least on some areas or in certain parts of the rotation—farmers perceived their soil type as being incompatible with the successful use of no-till. Addressing soil compaction typically called for more tillage that was targeted to the problem areas, but some farmers discussed tilling less to avoid soil compaction. 47 Managing soil temperature and moisture was mentioned as reasons for both tilling more and tilling less. Less tillage was described as both retaining moisture as well as managing excess moisture. However, more tillage was said to be needed to dry and warm the soil in the spring, especially prior to planting corn. Extreme weather events were mentioned by a couple of farmers in relation to their tillage practices, with one specifically saying “I think we’re gonna get more rain more often. I believe that our climate is changing and we’re having an effect on it. So… that’s largely why I went to no till.” IA09. However, this was not a common sentiment. Table 2.7. Examples of quotes pertaining to the biophysical conditions that influence farmers’ tillage decisions. Less tillage to reduce soil erosion “the dirt running off the land is what we try to conserve.” IA15 [strip till farmer] “I think we’re just going to have to get used to working with no till if we’re gonna stay in this corn bean rotation. We’re just losing too much soil.” IA24 “We have some places that are really eroded over the years; it’s one reason we went to no-till, to try to curb some of that.” IN#36 “And especially in the spring when we get high winds; I’ve seen a lot of property change hands in the wind, the dust and dirt, and I don’t lose it. Nope.” MI-30 [mostly or nearly no-till] Less tillage to improve soil “And we no-till, we haven’t worked any ground since ‘91, we went 100% no till, and what we’ve noticed over the years is our organic matter is up probably 2 to 2 ½ percentage points since then, and of course that’s a lot more available nitrogen right there.” IN#37 “now we have a deeper profile of a root zone to work with and I think we have much more volume of earth to hang onto what’s there.” IN#49 [a no-till/vertical tillage farmer] “I don’t think we use quite as much fertilizer because I think the soil is healthier. We’re getting good microbial activity. And breaking down the left overs you might say. So we’ve had success that way with some of our soil… I just like no till, I think its better for the soil” MI-2 More tillage due to soil texture “Our soils are really deep and black and fertile, so no till doesn’t seem to work very well around here.” IA63 “Our type of ground… No-till has been tried several times in our area and it just seems like everybody that… We’ve never tried it. We have no-tilled beans into corn ground that has not been tilled, that works half-way, well three-fourths way I’ll say. But corn into bean ground no-till it just seems like it does not work in this area. A lot of it is the type of ground we’ve got I think.” IN#13 “About every 3-4 years I like to rip it. Go in there, even when I no till, just trying to pick it up and drop it. Kinda loosens the soil up. I had some hills that were heavy clay and we couldn’t get anything to grow on them since I was a kid and now I can have just as good of a crop on those hills as any other spot.” MI-8 [farmer has been using no-till for several years] More tillage to reduce compaction “Yeah, we started no tilling in the early 90s and… Once in a while, maybe every… I’ll take a wheat stubble field if it shows some compaction and I’ve got a ripper stripper that I’ll run in the fall then when I plant corn I’ll plant right on those same strips, and I don’t do it every year but…” IN#41 “Well, I run primarily vertical tillage. Usually…I will use a chisel plow some, but predominately I’m probably on 75% of my acres I strictly use vertical tillage in the spring. If I’m trying to fight some compaction issues I’ll chisel plow in the fall.” MI-45 48 Table 2.7. (cont’d). Less tillage to reduce compaction “Before I started no tilling, it was a lot more variability in the yield, now it seems like its evened out a lot more… Seemed like we had a lot of compaction when we were doing tillage, so we had areas that were really bad. But that’s evened out now. When we went to no till, we bought an inline ripper and I ripped everything 2 years in a row to try to relieve that compaction and that made a big difference and I think that’s why we’ve been successful no tilling” IA48 Our ground tends to be on the wet side, it’s gummy, if you mess with it much deeper than that in the spring you bring nothing up the big old gummy balls of mud about like that, and they get harder’n a rock, and then it dries out and then you don’t grow anything. So that’s the reason we only scratch about that much in the spring, because if you go any deeper you’ve got mud, and you also start struggling with compaction. I mean, you don’t compact it unless you try to pull something hard and deep; if you just drive over it it doesn’t seem like it compacts it, but if you try to start pulling something and it makes the tractor work is when you start compacting. IN#13 Less tillage to manage moisture “My opinion, and experts talk about this too, 80% of our yield is determined by the weather and we can’t control that necessarily. We can do things to minimize the impact of the weather, you know, my opinion the no-till minimizes the impact of the weather.” IN#12 [goes on to describe having more moisture in a dry year] “I have certain fields that just tend to be wet fields, and I tend to try to keep those no-tilled just so that they’re a little easier to deal with.” MI-7 “I like no-till. And then in the fall if you get a wet year if you haven’t worked it for several years the ground’s nice and firm … You know, you get all the worm holes out there and they take the water” MI-36 More tillage to manage moisture “I have kinda stayed away from the no till for the corn crop anyway because of some of the, those soils need to be dried and warm. And the bottom grounds tend to not, depending on planting dates and that type of thing.” IA38 “Last year was the first year its been a detriment to us. We were so so terribly wet and with strip till it seems like you always have to wait about 1 day more than the conventional, they were going in in really wet conditions with a field cultivator and that would dry it up enough that they could get in with a planter and we had to wait.” IA19 “And I don’t strip till everything. I have some farms that aren’t drained very well. Just haven’t been brave enough to try the strip till on them yet.” IA54 More tillage to increase soil temperature “we do our low soils, we like to do some tillage because it helps them warm up. On the clay soils or sandy soils, there we’ll no till some. One of the things that’s been a concern is how cool we’ve been in the spring and actually have gone back to a little more tillage because of the cool spring because without the heat units the crop is off to a very slow start.” IA10 “I’ve tried no-tilling on our clay-loam soils, and some years it’s okay, but you get the wrong year the soil doesn’t warm-up and you’re just… You start out behind the eight ball. So, by doing some tillage in the spring to prepare for corn planting it warms the soil up and it gives me more consistent yields.” MI-33 “The problem with no till in this area is we don’t warm up in the spring enough and the crops don’t come out of the ground.” MI-34 Operational needs and compatibility. This theme focuses on the interplay between different crops and operations that take place on a field from one year to the next. It includes two sub-topics: crop rotation and residue management, and fertilizer application. As has previously been highlighted in the no-till section, many farmers till differently at different points in their crop rotation, and this is not limited to no-till. For example, among the farmers that elaborated on 49 their chisel plow use, 12 percent reported chisel plowing according to their rotation, or in other words, following certain crops but not others. Most commonly mentioned was chisel plowing between corn crops or prior to planting corn. Some of this tillage variation based on crop rotation can also be seen in a couple of the quotes in Table 2.7 above. Table 2.8. Examples of quotes pertaining to the crop rotation and residue management aspects of operational needs and practice compatibility features that influence farmers’ tillage decisions. Tillage in rotation and for residue management “Generally, second year corn, we will till with a chisel plow or some kind of tillage. Just to handle the residue and what not.” IA36 “Part of this vertical tillage and continuous corn is getting this tough crop mixed into the soil to get it breaking down faster.” IA51 “We do some tillage on our corn stalks. Basically for residue management to deal with the corn stalks.” IA63 “we’re making a tillage pass in front of everything. Doing it to size residue, with the twin row corn, with the higher populations we’ve got a lot of residue, so we’re doing it to size residue and to bury some residue and then for weed control and just to aerate and warm-up the soil.” IN#45 Effects of newer corn varieties “Now you’re starting to see some tillage come back into soybeans more because of that darn residue with the corn. We’re putting fungicides on which make it not deteriorate, the higher populations, the GMO corns, they did exactly what we wanted, we wanted a strong stalk to hold the plant up but that strong stalk doesn’t break down. You almost need to break that stalk down so you can plant a good crop of soybeans next year.” MI-18 “We got a tool they call vertical till that we go over the bean ground with and its mainly to break down the corn stalks. The corn stalks are greener and tougher now than they used to be. They bred that in to the seed.” MI-41 “you get your soil so that it’s kind of working, breaking down stuff, it was working good until the Bt corn came into practice, and then you can break that stuff down, and we never applied any nitrogen on the stalks or anything like that.” MI-50 Tillage in relation to planting equipment “Its [tillage] more to get the stand of the crop. We’ve got a no till drill and with these heavier corn stalks and stuff we were having an issue getting the stand that we want. Even in the wheat and I don’t know why. We just feel its worth that [tillage] trip, we get a better stand.” MI-26 “Corn we usually… we do a minimum till with the corn, just because the corn planter… We have some issues with our corn planter so we do a minimum till; we’ll go in and run our finisher in front of it and just get the soil loosened up and stuff so we have a better… We feel we have a better seed rate.” IN#48 At least some of the reason for tilling based on rotation is to handle the residue that remains from the previous crop. The higher volumes of residue from corn (versus that of soybeans for example) were most often cited as the focus of tillage for this purpose. Several farmers mentioned residue management being a bigger issue in recent years, with newer corn varieties having stronger stalks that take longer to break down. A few farmers also discussed 50 residue management in relation to the operation of their planting equipment—needing to reduce the residue so their planting equipment would work better. There are also certain timing aspects to tillage that are based on rotation since harvest times differ between crops. This creates more opportunity for additional operations in the fall after some crops rather than others. For example, wheat that is typically harvested in July in the study area, versus October for corn and soybeans (NASS 2010) (see IN#41 quote in Table 2.7). Fertilizer application was the second sub-topic of the operational needs and compatibility theme. This has to do with the interaction between tillage practices and fertilizer practices, and how some fertilizer practices are incompatible with certain tillage practices, particularly no-till. The major issue was the desire to incorporate fertilizer into the soil through some sort of tillage, rather than leave them on the soil surface, though not all no-till farmers were concerned about it. This factor either put farmers off total no-till or pushed them to change or select their fertilizer practices to be compatible. As can be seen in the examples in Table 2.9, it was relatively common for farmers to switch from no-till to strip till in order to incorporate their applied nutrients. The use of anhydrous ammonia15 as a nitrogen source was particularly mentioned as being incompatible with no-till, leading farmers to either not fully no-till or not use anhydrous ammonia. The use of animal manure as a nutrient source was also mentioned as being a particular challenge with no-till, either through the injection process, the need to incorporate the manure if it has been spread on top of the soil, and/or the need to smooth the field following manure application and prior to planting. Manure is a resource but also a waste product and spreading manure on fields is a way to dispose of it, even if it is not optimal from a cropping standpoint. 15 Anhydrous ammonia is a gas and has to be injected under the soil surface using a ripper-style shank. 51 Because it is so bulky, manure does not tend to be hauled very far, so it is not necessarily spread evenly across an operation’s fields. Table 2.9. Examples of quotes pertaining to the fertilizer application aspects of operational needs and practice compatibility features that influence farmers’ tillage decisions. Need for fertilizer incorporation “Well, I actually switched to strip till because of the fertilizer management… When we went to straight no till, that P and K was just sitting on the surface. And people debate, about 50 percent say it doesn’t matter. But I feel like you need to get it incorporated, so that’s why I went to strip till. The strip till applicator actually injects that fertilizer about 4-6 inches deep, so I’m not spreading it on the surface anymore. So did tillage effect the way fertilizer, its probably the other way around, the way I fertilize effected how I wanted to do my tillage.” IA06 “I didn’t start out trying to go to strip tillage, my real goal was to get the fertilizer in the ground rather than lay it on top. And I did not like that fact that I was practically giving up the nitrogen that was in my MAP and DAP. That’s practically free nitrogen. And if you’re putting on 200 pounds an acre, that’s 36 units at say 15 cents a unit, that’s 15 dollars worth of nitrogen. Why not take advantage of it?” IA64 “when we started no tilling is when we started side dressing nitrogen, because historically my father had put all the nitrogen on with the local ag retailer, broadcast before we planted and we would work it in, and I just didn’t feel like leaving 200 pound of N lying on top. And I never thought it was a good idea to put all the N in spring.” IN#12 Anhydrous ammonia and no-till “Minimum till. We’re really not in a no-till situation; the anhydrous tracks pre-plant makes it tough to plant behind.” IN#47 “Well, we don’t use anhydrous for any purpose, and part of that is if you’re going to no till I don’t really want an anhydrous bar running through it prior. We primarily use 28…” IN#16 “…although they’ve got better equipment now, but that’s why we got set up with 28 because pre-plant anhydrous and no-till doesn’t work.” IN#33 Manure use For the strip tilling, does that encourage you towards manure use or away? “Probably away. They’re diabolically opposed based on the seasonality and the till, the strip till that goes in the bean stubble for residue management, that’s a really nice way to do that, but obviously with the hog manure, between compaction and the way the hog manure is injected, strip till may not work with that very well.” IA62 “We have all this manure to deal with, so no-till, just isn’t a good fit if you have a lot of manure, because you’ve got to kinda work it in.” IA23 “Where we have been able to no-till and avoid hauling manure on it definitely has an impact, but manure application just destroys a lot of the benefits… but we’ve got to spread manure around so…” MI-28 Operational savings and costs. The third major influence on tillage practice decisions is related to operational savings that come with less tillage in the form of time, labor, equipment, and/or fertilizer. The saving of time/labor was the most frequently and clearly articulated and came through fewer trips through the field and the ability to farm with fewer people. Fewer trips through the field also saves fuel and wear and tear on equipment. For some farmers less tillage was not their goal, but was an acceptable enough option when they ran out of time tilling all of 52 their land with their preferred method. There was also the potential savings on fertilizer, either through strip tilling and restricting fertilizer application to the strips, or through improved soil quality and biological functions. However, some farmers mentioned that savings came at the cost of yield, or there were other disadvantages. Table 2.10. Examples of quotes pertaining to the operational savings and costs that influence farmers’ tillage decisions. Time, labor, fuel and equipment savings “A side benefit of going to no till is that I don’t need to hire too many people. I can keep working my 82 year old father to death. There is a lot less trips across the fields. A lot less tractor time.” IA09 “I mean we do do some no-tilling, you know, because you can’t keep a field cultivator far enough ahead sometimes, so, you know, you may do 400 or 500 acres no-till.” IN#38 “I’m probably leaning more towards no till. Some of my fields are getting further and further away. And just the time and the fuel costs. If I have a planter that is set up to do it, why not?” MI-10 How long have you been doing the minimum till? “Oh, my dad and my grandpa started in 1980. Early. They were one of the first ones in the county to start, and really they started out of necessity because it was just two guys, and my dad was doing the same thing I’m doing now, he worked full-time down at General Motors and they just didn’t have time to be working ground.” MI-25 “We farmed 3000 acres when my dad was alive and we made some major changes [after he died] because it was just overwhelming, and we got rid of some of the less productive farms, and it was just my brother and I and we had a couple hired men and we went to no-till… I would never go back to conventional till; it’s just made our lives so much simpler… We use to just go through an enormous amount of diesel fuel and cultivator shovels when you’re out there working the ground, working the ground, working the ground, then you’ve gotta pick up stones, then you’ve got a clean out the ditches in the field; I mean, it just eliminated so much work it’s phenomenal. I mean, you could farm two or three times more ground because you’re not out there beating it all to death.” MI-36 “I wanted to get to this no till stuff. We were starting out farming and I didn’t want to go out and buy all this huge equipment.” IA51 Fertilizer savings “It [strip till] makes my fertilizer more efficient, it put it on right where the corn plant is gonna be.” IA26 So you were saying you’re strip tilling. Do you feel like that impacts nitrogen use compared to other forms of tilling that you might use or could use? “Cuts it by a third, minimum. Everything’s in the band where the crops going to use it.” MI-35 “Well, low crop prices, you know, we’ve had a stretch where margins were very tight, so… But when I started strip tilling we cut fertilizer 40%, you know, because you weren’t spreading 200 pounds of potash, you only needed 60 pounds in the band;” MI-35 “The no-till system… once it’s established I think the nitrogen use is probably a little bit more efficient, there’s a transitional period where it takes a little bit more nitrogen to feed that a little more biological system. For us we’ve been no-tilling for about 20 years and so with the exception of new farms that we pick up. We feel overall… I’m going to say probably that we’re about 30 pounds less nitrogen required and that comes from the cycling of the stuff back from the biology that it seems to be that we’re seeing there, compared to the kind of best guess we can get from guys around here on conventional systems.” IN#18 53 Table 2.10. (cont’d). Yield costs “So, if we keep the soil in place, first and foremost, and the labor savings are a side benefit. But, I’m also leaving a little bit of money on the table by going to no till. If no till was the most profitable way to go, everybody would switch over in a season. As it is, hardly anybody switches over. Cause there is more money to be made, more yields to be had, planting that seed in freshly turned over black warm dirt, instead of unturned over soil that’s had a crop on it. So, people are not streaming to no till, but they are gradually coming around.” IA09 “And I realized I suffered some yield, but I’m a no tiller. Cause the earth will be hear a lot longer than I am. But we don’t know, that’s the philosophy that I’m farming with.” IA53 “I’ve kinda been going back and forth whether to go back to tilling or not. You know, advantageous and disadvantages. You save fuel and stuff. For the most part, we’ve been fairly satisfied with the yields. It’s kinda a pain…at the end, when you strip till you pick up a big chunk and its harder to keep your fields level. You need a finish tool.” IN#3 DISCUSSION Variation in Tillage Practices and Meanings The results of this analysis show that there are several types of variation in tillage tool/type use among Midwestern farmers. The most obvious variation is in tillage tool/type use within a single farm: many farmers in the sample reported using multiple tillage tools/types, either consistently, based on soil type or crop rotation, or on an “as-needed” basis, such as to address soil compaction in specific places. The other type of variation is between farmers in the specifics of how tillage tools and types are used. This variation includes differences in how tools are set up—such as the size and shape of chisel plow and ripper shanks and the depths at which they are used—as well as what tools are used for different types of tillage. A prime example of variation in tillage tools and types this is whether chisel plows are an acceptable part of “minimum” tillage, with farmers expressing both views: “We are minimum till; we try to… most everything gets chisel plowed in the fall but we leave a lot of residue on the top.” MI-19 “We’re all minimal till. We don’t chisel plow.” MI-12 The demonstrated variation in spatial and temporal dimensions of tillage use, as well as in the meanings of different tillage types, suggests that recent research has glossed over 54 potentially important variation in tillage practices either by asking about tillage in too general a way and/or analyzing it in too general a way. Ryan, Erickson, and De Young (2003) recognize that they likely asked about the use of no-till in a way that potentially included a variety of actual practices. Vitale et al. (2011) and Andrews et al. (2013) do better by defining their categories, and in the case of Andrews et al. (2013) including differences in tillage by crop rotation in the question. However, neither Vitale et al. nor Andrews et al. seem to make use of the richness of their data as they collapse the variation into the typical categories. Reimer, Weinkauf et al. (2012) similarly recognize a more complex range of tillage practices, but did not seem to consider the meaning of this range—though their smaller sample size would have made this difficult. There is an argument to be made for the relevance of general or collapsed categories of tillage practices—convenience, interest in relative versus absolute categories, the value of how farmers self-identify and their ostensible intention, the importance of what farmers do on the majority of their land etc. However, there are potentially important environmental differences between the use of practices within the common tillage categories, and the use of these common categories sets researchers up to ignore an interesting line of research on the significance of using multiple types of tillage. The present analysis and previous studies (Andrews et al. 2013; Reimer, Weinkauf, et al. 2012; Vitale et al. 2011) indicate that many farmers do use multiple tillage practices on their farms, and the present analysis suggests several reasons why they might do so, such as soil qualities, crop rotation, compatibility with other practices etc. But why do all farmers not do this? The descriptives tables in Vitale et al. (2011), and Andrews et al. (2013) suggests that there could be some notable differences between farmers who use all one kind of tillage and those who 55 use multiple types, which suggests an interesting line of future investigation. Are these farmers simply incomplete adopters who need more encouragement to complete the transition process to no-till? Or are these farmers actually the ideal? Are they the farmers with the skill, knowledge and equipment to tailor their practices to their crops, soil and weather conditions, such that they are flexible and resilient to a range of social, personal, economic and climate conditions? Variation in Tillage Reasons The results of this analysis suggest that biophysical conditions, operational needs and compatibility, and operational savings and costs are the dominant decision factors for tillage in Midwest corn agriculture. In contrast to the implied importance of attitudes and values in the use of social-psych theories to model farmer BMP decision-making, few farmers in the study mentioned values they hold as influencing, or at least being consistent with, their tillage practices. This suggests that farmers’ environmental attitudes and values are not playing a conscious role in their tillage decisions. While the unconscious role that attitudes and values might play may be meaningful, my results suggest that additional logistical and biophysical factors should at least be considered in conjunction with them. Vitale et al. (2011) and Reimer, Weinkauf et al. (2012) both consider perceived characteristics of conservation tillage and found them to be important positive and negative factors on practice use. They both find the perception that conservation tillage is good for reducing erosion and conserving soil as an important positive predictor, along with savings on inputs and labor, which is consistent with the findings of my analysis. Vitale et al. (2011) especially found soil type to be in important reason for not using conservation tillage, with soil type and trouble with soil compaction having negative influences on conservation tillage use. Similarly, Reimer, Weinkauf et al. (2012) found “lack of perceived need” as a reason for not 56 adopting conservation tillage. These previous findings are also consistent with the results of my analysis and reinforce the importance of perceived practice characteristics and what Rogers (2003) calls the “relative advantage” and “compatibility” characteristics of tillage practices, where the practice being considered needs to be better in some way than the current practice (in this case, provides economic savings in some form), but also needs to be compatible with the existing system—in this case both the biophysical conditions as well as fertilizer application practices. What is perhaps most surprising in my results, is that farmers reported different decisions to ostensibly the same soil conditions. This suggests that attitudes and values could possibly be influencing tillage decisions indirectly by influencing how farmers perceive their biophysical conditions, and how they prioritize different aspects of their operation (in other words, how do they define what they “can” and “cannot” do?). This is most apparent in Table 2.7, where soil compaction, and wet soils can either push farmers to till more or till less. This could be a case where the biophysical context is different enough to call for or make possible different responses, or it could be that farmers are in fact choosing to handle the same conditions in opposite ways. In either case, combined with the overall importance of soil qualities and conditions as a basis for tillage tool/type use, it suggests that researchers need to better consider the specific biophysical conditions in which a particular farmer is operating. “It encouraged more of us to use no till. Especially planting soy beans back into corn. And then, we could see the variations from field to field, even doing the same practices with different soils. Some of my lighter soils I went stronger into no till and then some of my fields where I was like, I could soil condition index there better with corn on corn even with tillage. It kinda tuned us in to how different practices on different fields was appropriate. Since then, I’ve moved into more of a resiliency concept, how do we handle these extreme weather events and variations. And how do we prepare our land for that.” IA04 57 Implications and Directions for Future Research Future social research on tillage practices is advised to consider three elements. 1) When collecting data on tillage practices, design questions to capture, or at least not preclude, the use of multiple tillage tools/types on the farm at the same time, the variation in tillage by crop, and other forms of temporal and spatial variation in tillage. Assumptions should not be made on the consistency of tillage practices from year to year, or extrapolate the practices used on a single field to those used on the whole operation. 2) Incorporate research from other disciplines; such as by working with soil scientists to focus interview/survey questions on the specific characteristics of tillage practices that are most important from an environmental perspective—be it season of tillage, depth of tillage, use or non-use of specific tools, number of tillage passes etc. 3) Consider the significance of farmers using multiple types of tillage. It is also worth noting that none of the farmers interviewed described their tillage practices in terms of the percentage of residue that remained—not for a specific tool or a type of tillage. This suggests that a common feature of tillage type definitions used on surveys may not be very helpful to farmers in accurately selecting the response category that reflects their tillage practices, even if the question were to be specific to a single field and year and thus removes the use of multiple tillage types from consideration. It seems likely farmers are categorizing their tillage practices when asked to do so based on broad conceptions of the practice and/or on the labels attached to the tillage tools they use, which may or may not match with definitions used on surveys. A final note should be made on the limitations and advantages that result from the design of the present study. The largest limitation is that the question about tillage practices was very general and was often asked in conjunction with the interaction between tillage and fertilizer 58 practices, resulting in many farmers focusing their responses on the later aspect and thus providing little detail on their tillage practices. The interviews were also focused on corn, so it is possible that many farmers only discussed their tillage on corn and did not mention how they till in other parts of their rotation. The result is that these interviews likely under-report farmers use of multiple tillage tools/types and their use in their crop rotation. The advantage of the general nature of how the tillage question was asked is that the reasons that farmers mentioned for their tillage practices were un-prompted (except for the fertilizer interaction and savings) and thus are expected to much more accurately represent the reasons that are most important to them. CONCLUSION This exploratory analysis had two research questions: How and why do Midwest farmers till? I find that how farmers till varies widely, with farmers utilizing a wide range of tillage tools and frequently use multiple tillage tools/types on different parts of their farm and at different times. Notably, the common tillage categories were found to contain great internal variation, especially in the case of no-till, which showed important temporal variation in its use that is rarely acknowledged in the literature. This intra-category variation poses a great challenge to the utility of the tillage categories currently in use in social science research as these categories are likely not measuring what the researchers think that they are, even when category definitions are provided, as spatial and temporal variation are rarely incorporated. The reasons for tillage decisions that emerged as most important were soil conditions, operational compatibility, and savings, though not always in consistent directions. The inconsistency in farmers’ tillage responses to soil conditions suggests the importance of 1) better understanding the soil conditions that farmers are actually dealing with so that their responses to 59 soil conditions can be assessed, and 2) the biophysical knowledge and social constructions that farmers have that influence their perceptions of their soil conditions and what actions will lead desirable outcomes. This later point pushes beyond the capacity of the adoption-diffusion model and suggests the utility of adding a critical realism perspective to efforts to understand farmer decision-making (Carolan 2006). Critical realism assumes the existence of a tangible, biophysical reality but also acknowledges that there can be multiple meanings assigned to it, making possible multiple, and even conflicting, knowledge claims about the same ostensible reality (Carolan 2005). The spatial and temporal variation in tillage practices reported, and the importance of the biophysical context in which the farmer is operating should be examined more systematically in future research. The exciting aspect of the diversity in tillage practices is that tillage is not a single binary decision, which leaves opportunity for experimentation and incremental change towards more sustainable farming practices. Better understanding why farmers use multiple types of tillage, can advise ways of encouraging farmers to take the first step and start using a practice on some of their fields and the potential to get farmers to use a given practice on a greater portion of their land. The work of soil scientists and those in related fields should be utilized to focus research efforts on the specific aspects of tillage practices that are most important environmentally. 60 REFERENCES 61 REFERENCES Allmaras, R. R. and R. H. Dowdy. 1985. “Conservation Tillage Systems and Their Adoption in the United States.” Soil and Tillage Research 5(2):197–222. Allmaras, R. R., D. W. Wilkins, O. C. Burnside, and D. J. Mulla. 1998. “Agricultural Technology and Adoption of Conservation Practices.” Pp. 99–158 in Advances in Soil and Water Conservation, edited by F. J. Pierce and W. W. Frye. Chelsea, MI: Sleeping Bear Press. Andrews, A. C., R. A. Clawson, B. M. Gramig, and L. Raymond. 2013. “Why Do Farmers Adopt Conservation Tillage? An Experimental Investigation of Framing Effects.” Journal of Soil and Water Conservation 68(6):501–11. Aneja, Viney P., William H. Schlesinger, and Jan Willem Erisman. 2009. “Effects of Agriculture upon the Air Quality and Climate: Research, Policy, and Regulations.” Environmental Science & Technology 43(12):4234. Baumgart-Getz, Adam, Linda Stalker Prokopy, and Kristin Floress. 2012. “Why Farmers Adopt Best Management Practice in the United States: A Meta-Analysis of the Adoption Literature.” Journal of Environmental Management 96(1):17–25. Belknap, John and William E. Saupe. 1988. “Farm Family Resources and the Adoption of No- Plow Tillage in Southwestern Wisconsin.” North Central Journal of Agricultural Economics 10(1):13–23. Bultena, Gordon L. and Eric O. Hoiberg. 1983. “Factors Affecting Farmer Adoption of Conservation Tillage.” Journal of Soil and Water Conservation 38(3):281–84. Carolan, Michael S. 2005. “Society, Biology, and Ecology: Bringing Nature Back Into Sociology’s Disciplinary Narrative Through Critical Realism.” Organization & Environment 18(4):393–421. Carolan, Michael S. 2006. “Do You See What I See? Examining the Epistemic Barriers to Sustainable Agriculture.” Rural Sociology 71(2):232–60. Carpenter, S. R., N. F. Caraco, D. L. Correll, R. W. Howarth, A. N. Sharpley, and V. H. Smith. 1998. “Nonpoint Pollution of Surface Waters with Phosphorus and Nitrogen.” Ecological Applications 8(3):559–68. Caswell, Margriet, Keith Fuglie, Cassandra Ingram, Sharon Jans, and Catherine. Kascak. 2001. Adoption of Agricultural Production Practices: Lessons Learned from the U.S. Department of Agriculture Area Studies Project. Agricultural Economic Report No. 792. Resource Economics Division, Economic Research Service, U.S. Department of Agriculture. Retrieved August 25, 2017 (https://naldc.nal.usda.gov/download/38367/PDF). 62 Coughenour, C. Milton. 2003. “Innovating Conservation Agriculture: The Case of No-till Cropping.” Rural Sociology 68(2):278–304. Creswell, John W. 2007. Qualitative Inquiry & Research Design: Choosing Among Five Approaches. 2nd ed. Thousand Oaks, CA: Sage Publications, Inc. Crosson, Pierre. 1981. Conservation Tillage and Conventional Tillage: A Comparative Assessment. Ankeny, IA: Soil Conservation Society of America. Crosson, Pierre. 1984. “New Perspectives on Soil Conservation Policy: Conservationists Must Begin to Address Some Tough, New Policy Questions.” Journal of Soil and Water Conservation 39(4):222–25. CTIC. 2002. “Tillage Type Definitions.” Conservation Technology Information Center. Retrieved April 20, 2018 (http://www.ctic.purdue.edu/resourcedisplay/322/). Dillman, Don A. and John E. Carlson. 1982. “Influence of Absentee Landlords on Soil Erosion Control Practices.” Journal of Soil and Water Conservation 37(1):37–41. Ervin, Christine A. and David E. Ervin. 1982. “Factors Affecting the Use of Soil Conservation Practices: Hypotheses, Evidence and Policy Implications.” Land Economics 58(3):277–92. Glaser, Barney G. 1978. Theoretical Sensitivity: Advances in Methodology of Grounded Theory. Mill Valley, CA: Sociological Press. Glaser, Barney G. 1992. Basics of Grounded Theory Analysis. Mill Valley, CA: Sociological Press. Glaser, Barney G. and Anselm L. Strauss. 1967. The Discovery of Grounded Theory: Strategies for Qualitative Research. Chicago: Aldine Publishing Company. Gould, Brian W., William E. Saupe, and M. K. Richard. 1989. “Conservation Tillage : The Role of Farm and Operator Characteristics and the Perception of Soil Erosion.” Land Economics 65(2):167–82. Halcrow, Harold G., Earl O. Heady, and Melvin L. Cotner, eds. 1982. Soil Conservation Policies Institutions and Incentives. Ankeny, IA: Soil Conservation Society of America. John Deere. 2007. Tillage: A Practical Guide to the Latest Tillage Method, Conservation Planning, Crop Residue Management, and Solutions to Soil Compaction. 4th ed. Moline, IL: Deere & Company, John Deere Publishing. Kanicki, Dave. 2014. “What Is Vertical Tillage Anyway?” Farm Equipment. Retrieved April 20, 2018 (https://www.farm-equipment.com/articles/10122-from-the-march-2014-print-edition- what-is-vertical-tillage-anyway). Knowler, Duncan and Ben Bradshaw. 2007. “Farmers’ Adoption of Conservation Agriculture: A Review and Synthesis of Recent Research.” Food Policy 32(1):25–48. 63 Korsching, Peter F., Curtis W. Stofferahn, Peter J. Nowak, and Donald J. Wagener. 1983. “Adopter Characteristics and Adoption Patterns of Minimum Tillage: Implications for Soil Conservation Programs.” Journal of Soil and Water Conservation 38(5):428–31. Kumar, S. et al. 2014. “Long-Term Tillage and Drainage Influences on Soil Organic Carbon Dynamics, Aggregate Stability and Corn Yield.” Soil Science and Plant Nutrition 60(1):108–18. Lal, R., D. C. Reicosky, and J. D. Hanson. 2007. “Evolution of the Plow over 10,000 Years and the Rationale for No-till Farming.” Soil and Tillage Research 93(1):1–12. Lee, Linda K. and William H. Stewart. 1983. “Landownership and the Adoption of Minimum Tillage.” American Journal of Agricultural Economics 65(2):256. Liu, Tingting, Randall Bruins, and Matthew Heberling. 2018. “Factors Influencing Farmers’ Adoption of Best Management Practices: A Review and Synthesis.” Sustainability 10(2):432. López-Fando, C. and M. T. Pardo. 2012. “Use of a Partial-Width Tillage System Maintains Benefits of No-Tillage in Increasing Total Soil Nitrogen.” Soil and Tillage Research 118:32–39. MWPS. 2000. Conservation Tillage Systems and Management: Crop Residue Management with No-till, Ridge-till, Mulch-till, and Strip-till (MWPS-45). 2nd ed. Ames, IA: MidWest Plan Service, Iowa State University. Napier, TL, CS Thraen, and A. Gore. 1984. “Factors Affecting Adoption of Conventional and Conservation Tillage Practices in Ohio.” Journal of Soil and Water 39(3):205–9. NASS. 2010. Field Crops: Usual Planting and Harvesting Dates. Agricultural Handbook Number 628. National Agricultural Statistics Service, U.S. Department of Agriculture. Retrieved April 21, 2018 (http://usda.mannlib.cornell.edu/usda/current/planting/planting- 10-29-2010.pdf). Norris, Patricia E. and Sandra S. Batie. 1987. “Virginia Farmers’ Soil Conservation Decisions: An Application of the Tobit Analysis.” Southern Journal of Agricultural Economics 19:79– 90. Nowak, Peter J. 1987. “The Adoption of Agricultural Conservation Technologies: Economic and Diffusion Explanations.” Rural Sociology 52(2):208–20. Nowak, Peter J. and Peter F. Korsching. 1985. “Conservation Tillage: Revolution or Evolution?” Journal of Soil and Water Conservation 40(2):199–201. NRCS. 1992. Farming with Crop Residues. Natural Resources Conservation Service, U.S. Department of Agriculture. Retrieved (https://www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/nrcs141p2_029000.pdf). 64 NRCS. 2012a. Conservation Practice Standard Overview: Residue and Tillage Management (329) No-Till/Strip Till/Direct Seed. Natural Resources Conservation Service, U.S. Department of Agriculture. Retrieved (https://www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/stelprdb1254958.pdf). NRCS. 2012b. Conservaton Pracice Standard Overview: Residue Management, Mulch Till (345). Natural Resources Conservation Service, U.S. Department of Agriculture. Retrieved April 20, 2018 (https://www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/stelprdb1254982.pdf). Prokopy, L. S., K. Floress, D. Klotthor-Weinkauf, and A. Baumgart-Getz. 2008. “Determinants of Agricultural BMP Adoption: Evidence from the Literature.” Journal of Soil and Water Conservation 63(5):300–311. Rahm, Michael R. and Wallace E. Huffman. 1984. “The Adoption of Reduced Tillage: The Role of Human Capital and Other Variables.” American Journal of Agricultural Economics 66(4):405–13. Rasmussen, Wayne D. 1982. “History of Soil Conservation, Institutions, and Incentives.” Pp. 3– 18 in Soil Conservation Policies, Institutions, and Incentives, edited by H. G. Halcrow, E. O. Heady, and M. L. Cotner. Ankeny, IA: Soil Conservation Society of America. Reicosky, D. C. and R. R. Allmaras. 2003. “Advances in Tillage Research in North American Cropping Systems.” Journal of Crop Production 8(1/2):75–125. Reimer, Adam P. Aaron Thompson, Linda Stalker Prokopy, J. Gordon Arbuckle, Ken Genskow, Douglas Jackson-Smith, Gary Lynne, Laura McCann, Lois Wright Morton, and Pete Nowak. 2014. “People, Place, Behavior, and Context: A Research Agenda for Expanding Our Understanding of What Motivates Farmers’ Conservation Behaviors.” Journal of Soil and Water Conservation 69(2):57A–61A. Reimer, Adam P., Aaron W. Thompson, and Linda S. Prokopy. 2012. “The Multi-Dimensional Nature of Environmental Attitudes among Farmers in Indiana: Implications for Conservation Adoption.” Agriculture and Human Values 29(1):29–40. Reimer, Adam P., Denise K. Weinkauf, and Linda S. Prokopy. 2012. “The Influence of Perceptions of Practice Characteristics: An Examination of Agricultural Best Management Practice Adoption in Two Indiana Watersheds.” Journal of Rural Studies 28(1):118–28. Robertson, G. Philip Katherine L. Gross, Stephen K. Hamilton, Douglas A. Landis, Thomas M. Schmidt, Sieglinde S. Snapp, and Scott M. Swinton. 2014. “Farming for Ecosystem Services: An Ecological Approach to Production Agriculture.” BioScience 64(5):404–15. Robertson, G. Philip, Eldor A. Paul, and Richard R. Harwood. 2000. “Greenhouse Gases in Intensive Agriculture: Contributions of Individual Gases to the Radiative Forcing of the Atmosphere.” Science 289(5486):1922–25. Rogers, Everett M. 2003. Diffusion of Innovations. Fifth. New York, NY: Free Press. 65 Ryan, Robert L., Donna L. Erickson, and Raymond De Young. 2003. “Farmers’ Motivation for Adopting Conservation Practices along Riparian Zones in a Mid-Western Agricultural Watershed.” Journal of Environmental Planning and Management 46(1):19–37. Shortle, James S. and John A. Miranowski. 1986. “Effects of Risk Perceptions and Other Characteristics of Farmers and Farm Operations on the Adoption of Conservation Tillage Practices.” Applied Agricultural Research 1(2):85–90. Uri, Noel D. and Herby Bloodwirth. 2000. “Global Climate Change and the Efect of Conservation Practices in US Agriculture.” Global Environmental Change 10:197–209. Vitale, J. D., C. Godsey, J. Edwards, and R. Taylor. 2011. “The Adoption of Conservation Tillage Practices in Oklahoma: Findings from a Producer Survey.” Journal of Soil and Water Conservation 66(4):250–64. Walker, Diane and Florence Myrick. 2006. “Grounded Theory: An Exploration of Process and Procedure.” Qualitative Health Research 16(4):547–59. Warriner, G. Keith and Trudy M. Moul. 1992. “Kinship and Personal Communication Network Influences on the Adoption of Agriculture Conservation Technology.” Journal of Rural Studies 8(3):279–91. Wauters, Erwin and Erik Mathijs. 2014. “The Adoption of Farm Level Soil Conservation Practices in Developed Countries: A Meta-Analytic Review.” International Journal of Agricultural Resources, Governance and Ecology 10(1):78–102. 66 CHAPTER 3: DRIVERS OF THE USE OF FIVE BMPS IN MIDWEST ROW-CROP AGRICULTURE The need for US commodity crop farmers to use increasing numbers of agricultural best management practices (BMPs) in order to reduce their effect on the surrounding environment has been well established (Holmes, Armanini, and Yates 2016; Lam, Schmalz, and Fohrer 2011; Liu et al. 2016; Sowa et al. 2016). At the same time, there has long been an interest in understanding the reasons farmers choose to use or not use these practices. In the last decade several meta- analyses have attempted to summarize the findings of this expansive research across practices and conceptual approaches (Baumgart-Getz, Prokopy, and Floress 2012; Knowler and Bradshaw 2007; Liu, Bruins, and Heberling 2018; Prokopy et al. 2008; Wauters and Mathijs 2014). However, there remains little agreement on what variables drive BMP use, and neither of the main conceptual approaches—the adoption-diffusion model and several social-psych theories, particularly those in the Theory of Planned Behavior (TPB) tradition—has emerged as a definitive winner. In addition, little research has been done that compares influential drivers across practices, though called for by Reimer et al. (2014). This analysis takes an exploratory approach and compares the relative importance of a wide range of predictors of BMP use from the literature, using the meta-analyses as a guide, on the use of five specific BPMs: pre-sidedress nitrate tests (PSNTs), nutrient maps, variable rate nitrogen (N), variable rate phosphorus (P) or potassium (K), and cover crops. The goal is to identify the common drivers of BMP use and assess their relative importance. It also explores the use of practice knowledge as a predictor of BMP use that has been underutilized in previous research. Rather than being a meta-analysis like previous studies, this one uses data from a single four-state farmer survey. 67 APPROACHES TO UNDERSTANDING BMP ADOPTION/USE Two main conceptual approaches have been used to explain farmers’ use of BMPs: social-psychological theories, and the adoption-diffusion model. Over the years neither approach has displaced the other, though the adoption-diffusion model is not as popular for agricultural research as it once was (Ruttan 1996). This analysis does not take either approach as the a priori model to test, but rather includes variables commonly used in analyses based on both approaches in order to compare their relative importance along with common control variables. These two approaches also offer different ways to think about the role of knowledge in BMP use. Social-Psychological Theories One family of theories from social-psychology has become a very common conceptual approach for understanding farmer adoption of conservation practices. This family consists of the theory of reasoned action (TRA), the theory of planned behavior (TPB), and the reasoned action approach (RAA) (Ajzen 1991; Fishbein and Ajzen 2015). These three theories are related and began with the theory of reasoned action (TRA). In the TRA attitude towards the behavior and subjective norms—the perceived expectations of others–drive behavioral intention, in turn drives behavior. The use of behavioral intention as a mediator between attitudes, norms and behavior is a distinguishing feature the TRA and the theories derived from it. The theory of planned behavior (TPB) is an extension of the TRA that adds perceived behavioral control (PBC) as a predictor of behavioral intention and actual behavior (Madden et al. 1992 highlight the differences). This addition incorporates a contextual element to the theory and improves its predictive power, especially for behaviors that have greater actual or perceived external constraints (Ajzen 1991; Madden et al. 1992). The reasoned action approach (RAA) is the most recent incantation of this family of theories and extends the TPB by adding “individual,” 68 “social,” and “information” background factors, along with a new tier of belief variables and an actual control variable (Fishbein and Ajzen 2015). The belief components underlie the attitude, subjective norms and perceived behavioral control variables, and typically measures of the different types of beliefs that are used to measure attitudes, subjective norms and perceived behavioral control using factor analysis (Fishbein and Ajzen 2015). Adoption-Diffusion Model The adoption-diffusion model describes the process through which a new idea, technology, or practice (i.e., innovation) spreads through a social system, and is ultimately adopted (or not) by individuals over time (Rogers 2003). Originally developed in large part to understand the spread and adoption of new agricultural technologies (Rogers 1958; Ryan 1948; Ryan and Gross 1943), it has been applied to a wide range of topics and utilized by a number of different disciplinary fields, largely through the work of Everett Rogers (Rogers 2003). Rogers (2003:5) describes diffusion as “the process in which an innovation is communicated through certain channels over time among the members of a social system.” It is a macro process (as opposed to the micro-level adoption decision) that is heavily dependent on communication, and where the newness of the innovation and its effects creates uncertainty about the innovation (Rogers 2003). Over time the innovation is communicated to more people and uncertainty is reduced through greater information, sometimes resulting in the full adoption of the innovation in the population (Rogers 2003). The result of diffusion (and the adoption process that results) makes it a form of social change, in that it can alter the structure and function of the social system (Rogers 2003). As diffusion happens, individuals (or other decision-making entities) gain information about the innovation and ultimately decide to use it or not use it. The “innovation-decision 69 process,” as described by Rogers (2003:14), “is essentially an information-seeking and information-processing activity in which an individual is motivated to reduce uncertainty about the advantages and disadvantages of the innovation.” The Potential Importance of Knowledge Knowledge is included implicitly and explicitly in the above discussed theories/models, but it is rarely included in empirical analyses. One potential reason for this is that “knowledge” can be referenced with a wide variety of other terms depending on how subjectively one is willing to consider it. If we are interested in what a participant “knows” then we must extend our concept of “knowledge” to include what researchers might typically term beliefs, awareness, and even perceptions, because we are interested in how the subject identifies the parameters of the world in which they operate (Carolan 2006). “Factual knowledge” is included outright in the TRA, as described by Kaiser, Wölfing, and Fuhrer (1999), as the driver of attitude towards the behavior. Using a broad conceptualization of knowledge, the TPB has perceived behavior control in it, as does RAA along with knowledge as an information background factor and the three belief components, which are subjective knowledge measures. The role of knowledge is more explicit in the adoption-diffusion model, where the spread of information through communication channels is a key aspect of the diffusion process and knowledge of the innovation is the first stage of the innovation-decision process. Rogers (2003) further breaks down knowledge of the innovation in to three types: 1) Awareness-knowledge, which is simply awareness that the innovation exists. 2) How-to knowledge, which is knowledge of how to correctly use and/or implement the innovation. 70 3) Principles-knowledge, which is knowledge of the systems and processes that underlay how the innovation works. These types of knowledge, along with a couple other components of the adoption- diffusion model are notably similar to the four types of knowledge proposed by Kaiser and Fuhrer (2003) in the social-psych tradition. They posit that the role of knowledge in explaining ecological behaviors has been underestimated in past research because either education is used as a proxy for knowledge, different forms of knowledge are not considered simultaneously, and because knowledge is necessary but not sufficient for behavior, since other factors can constrain action even with sufficient knowledge. Kaiser and Fuhrer’s (2003) four types of knowledge are: 1) Declarative knowledge is knowledge of environmental processes or how an environmental system works. 2) Procedural knowledge is knowledge of how to implement a behavior or achieve a particular goal. 3) Effectiveness knowledge is knowledge of the relative effectiveness of a given behavior to achieving the desired outcome. 4) Social knowledge is knowledge of social norms and expectations (“conventional social norms”), as well as personal, “moral social norms.” Figure 3.1 shows how Kaiser and Fuhrer incorporate their four types of knowledge into a version of the TPB. However, they leave out perceived behavioral control, though they mention the importance of external constraints on behavior even when sufficient knowledge is present in the text. Presumably, perceived behavioral control would predict intention and ecological behavior in their model, in keeping with their theoretical foundations in TPB. 71 Figure 3.1. Kaiser and Fuhrer’s (2003:606) conceptual model of how know knowledge types influence each other and ultimately ecological behavior. There are clear correlations between Rogers’ principles-knowledge and Kaiser and Fuhrer’s declarative knowledge, as well as between how-to knowledge and procedural knowledge. Kaiser and Fuhrer’s effectiveness knowledge and social knowledge seem to both be included in Rogers’ relative advantage component of perceived innovation attributes. The perceived innovation attributes are not discussed overtly as knowledge, but they do fit in the broad conceptualization of knowledge discussed above. How knowledge has been operationalized and included in empirical analyses related to agriculture has varied, but the most common form is some type of awareness knowledge— awareness of problems (also called ‘concern’) and practices (e.g., Duff et al. 1992; Ervin and Ervin 1982; Napier, Thraen, and Gore 1984; Reimer and Prokopy 2012; Ulrich-Schad et al. 2017), and exposure to information, communication networks, and education (e.g., Duff et al. 1992; Dunn et al. 2016; Ervin and Ervin 1982; Napier et al. 1984; Ulrich-Schad et al. 2017). The 72 variable categories used in the meta-analyses by Prokopy et al. (2008) and Baumgart-Getz et al. (2012), highlight the use of principles-knowledge or declarative knowledge, which describes their “cause” “consequences” and “knowledge” categories, while their “program” category is either awareness knowledge or perhaps procedural knowledge of conservation programs (see Table 3.2). Figure 3.2. Combining perceived practice characteristics from the adoption-diffusion model with the Reasoned Action Approach by Reimer, Weinkauf, et al. (2012). Knowledge of practice characteristics has also been used in empirical studies, drawing from the adoption-diffusion approach. Reimer, Weinkauf et al. (2012) merge the five perceived practice characteristics of Rogers (2003) with the RAA in their approach to understanding farmers’ reasons for adopting or not adopting best management practices in Indiana (see Figure 3.2). They find that relative advantage and compatibility of BMPs (or lack thereof) to be the most important practice characteristics motivating or discouraging practice adoption. Arbuckle and Roesch-McNally (2015) use this combined conceptual approach to study cover crop adoption in Iowa qualitatively and quantitatively. Similarly to Reimer, Weinkauf et al. (2012), 73 Arbuckle and Roesch-McNally (2015) find variables related to the relative advantage and compatibility (or lack thereof) of cover crops to be important drivers of cover crop adoption or non-adoption. Rogers’ (2003:173) observed that “few diffusion investigations deal with how-to knowledge, although it must be a fundamental variable in the innovation-decision process,” and an extensive search of the literature supports this. The only recent study I could find that considered how-to BMP knowledge was Vitale et al. (2011), who included a knowledge of conservation tillage practices measure in their survey of Oklahoma farmers. They found conservation tillage knowledge to have a significant, positive effect on the use of reduced tillage and conservation tillage and a significant, negative effect on the use of conventional tillage. Nowak and Korsching (1985) long ago recognized the importance of knowledge, learning and experience in the adoption practice, pointing to the example of farmers who had been using conservation tillage longer retained more crop residue than more recent adopters. 74 Table 3.1. Overview of conservation/best management practice adoption meta-analyses. Knowler and Bradshaw (2007) Prokopy et al. (2008) Number of studies included Number of analyses included 23 31 55 at least 329 Baumgart-Getz, Prokopy and Floress (2012) 46 (published and unpublished) [not provided] Type of analysis included quantitative quantitative with significant results at quantitative with "enough information to calculate 0.05 level effect size" Wauters and Mathijs (2014) 38 69 quantitative Number of predictor variables included (after reduction) Year range of included studies 46 34 30 87 1984-2002 1982-2007 1982-2007 1982-2006 Study location world-wide US at multiple scales US at multiple scales Outcome variable criteria "farm-level adoption of a actual adoption of one number of soil management and wider conservation practices consistent with conservation or more practices expected to lead directly to water agriculture" quality improvement actual BMP adoption developed country (most in US) at multiple scales use of soil conservation practices broadly defined, many on conservation tillage of some type Method of analysis vote-count vote-count mixed effects model using Hedges d vote-count Sub-analyses included: Practice / category Analysis method Predictor variable type Region Journal quality X X X X X X 75 X X X PREVIOUS RESEARCH ON DRIVERS OF BMP USE As discussed above, the inclusion of knowledge, especially how-to or procedural knowledge, is ripe for further explanation, but is not expected to be the only factor of importance in predicting farmers’ use/adoption of conservation and best management practices. Much of the more recent work on this topic, using both adoption-diffusion and social-psych approaches among others, has been analyzed in several recent meta-analyses (Baumgart-Getz et al. 2012; Knowler and Bradshaw 2007; Prokopy et al. 2008; Wauters and Mathijs 2014)16. However, the results of these analyses have found little consistency in sign and significance of the examined predictors. These analyses do provide a useful starting point for the present analysis though. These four meta-analyses (summarized in Table 3.1) include 23 to 55 studies from the early 1980s through the mid-2000s that used actual adoption (as opposed to intended adoption) of a range of BMPs as their dependent variables. Knowler and Bradshaw (2007) include studies from all parts of the world, making it less directly applicable than the other three studies, while Prokopy et al. (2008) and Baumgart-Getz et al. (2012) restrict their sample to the US and Wauters and Mathijs (2014) limit theirs to studies of “developed” countries. Knowler and Bradshaw (2007), Prokopy et al. (2008), and Wauters and Mathijs (2014) all use vote-count methodology, while Baumgart-Getz et al. (2012) use a mixed effects model. This later analysis method provides effect sizes for the variables considered, while vote-count methodology counts the models for which a given variable has a significant positive effect, significant negative effect, or insignificant effect. The vote-count analyses find a definite overall trend towards insignificance, though some variables are more often significant than not. The more sophisticated mixed effect model finds a large portion of 16 A very recent meta-analysis (Liu et al. 2018) is not included because they include studies that used program participation as the outcome of interest, making their results not directly comparable to the other studies referenced here. 76 variables with very small effect sizes and/or non-significant effects, but it allows for assessment of the relative importance/influence of different variables in a way that vote-count methods do not. Connecting Variables and Theories While there is overlap in the types of variables that are used to operationalize the theoretical approaches discussed above, the different approaches emphasize different categories of variables. The social-psych theories naturally call for variables that measure respondents’ beliefs, attitudes and values as they relate to the environment, other entities, and the behavior in question (Baumgart-Getz et al. 2012; Prokopy et al. 2008). Beliefs, attitudes, and values are commonly measured as multi-item latent variables (see for example Price and Leviston 2014; Ulrich-Schad et al. 2017; Werner et al. 2017). While the adoption-diffusion model includes attitudes towards the practice as an important predictor of the adoption-decision process, it also focuses on the characteristics of the practice, and stresses the importance of education, and information sources (Nowak 1987; Nowak and Korsching 1985; Prokopy et al. 2008; Rogers 2003). Some, though not recently, have also operationalized farmers’ innovativeness (Rollins 1993). METHODS AND ANALYSIS The goal of this analysis is to look for the common drivers of BMP use across the five practices considered, and to assess the relative importance of significant drivers. This analysis does not take either the adoption-diffusion model or any of the social-psych approaches as an a priori model to test, but rather includes variables commonly used in analyses based on both approaches in order to compare their relative importance along with common control variables. I 77 use structural equation modeling with latent variables (SEMLV) with binary BMP use outcome variables, using 2017 farmer survey data from four states. Sample The data used comes from a 2017 survey of corn growers in Illinois, Indiana, Michigan, and Ohio. The questionnaire asked farmers about the 2016 growing season and included questions on their overall operation and cropping practices, specific practices used on their largest corn field, the frequency of their information source use and trust of those sources, their attitudes and values towards farming and the environment, and their demographic characteristics. Surveys were sent to 10,582 corn growers using a two-wave, modified Dillman (Dillman, Smyth, and Christian 2014) mailing procedure, with a response rate of 31 percent, of which 2289 surveys were at least partially filled out (22 percent useable response rate). After dropping cases with missing values on key variables, the working sample was 1958 cases. Case-wise deletion of missing values on independent variables was required by the analysis method, leaving a final sample of approximately 814 cases17 (a few less on the dependent variables) (see Appendix A for details on data cleaning). The final sample response numbers by state are: IL = 282, IN = 212, MI = 114, OH = 256. As shown in Table 3.2, the sample contains more full-time farmers with larger farms than average. While this means that the sample is not representative of all farms in these states, it may be more representative of corn and soybean farms, and corn-soybean-wheat growing farms as these have larger average sizes. The farmers in this sample do represent those who can have the greatest environmental impact by using BMPs. 17 There are fewer than 814 observations on some of the latent variable measures as noted in Table 3.5. 78 Table 3.2. Comparison between final sample (n=814) and 2012 Census of Agriculture on key demographic variables by state. 2016 Survey--Final Sample 2012 Census of Agriculture† IL IN MI OH Years farming Age in years Percent working X days off-farm 0 days 1 to 49 days 50 to 99 days 100 to 199 days 200 or more days 29.0 58.2 54.8 18.6 3.4 8.0 15.2 Percent of household income from farming less than 25% 25 to 49% 50 to 74% 75 to 100% 10.6 17.9 20.5 51.0 1192.1 Average cropland per farm Average acres of corn-soy operation* Average acres of corn-soy-wheat operation* †Data except for cropland for principle operator *Only including farms that grow these crops Variables 28.7 57.0 59.1 13.1 4.0 6.6 17.2 15.2 15.7 18.7 50.5 26.0 55.3 30.0 58.9 58.7 15.6 3.7 4.6 54.1 13.9 4.1 9.4 17.4 18.4 17.4 11.0 16.5 55.0 18.0 23.8 20.1 38.1 1289.4 1515.6 935.0 IL 27.3 57.8 IN 26.0 55.8 MI OH 26.0 57.6 25.9 56.8 42.4 35.5 41.3 36.7 9.4 4.0 7.3 7.6 3.4 7.9 7.9 4.0 8.1 8.1 3.8 7.9 36.9 45.5 38.7 43.5 57.1 10.1 13.3 19.6 65.3 70.5 68.1 9.5 10.6 14.6 7.8 9.2 12.4 9.6 10.2 12.1 351.3 257.7 171.7 167.3 591.8 684.1 497.5 567.6 336.6 423.0 331.4 385.8 There are five outcome variables: use of pre-sidedress nitrate tests (PSNT), use of nutrient maps to inform fertilizer decisions, use of variable rate nitrogen (N), use of variable rate phosphorus (P) or potassium (K), and use of winter cover crops. PSNTs and nutrient maps are both decision aids for fertilizer applications, while the variable rate N and variable rate P or K are two examples of precision fertilizer application technology; cover crops represent a biological soil and nutrient management practice. The questions about these practices had five response categories18, and the variables used were created by combining the “yes, regularly do this,” with the “yes, sometimes do this” responses into a binary variable reflecting current use of 18 The three response categories that were not used are: “Used to, but no longer do,” “Never used, but might,” and “Never used, don’t want to.” 79 the practices19. The choice to use the collapsed, binary variables as opposed to the full range of responses was made based on my interest in actual practice use, rather than past use or possible future use. The above mentioned meta-analyses were used as guides for the present analysis. The categories and sub-categories of Prokopy et al. (2008) and Baumgart-Getz et al. (2012) were combined (Table 3.3)20, and the longer variable list from Wauters and Mathijs (2014) was used as examples of variables for each categories. These examples informed the selection and calculation of variables from those available to represent each sub-category in Table 3.3 as much as possible. This resulted in 53 independent variables, including two latent variables and two variables that are excluded from the analysis to serve as reference categories (see Table 3.4), representing 30 of the 39 sub-categories. Two common variables were not included in the analysis for technical reasons though they were in the dataset: age and operator gender. Age was highly correlated with farming experience (> 0.75) so both could not be included without unacceptably high multicollinearity. Operator gender was left out due to the very small number of women in the overall sample (1.5 percent) and in the final sample (0.7 percent), which resulted in problematically low variation in the operator gender variable. The independent variables used in this analysis are described in Table 3.4. To facilitate analysis and discussion, the variables have been re-organized into six categories, rather than the four shown in Table 3.3; this was done to group more similar variables together, in some cases based on the actual measures being used in the present analysis. Changes were made to all the 19 This is a useful and simple measure, though it imperfectly reflects temporal variability in use and does not reflect the area on which the practice is used (spatial extent) at all. 20 Many of the variable categories and definitions are the same between these two studies, which made combining them straightforward. 80 categories shown in Table 3.3, though the new “attitudes and values,” and “environmental awareness” categories are similar those of Prokopy et al. (2008) and Baumgart-Getz et al. (2012). The most substantial change was dividing the “capacity” variables into background factors, social farm characteristics, and information categories; the sub-category names are the same as in Table 3.3. Table 3.3 Combined categories used in Prokopy et al. (2008) and Baumgart-Getz, Prokopy, and Floress (2012). Categories and sub-category Capacity Description Acres / farm size Number of acres farmed Age Capital Diversity Education Farmer age Measure of investment into farm (excluding acres) Measures that capture diversity of farm operation Farmer education Extension Training Subgroup of EDUCATION includes only extension training Formal education Subgroup of EDUCATION- includes only formal education Farming experience Years farming Income Information Labor Networking Agency Business Local Measures of wealth such as income, crop value, etc. Access to and quality of information Measures of increased labor available to the farm Overall measure of networking capacity Subgroup of NETWORKING: Connectivity to and familiarity with agency personnel and procedures Subgroup of NETWORKING: Measures of networking capacity in the agri- business sector Subgroup of NETWORKING: Interacts with neighboring farms as well as any grass roots organization University Extension Subgroup of NETWORKING: Exposure to a university extension office Institutional / Ownership Type Farm structure or organization measures e.g. individually owned versus Tenure Attitudes corporate-owned farms Whether operator owns farmland Overall attitude The attitude category as a whole Adoption payments Farmer receiving payments for participating in conservation programs Environmental Importance individual places on environmental quality Profitability of practice Farmer places financial gain as primary purpose of farm Heritage Farm will be taken over by a family member Quality of environment Farmer’s perception of the current quality of the environment Regulatory Farmer feels government can/should regulate agriculture 81 Table 3.3. (cont’d). Risk Scientific Environmental awareness A measure of risk averseness Values scientific research Overall environmental awareness The environmental awareness category as a whole Cause Consequences Knowledge Program Understanding how agriculture can impact environmental quality Understanding the consequences of a degraded system Knowledge of general terms or facts related to environmental quality Knowledge of nonpoint source programs or efforts Farm characteristics Animal Grain Operator gender Other farm types River Slope Soil quality Livestock or dairy operations Primarily a grain farm Primary operator is male Other/unspecified type of farm Farm located near a stream or in a river bottom Fields are hilly or have steeper slopes Measures of soil quality Note: Sub-categories in italics are those which are not included in the present study. Age and gender were not used for technical reasons; the rest did not have relevant measures in the data used. Some variables were not easily categorized, such as crop insurance use, which could reasonably go in the social farm characteristics category, but which was kept in the attitudes and values category as a measure of risk aversion. Similarly, participation in conservation programs, either in general or working land programs specifically, could be included in the attitudes and values category or the environmental awareness categories as they are in Table 3.3. However, they have been included in the social farm characteristics category for this analysis because farmers participate in conservation programs for a range of reasons and also use conservation practices without participating in a program (Reimer and Prokopy 2014); thus program participation does not inherently reflect a particular attitude, value or awareness. 82 Table 3.4. Variables used in analysis, their definition and scale, grouped by author-defined categories, and using sub-categories from the literature. All variables are for 2016. Categories Variable Name Definition Scale Individual Characteristics Farming experience Years farming Years as primary farm decision-maker Continuous Education Education level Education level Categorical with 4 ordered categories: 1 = less than high school, 4 = bachelor's degree or higher College ag courses Took college courses in agriculture Binary Knowledge Knowledge Self-reported knowledge of practices for nutrient Categorical with 5 levels of knowledge: management and soil conservation 1 = nothing at all, 5 = a great deal Income Value of sales Gross value of farm products sold Categorical with 4 ordered categories: Household income Household adjusted gross income Categorical with 6 ordered categories: 1 = less than $25,000, 6 = greater than $500,000 Ag. income Portion (%) of household gross adjusted income Categorical with 4 ordered categories: coming from farming 1 = less than 25%, 4 = greater than 74% 1 = less than $100,000, 4 = greater than $1 million Farm Characteristics—Physical Acres / farm size Diversity Animal Cropland area Corn-soy ratio Animal farm Cropland planted in 1000 acres Corn acres to soy acres ratio (from a3) Dairy production or other livestock (including poultry) account for more than 10% of farm revenue Continuous Continuous Binary Grain Grain farm Field or grain crops account for more than 10% of Binary Other farm types Other farm type farm revenue Something other than grain or livestock provides more than 10% of revenue—includes fruit, nut and vegetable crops, and flowers, ornamentals and live plants Binary Slope Field slope Slope of largest, owned corn field Categorical with 4 ordered categories of slope: 1 = flat (0-2% average grade), 4 = severely hilly (10% or more average grade) 83 Table 3.4. (cont’d). Soil quality Soy yield Corn yield Loam Clay Average soy yield Average corn yield Main soil texture of largest, owned corn field is Binary loam Main soil texture of largest, owned corn field is Binary clay Continuous Continuous Clay-loam Main soil texture of largest, owned corn field is Binary clay-loam (reference) Silty-loam Main soil texture of largest, owned corn field is Binary silty-loam Sandy-loam Main soil texture of largest, owned corn field is Binary sandy-loam Other soil type Main soil texture of largest, owned corn field is Binary "other"—includes sand and silt textures Farm Characteristics—Social Tenure Labor Cropland owned Percent of operation cropland that respondent Continuous Days worked off farm Days worked off farm owns Categorical with 5 ordered categories: 1 = none, 5 = 200 days or more Number of employees Number of family and non-family employees, not Continuous Ownership Type/structure Sole proprietorship Sole proprietorship (reference) including the respondent Partnership LLC S-corporation C-corporation Partnership Limited liability corporation Small corporation (subchapter S) Corporation (subchapter C) Binary Binary Binary Binary Binary Program Conservation program Participate in any kind of conservation program Binary Other business structure Other business structure--mostly combinations of Binary the above Adoption payments Working land program Participates in a working land program (EQIP or Binary participation participation CSP) 84 Table 3.4. (cont’d). Information Information access Info index Sum of frequency of use and level of trust for 9 Continuous info sources Networking Associations info source Frequency of use of growers associations and Categorical with 5 levels of frequency: conferences as info source 1 = never, 5 = once a day Business Chemical dealers info Frequency of use of chem dealers as info source Categorical with 5 levels of frequency: source 1 = never, 5 = once a day Seed dealers info source Frequency of use of seed dealers as info source Categorical with 5 levels of frequency: 1 = never, 5 = once a day Crop consultants info Frequency of use of independent consultants as Categorical with 5 levels of frequency: source info source 1 = never, 5 = once a day Local Other farmers info Frequency of use of family and other farmers as Categorical with 5 levels of frequency: source info source 1 = never, 5 = once a day University Extension Campus extension info Frequency of use of campus extension as info Categorical with 5 levels of frequency: source source 1 = never, 5 = once a day Local extension info Frequency of use of local extension educators Categorical with 5 levels of frequency: Attitudes and Values source 1 = never, 5 = once a day Importance of environment Look after environment Importance of looking after the environment Categorical with 5 levels of importance: Hunting and fishing Importance of maintaining good hunting, fishing, Categorical with 5 levels of importance: environment hiking 1 = low, 5 = high 1 = low, 5 = high Profitability of practice Economic attitude Economic attitude latent variable Latent Wealth Profit Income Importance of building up wealth and family Categorical with 5 levels of importance: assets (used in latent variable) 1 = low, 5 = high Importance of maximizing farm/company profit Categorical with 5 levels of importance: (used in latent variable) 1 = low, 5 = high Importance of earning a high income (used in Categorical with 5 levels of importance: latent variable) 1 = low, 5 = high Heritage Passing on good land Importance of passing on land in good condition Categorical with 5 levels of importance: Farming tradition Importance of keeping family farming tradition Categorical with 5 levels of importance: alive 1 = low, 5 = high Passing on farm Child or other relative expected to take over farm Binary 1 = low, 5 = high after respondent retires 85 Table 3.4. (cont’d). Risk Debt aversion Importance of minimizing debt Categorical with 5 levels of importance: 1 = low, 5 = high Crop insurance use Percent of corn and soy acres covered by crop Continuous Environmental Awareness insurance Cause Concern ag. contributes to water problems Concern agriculture contributes to water problems latent variable Latent Groundwater contribution Algal bloom contribution Concern that agriculture contributes to Categorical with 5 levels of concern: groundwater contamination (used in latent variable) Concern that agriculture contributes to algal blooms in lakes (used in latent variable) 1 = low, 5 = high Categorical with 5 levels of concern: 1 = low, 5 = high Hypoxia contribution Concern that agriculture contributes to hypoxia Categorical with 5 levels of concern: ("dead zones") in oceans (used in latent variable) 1 = low, 5 = high Erosion contribution Concern that agriculture contributes to soil Categorical with 5 levels of concern: erosion (used in latent variable) 1 = low, 5 = high Consequences Erosion impacts Concern about on-farm impacts from soil erosion Categorical with 5 levels of concern: Pest resistances impacts Concern about on-farm impacts from weeds and Categorical with 5 levels of concern: insects developing resistance to pesticides 1 = low, 5 = high Extreme weather impacts Concern about on-farm impacts from droughts, Categorical with 5 levels of concern: floods, and other extreme weather events 1 = low, 5 = high Warming impacts Concern about on-farm impacts from warmer Categorical with 5 levels of concern: temperatures 1 = low, 5 = high 1 = low, 5 = high 86 Analysis Structural equation modeling with latent variables was used for this analysis and was executed in Mplus8. A separate model was run for each of the five outcome variables using the same independent variables. Due to the use of binary outcome variables and categorical variables in the latent variables, a robust, diagonally weighted least squares, means and variance adjusted estimator (specifically the WLSMV estimator in Mplus8) was used. Due to the evidence of state- level differences for three of the five practice-use variables a “complex” analysis is specified in Mplus, which adjusts the standard errors and chi-square tests of model fit for clustering by state. See Appendix B for additional explanation and justification for these estimator and analysis type decisions. Use of the WLSMV estimators results in estimated probit coefficients. The probit coefficients can be thought of as reflecting the rate of change in probability of practice use over different values of the variable, and significance indicates if the slope is significantly non-zero. The probit coefficients themselves do not indicate anything about the probability of practice use, but the probability of practice use can be calculated from the intercept21, probit coefficients, and specified values for each variable using the following equation (Muthén, Muthén, and Asparouhov 2016) where F is the standard normal distribution function22 (see Appendix B for more explanation): P(ui = 1 | xi) = F(β0 + β1 x1 + β2 x2 + … + βi xi) By calculating the probabilities of practice use at different values of a variable of interest while holding all other variables at a constant value the total effect of the variable on practice use 21 The intercept for a binary outcome variable is the negative threshold value that Mplus reports. See Appendix B for more information about the meaning of threshold values with categorical outcomes and how the WLSMV estimator calculates them. 22 A z score is calculated by the intercept, coefficients and specified variable values, then a left sided z table is used to determine the probability of F(z)—this is equal to the area under the curve to the left of the z value. 87 can be evaluated in a standardized way that is comparable across variables (Muthén et al. 2016). The “baseline” value used for each variable was the mean value for continuous variables (including latent variables) and the mode value for categorical variables, resulting in the probability of practice use by an “average” farmer. Due to the more meaningful interpretation of the probabilities versus the probit coefficients, the probabilities are the focus on the following section. FINDINGS General Approach to Results Table 3.5 provides the descriptive statistics for the analysis sample, while tables 3.6, 3.7 and 3.8 present select probit results from the five models (see Appendix C for full probit results). Conceptually similar practices have been paired for sake of space and comparability, and only effects significant or marginally significant for at least one of the practices are included in these tables. PSNT and nutrient maps are paired because they are both decision support tools that are independent of the crop in a given year. Variable rate N and variable rate P or K are paired because they are both using similar technology. Cover crops is by itself since it is the most different from the other practices included, by virtue of being a practice that is not directly related to target crop production and requiring additional activities to the crop growing cycle to plant it in the fall and then kill it before planting the regular crop in the spring. The sign of significant variables for all five practices are summarized in Table 3.9. One of the most notable results is the distinct lack of overlap in significant variables between the paired practices especially but also across all the models. In fact, of the fifty-three predictive 88 variables in the models only one, use of independent crop consultants as an info source, is significant in all five models. Table 3.5. Analysis descriptives (n = 814 unless otherwise noted). Mode Std Dev Minimum Maximum Mean 0.182 0.833 0.278 0.643 0.355 0 1 0 1 0 0.386 0.373 0.448 0.480 0.479 28.814 37 14.128 2.937 0.434 3.550 2.732 3.173 2.996 1.182 1.295 0.221 0.966 0.049 1.483 2 0 3 3 3 4 1 1 0 1 0 1 58.825 179.712 55 170 0.053 0.076 0.461 0.200 0.140 0.070 42.552 2.138 1.373 0.636 0.087 0.128 0.064 0.076 0.009 0.420 89 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0.852 0.496 0.805 0.872 1.035 1.120 1.194 1.574 0.415 0.182 0.216 0.646 9.082 34.492 0.224 0.265 0.499 0.400 0.347 0.255 33.640 1.554 4.227 0.481 0.282 0.334 0.245 0.265 0.092 0.494 PSNT Nutrient maps Variable rate N Variable rate P or K Cover crops Individual Characteristics Years farming Education level College ag courses Knowledge Value of sales Household income Ag. income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Grain farm Other farm type Field slope Soy yield Corn yield Loam Clay Clay-loam Silty-loam Sandy-loam Other soil type Farm Characteristics—Social Cropland owned Days worked off farm Number of employees Sole proprietorship Partnership LLC S-corporation C-corporation Other business structure Conservation program participation 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 1.000 1.000 1.000 1.000 0.050 0.012 0.000 0.000 0.000 1.000 1.000 1.000 1.000 1.000 1.000 65.000 4.000 1.000 5.000 4.000 6.000 4.000 12.500 25.000 1.000 1.000 1.000 4.000 20.000 56.250 82.000 270.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 1.000 1.000 1.000 1.000 1.000 100.000 5.000 68.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 Table 3.5. (cont’d). Working land program participation Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values Look after environment Hunting and fishing environment Economic attitude Income (n=813) Wealth (n=812) Profit Passing on good land Farming tradition Passing on farm Debt aversion Crop insurance use Environmental Awareness Concern ag. contributes to water problems Groundwater contribution (n=802) Algal bloom contribution (n=803) Hypoxia contribution (n=799) Erosion contribution (n=805) Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts 0.157 51.996 1.813 2.924 2.802 1.819 3.178 1.700 1.736 4.290 2.763 3.696 4.145 4.348 4.576 4.504 0.764 4.216 0 55 2 3 3 1 3 1 2 4 3 4 4 5 5 5 1 5 7.716 0.758 0.708 0.747 0.977 0.951 0.741 0.675 0.742 1.256 0.870 0.830 0.712 0.693 0.785 0.425 0.819 82.371 100 35.835 3.608 3.420 3.185 4.086 4.161 4.353 3.565 3.061 4 3 3 4 5 5 4 3 1.058 1.088 1.192 0.887 1.061 0.893 1.242 1.210 0.364 0.000 1.000 28.000 77.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.000 1.000 0.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 5.000 5.000 5.000 5.000 5.000 5.000 4.000 5.000 5.000 5.000 5.000 5.000 5.000 5.000 1.000 5.000 100.000 5.000 5.000 5.000 5.000 5.000 5.000 5.000 5.000 Twelve variables were not significant for any of the practices: years farming, ag income, grain farm, soy yield, clay soil type, other soil type, cropland owned, number of employees, s- corporation, c-corporation, campus extension info source, and hunting and fishing environment. Five additional variables were only of marginal significance for any of the five practices: college ag courses, sandy-loam soil, other business structure, passing on good land, and debt aversion. 90 Table 3.6. Select probit regression coefficients and standard errors for PSNT and nutrient maps (n=814). Individual Characteristics Education level College ag courses Knowledge Household income Farm Characteristics—Physical Other farm type Loam Farm Characteristics—Social Partnership Other business structure Information Seed dealers info source Crop consultants info source Local extension info source Attitudes and Values Passing on good land Farming tradition Crop insurance use Environmental Awareness Concern ag. contributes to water problems Erosion impacts Warming impacts Intercept R-Square Chi-Square RMSEA CFI TLI PSNT Estimate (S.E.) Nutrient maps Estimate (S.E.) -0.086 (0.112) 0.125 (0.070)t 0.178 (0.222) 0.083 (0.054) -0.458 (0.285) -0.558 (0.149)*** 0.310 (0.121)* 0.987 (0.584)t 0.035 (0.101) 0.295 (0.037)*** 0.067 (0.074) -0.091 (0.051)t 0.290 (0.185) 0.000 (0.002) 0.060 (0.006)*** 0.094 (0.035)** -0.035 (0.075) -4.325 (1.351)** 0.349 (0.073)*** 679.695*** 0.033 0.839 0.812 -0.084 (0.016)*** 0.271 (0.141)t 0.204 (0.097)* 0.060 (0.031)* -0.671 (0.298)* 0.265 (0.280) 0.206 (0.374) 0.127 (1.008) -0.155 (0.072)* 0.174 (0.056)** -0.129 (0.054)* -0.041 (0.040) -0.082 (0.034)* 0.003 (0.001)t 0.051 (0.015)*** 0.013 (0.077) -0.098 (0.040)* -2.937 (0.695)*** 0.371 (0.121)** 678.584*** 0.033 0.839 0.812 t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 91 Table 3.7. Select probit regression coefficients and standard errors for variable rate N and variable rate P or K (n=814). Variable rate N Estimate (S.E.) Variable rate P or K Estimate (S.E.) 0.034 (0.070) 0.123 (0.105) 0.045 (0.068) 0.156 (0.060)* -0.415 (0.314) 0.004 (0.000)*** -0.375 (0.148)* 0.091 (0.132) 0.000 (0.041) 0.306 (0.074)*** -0.029 (0.131) 0.236 (0.063)*** 0.011 (0.002)*** -0.013 (0.124) 0.134 (0.059)* 0.018 (0.107) 0.085 (0.035)* 0.251 (0.082)** 0.009 (0.054) -0.001 (0.001) 0.039 (0.012)** -2.878 (0.348)*** 0.237 (0.049)*** 679.410*** 0.033 0.840 0.813 -0.163 (0.094)t 0.244 (0.062)*** 0.198 (0.077)* 0.026 (0.102) -0.456 (0.260)t 0.006 (0.003)* -0.100 (0.099) 0.328 (0.174)t -0.019 (0.006)** 0.117 (0.118) 0.279 (0.073)*** -0.108 (0.188) 0.019 (0.021) 0.226 (0.044)*** 0.106 (0.039)** -0.080 (0.037)* 0.008 (0.007) 0.070 (0.090) -0.144 (0.076)t 0.003 (0.001)*** 0.002 (0.032) -2.758 (0.967)** 0.310 (0.028)*** 677.040*** 0.033 0.840 0.813 Individual Characteristics Education level Knowledge Value of sales Farm Characteristics—Physical Cropland area Other farm type Corn yield Silty-loam Sandy-loam Farm Characteristics—Social Days worked off farm LLC Conservation program participation Working land program participation Information Info index Chemical dealers info source Crop consultants info source Other farmers info source Attitudes and Values Economic attitude Passing on farm Debt aversion Crop insurance use Environmental Awareness Extreme weather impacts Intercept R-Square Chi-Square RMSEA CFI TLI t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 92 Table 3.8. Select probit regression coefficients and standard errors for cover crops (n=814). Cover crops Estimate (S.E.) 0.111 (0.064)t 0.257 (0.088)** -0.086 (0.044)t -0.035 (0.016)* 0.420 (0.164)* 0.338 (0.144)* 0.065 (0.022)** -0.006 (0.002)* -0.227 (0.048)*** 0.208 (0.059)*** 0.631 (0.127)*** 0.136 (0.033)*** 0.129 (0.049)** 0.292 (0.039)*** 0.248 (0.097)* -0.126 (0.046)** 0.040 (0.024)t 0.227 (0.017)*** -0.138 (0.051)** -2.726 (1.106)* 0.338 (0.02)*** 681.087*** 0.033 0.841 0.814 Individual Characteristics College ag courses Knowledge Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Other farm type Field slope Corn yield Silty-loam Farm Characteristics—Social LLC Working land program participation Information Associations info source Crop consultants info source Local extension info source Attitudes and Values Look after environment Economic attitude Environmental Awareness Concern ag. contributes to water problems Erosion impacts Pest resistances impacts Intercept R-Square Chi-Square RMSEA CFI TLI t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 93 Table 3.9. Sign of significant variables in analyses. Individual Characteristics Education level Knowledge Value of sales Household income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Other farm type Field slope Corn yield Loam Silty-loam Farm Characteristics—Social Days worked off farm Partnership LLC Conservation program participation Working land program participation PSNT Nutrient maps Variable rate N Variable rate P or K Cover crops - + - + + - 94 + + - + + + + + - + + - + + + - - + + Table 3.9. (cont’d) Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Local extension info source Attitudes and Values Look after environment Economic attitude Farming tradition Passing on farm Crop insurance use Environmental Awareness Concern ag. contributes to water problems Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts + + + + indicates positive effect and significant at p < 0.05 or better - indicates negative effect and significant at p < 0.05 or better + + + + + - + - - + - 95 + + - + + + + + - + - Figure 3.3 shows the predicted probability of practice use by an “average” farmer, that is, the probability of practice use when the mean or mode (for continuous and categorical variables respectively) of each variable is used in the equation discussed in the analysis section. These values represent the likelihood of a conceptually “average” of typical farmer using each practice, not the portion of the sample who reported using each practice (see Figure 3.5 in Appendix B for reported practice use by state). 77.0% 54.0% 46.0% e g a r e v a y b n o i t p o d a e c i t c a r p f o y t i l i b a b o r P 90% 80% 70% 60% r e m r a f 50% 40% 30% 20% 10% 0% 18.7% PSNT Nutrient maps Variable rate N Variable rate P or K Figure 3.3. Probability of practice use by a typical farmer. Results of PSNT Model Five variables are found to be significant predictors of PSNT use. As shown by the probabilities in Table 3.10, farmers with loam soil type on their largest corn field were 11 percent less likely to use PSNTs than farmers with clay-loam soil type (reference category), and partnerships were 9 percent more likely to use PSNTs than sole proprietorships (reference category). The more frequent use of independent crop consultants as an information source is 96 found to be a very important predictor of PSNT use, with farmers who report consulting with them on a daily basis found to be more than three times more likely to use PSNTs than farmers who never consult with them. Farmers with high concern that agriculture contributes to water problems, and those with high concern over erosion impacts on their farm are both more likely to use PSNT tests than farmers who have low concern on these variables, but only by 6 percent and 8 percent respectively. Table 3.10. Total effects for significant variables on PSNT use in terms of probabilities with all other variables at average. Prob. at min (min value) Prob. at max (max value) Farm Characteristics—Physical Loam (mode = 0) Farm Characteristics—Social Partnership (mode = 0) Information Crop consultants info source (mode = 1) Environmental Awareness Concern ag. contributes to water problems (mean = 3.575) Erosion impacts (mode = 5) 18.7% (0) 18.7% (0) 18.7% (1) 14.7% (1) 10.2% (1) 7.4% (1) 28.1% (1) 61.4% (5) 20.9% (5) 18.7% (5) Total effect -11.3% 9.4% 42.7% 6.2% 8.5% Note: Total effect is calculated as the probability at the variable's max value minus the probability at the variable's minimum value. Results of Nutrient Map Model Table 3.11 shows the total effects of the significant variables on nutrient map use. Farmers with higher levels of education are predicted to use nutrient maps less than their less educated counter-parts, with farmers with a bachelor’s degree or higher being about 8% less likely to use them than farmers with less than a high school degree. Knowledge, on the other hand has a positive effect on nutrient map use with a 25 percent total effect, and the second- largest probability at max at 87 percent predicted probability of nutrient map use. Household income has a positive if modest effect on nutrient map use with the difference between 97 household incomes less than $25,000, and greater than $500,000 being only about 9 percent. Other farm type, those that generate more than 10 percent of their production value from something other than grain/field crop or livestock are predicted to be 24% less likely to use nutrient maps than farms not in this category. Among the significant information sources, independent crop consultants are the only one with a positive effect on nutrient map use, while seed dealers and local extension have a negative effect. The difference in predicted probability of nutrient map use between farmers who use seed dealers as an information source daily, and those who never use them is about 19 percent. The difference for local extension is similar at 17 percent. The daily use of independent crop consultants versus non-use results in a 15 percent greater probability of nutrient map use. While this is not the largest total effect on nutrient map use, it notably produces the highest probability of use at 93 percent when all other variables are at their average. Table 3.11. Total effects for significant variables on nutrient map use in terms of probabilities with all other variables at average. Prob. at min (min value) Prob. at max (max value) Total effect Concern ag. contributes to water problems 72.9% (1) 79.1% (5) (mean = 3.575) Warming impacts (mode = 3) 82.6% (1) 70.5% (5) -12.1% Note: Total effect is calculated as the probability at the variable's max value minus the probability at the variable's minimum value. 98 Individual Characteristics Education level (mode = 2) Knowledge (mode = 3) Household income (mode = 3) Farm Characteristics—Physical Other farm type (mode = 0) Information Seed dealers info source (mode = 3) Crop consultants info source (mode = 1) Local extension info source (mode = 2) Attitudes and Values Farming tradition (5) Environmental Awareness 79.4% (1) 62.9% (1) 73.2% (1) 71.6% (4) 87.5% (5) 82.1% (6) 77.0% (0) 52.8% (1) 85.3% (1) 77.0% (1) 80.8% (1) 85.8% (1) 66.6% (5) 92.5% (5) 68.4% (4) 77.0% (5) -7.8% 24.6% 8.9% -24.3% -18.7% 15.5% -12.3% -8.7% 6.2% Farmers that feel it is important to keep the family farming tradition alive were 9 percent less likely to use nutrient maps than those that do not feel it is important to keep that tradition alive. Farmers with high concern that agriculture contributes to water problems were 6 percent more likely to use nutrient maps than those with low concern, but those with high concern that warming was impacting their farm were 12 percent less likely to use nutrient maps than those with low concern. Results of Variable Rate N Model Table 3.12 shows the significant total effects on the use of variable rate N technology. Farmers with larger farms and higher average corn yields are more likely to use variable rate N than those with smaller farms and lower average corn yields by about 22 percent. Farmers with silty-loam soil type on their largest corn field were 14 percent less likely to use variable rate than farmers with clay-loam soil type (reference category). Farmers with an LLC business structure were 12 percent more likely to use variable rate N than farmers in a sole proprietorship (reference category). Farmers who participated in a working land conservation program were 9 percent more likely to variable rate N than non-participants. Farmers with a high information index value, those who use more information sources and who trust those sources more highly, are more likely to Nat a variable rate than those with low index scores. The difference in predicted probability of variable rate N use between farmers who use independent crop consultants as an information source on a daily basis, and those who never use is 21 percent. While this is not the variable with the largest total effect in the model, it is the variable that predicts that highest likelihood of variable rate N use at its max value (67 percent) when all other variables are at their average. 99 Farmers with a strong economic attitude/orientation are predicted to be 13 percent more likely to use variable rate N than those with a weak economic attitude. Farmers who expected to pass on their farm to a child or other relative were 10 percent more likely to use variable rate N than those who did not expect to pass on their farm. Farmers with high concern over the impacts of extreme weather events on their farm were 6 percent more likely to use variable rate N than those with low concern. Table 3.12. Total effects for significant variables on variable rate N use in terms of probabilities with all other variables at average. Prob. at min (min value) Prob. at max (max value) Total effect Farm Characteristics—Physical Cropland area (mean = 1.182) Corn yield (mean = 179.712) Silty-loam (mode = 0) Farm Characteristics—Social LLC (mode = 0) 39.0% (0.050) † 60.6% (3.569) 35.2% (110.727) 56.8% (248.696) 46.0% (0) 31.6% (1) 46.0% (0) 57.9% (1) 55.2% (1) 21.7% 21.6% -14.5% 11.9% 9.2% 13.4% 20.6% 13.3% 9.7% 6.3% Working land program participation (mode = 0) 46.0% (0) Information Info index (mean = 51.996) 39.4% (36.564) 52.8% (67.429) Crop consultants info source (mode = 1) 46.0% (1) 66.6% (5) Attitudes and Values Economic attitude (mean = 4.063) Passing on farm (mode = 1) Environmental Awareness 35.9% (1) 36.3% (0) 49.2% (5) 46.0% (1) Extreme weather impacts (mode = 3) 41.3% (1) 47.6% (5) Note: Total effect is calculated as the probability at the variable's max value minus the probability at the variable's minimum value. For continuous variables the min and max values are two standard deviations below and above the mean respectively unless otherwise noted. † The sample minimum is used because one standard deviation below the mean is less than the sample minimum. Results of Variable Rate P or K Model The significant total effects for the variable rate P or K model are shown in Table 3.13. The effect of high knowledge on variable rate P or K is 37 percent higher than having low knowledge. This is the largest magnitude total effect in this model, as well as producing the 100 highest maximum probability at 72 percent. Farmers with farm sales greater than $1 million are predicted to be 23 percent more likely to use variable rate P or K than those with sales less than $100,000. Higher average corn yields were also related to more likely use of variable rate P or K. Part time farmers (those who worked off farm 200 days or more) were 3 percent less likely than full time farmers to use variable rate P or K. Participants in any kind of conservation program were 10 percent more likely to use variable rate P or K than non-program participant. Table 3.13. Total effects for significant variables on variable rate P or K use in terms of probabilities with all other variables at average. Individual Characteristics Knowledge (mode = 3) Value of sales (mode = 3) Farm Characteristics—Physical Prob. at min (min value) Prob. at max (max value) Total effect 34.8% (1) 38.2% (1) 71.9% (5) 61.4% (4) 37.1% 23.2% Corn yield (mean = 179.712) 37.5% (110.727) 69.5% (248.696) 32.1% Farm Characteristics—Social Days worked off farm (mode = 1) 54.0% (1) Conservation program participation (mode = 0) 54.0% (0) Information Chemical dealers info source (mode = 3) Crop consultants info source (mode = 1) Other farmers info source (mode = 3) 35.9% (1) 54.0% (1) 60.3% (1) Attitudes and Values 50.8% (5) 64.4% (1) 70.1% (5) 69.9% (5) 47.6% (5) -3.2% 10.5% 34.1% 15.9% -12.7% Crop insurance use (mean = 82.371) 45.2% (10.701) 56.0% (100.00) † 10.7% Note: Total effect is calculated as the probability at the variable's max value minus the probability at the variable's minimum value. For continuous variables the min and max values are two standard deviations below and above the mean respectively unless otherwise noted. † The sample maximum is used because one standard deviation above the mean is greater than the sample maximum. Farmers who use chemical dealers as an information source on a daily basis are predicted to be 34 percent more likely to use variable rate P or K than farmers who never use chemical dealers, almost as large an effect as knowledge. Farmers with daily versus non-use of independent crop consultants are predicted to be 16 percent more likely to use variable rate P or K, but farmers with daily use of other farmers as an information sources verses those who do not 101 consult other farmers were 13 percent less likely to use variable rate P or K. Farmers with all of their corn and soy acres enrolled in crop insurance were 11 percent more likely to use variable rate P or K than those who did not use crop insurance. Knowledge, corn yield, and chemical dealers as an information source have the three highest total effects in this model, and all three have similar probabilities of variable rate P or K at their max values. Independent crop consultants as an information source actually predicts a slightly higher probability of variable rate P or K use at its max value than does corn yield, in spite of having a total effect of about half the size. Results of Cover Crop Model Table 3.14 shows the significant total effects for the cover crop model. Farmers with a high level of knowledge are predicted to be 17 percent more likely to use cover crops than farmers with low knowledge. A high corn to soy ratio (i.e., farmer is growing mainly corn) is predicted to have a small negative effect on the probability of cover crop use. This variable is a measure of crop diversity, and cover crops are a type of crop diversity so this effect/relationship makes sense. Farmers having greater than 10 percent of farm revenue coming from livestock or from non-field crops or grain (e.g., other farm type), are also more likely to use cover crops by 9 percent and 7 percent respectively. Farmers who have largest corn fields with a greater than 10 percent grade are about 4 percent more likely to use cover crops than farmers with flat largest corn fields. This suggests that at least some cover crop use is related to how sloped the farm is, assuming the largest corn field is generally indicative of the topography of the area/farm. Farmers with high average corn yields are less likely to use cover crops than those with lower average yields. Farms with silty- loam soil texture in the largest corn field are predicted to be 3 percent less likely to use cover 102 crops than those with clay-loam as the dominant soil texture in the largest corn field. Having the farm structured as an LCC vs a sole proprietorship increased the predicted probability of using cover crops by 4 percent, while participation in a working lands conservation program increase cover crop use by an estimated 14 percent. Table 3.14. Total effects for significant variables on cover crop use in terms of probabilities with all other variables at average. Individual Characteristics Knowledge (mode = 3) Farm Characteristics—Physical Corn-soy ratio (mean = 1.295) Animal farm (mode = 0) Other farm type (mode = 0) Field slope (1) Prob. at min (min value) Prob. at max (max value) Total effect 3.0% (1) 19.8% (5) 16.8% 9.3% (0.012) † 6.9% (4.442) 8.5% (0) 8.5% (0) 8.5% (1) 17.1% (1) 15.2% (1) 12.1% (4) -2.4% 8.6% 6.6% 3.6% -13.4% -2.9% 3.8% 14.4% 10.2% 11.2% 16.9% 11.4% -9.5% 2.4% 7.4% -12.1% Corn yield (mean = 179.712) 17.1% (110.727) 3.8% (248.696) Silty-loam (mode = 0) Farm Characteristics—Social LLC (mode = 0) Working land program participation (mode = 0) Information Associations info source (mode = 2) Crop consultants info source (mode = 1) Local extension info source (mode = 2) Attitudes and Values Look after environment (mode = 4) Economic attitude (mean = 4.063) Environmental Awareness 8.5% (0) 5.6% (1) 8.5% (0) 12.3% (1) 8.5% (0) 23.0% (1) 6.7% (1) 8.5% (1) 4.9% (1) 1.7% (1) 16.4% (1) 16.9% (5) 19.8% (5) 21.8% (4) 13.1% (5) 6.8% (5) Concern ag. contributes to water problems 7.1% (1) 9.5% (5) (mean = 3.575) Erosion impacts (mode = 5) Pest resistances impacts (mode = 5) 1.1% (1) 20.6% (1) 8.5% (5) 8.5% (5) Note: Total effect is calculated as the probability at the variable's max value minus the probability at the variable's minimum value. For continuous variables the min and max values are two standard deviations below and above the mean respectively unless otherwise noted. † The sample minimum is used because one standard deviation below the mean is less than the sample minimum. 103 Daily versus non-use of associations, independent crop consultants, and local extension as information sources had total positive effects on cover crop use of 10 percent, 11 percent, and 17 percent respectively. Farmers who feel that looking after the environment is very important are predicted to be 11 percent more likely to use cover crops than farmers who feel it has low importance, while having a high economic attitude predicts a 10 percent lower probability of cover crop use than a low economic attitude. High concern that agriculture contributes to water problems and high concern over erosion impacts on the farm both predict higher likelihood of cover crop use than low levels of concern, by 2 percent and 7 percent respectively. Farmers who are highly concerned about the impacts of pest resistance to pesticides on their farm are predicted to be 12 percent less likely to use cover crops than those who are not concerned about pest resistance. This represents a key practice trait deterring farmers from using the practice, as cover crops are often killed with herbicides in the spring prior to planting the main crop especially if no-till is also being use. Additional comparative insight across predictors can be gained by graphing probabilities across variable value gradients, and variables on the same scale can be graphed together to visually demonstrate the dynamics at play between minimum and maximum probability, and total effect. An example of this is given in Figure 3.4 where we can see the important influence that local extension can have on cover crop use, even though its total effect is only slightly larger than that of knowledge. Independent crop consultants have a smaller total effect than knowledge, but they have the same predicted probability of cover crop use at their max values. These differences occur because the normal distribution function is S-shaped, and thus the slope (and curve) of the predicted probabilities depends on the where on that S-curve the variable is being measured (Muthén et al. 2016). 104 This graph also clearly shows which variables have the most potential to increase cover crop use, and which are “maxed out” so to speak. Because the majority of farmers already have a high level of concern over erosion impacts on their farm, there is little if any room for increasing concern to change the likelihood that they will use cover crops. In contrast, these results suggest that increasing the frequency of contact with local extension, increasing the use of independent crop consultants, and increasing knowledge have the potential to increase the likelihood of cover crop use. Even increasing contact with local extension from once a year to once a month has the potential to increase the probability of cover crop use by more than 5%. e s u p o r c r e v o c f o y t i l i b a b o r P Knowledge Crop consultants info source Local extension info source Look after environment Erosion impacts 25% 20% 15% 10% 5% 0% 1, Never 2, Once a year 3, Once a month 4, Once a week 5, Once a day Level of concern/importance, frequency of info source use Figure 3.4. Graphical depiction of the predicted probabilities of select variables on cover crop use across a range of values. Note: The points on the lines indicate the mode value of the respective variables. 105 DISCUSSION Independent Crop Consultants As previously mentioned, the use of independent crop consultants as an information source is the only variable that is significant in all five models of practice use; it also has a positive effect on the use of all five practices. Figure 3.5. shows the graphed predicted probabilities of the five practices across the frequency of use of independent crop consultants. Here we can see the effect of the more than 40 percent total effect of independent crop consultants on PSNT use, with the effect on the other practices ranging between 10 percent and 20 percent. This shows that while the use of independent crop consultants has the potential to increase the use of all five practices, it is expected to be most effective at increasing the use of PSNTs based on its large total effect. The question then remains, why do some farmers choose to use independent crop consultants, and how can their use be increased? Stuart et al. (2018) do not consider the use of independent crop consultants specifically in their analysis of information source use in relation to nitrogen fertilizer, but they find from qualitative analyses that lack of trust in consultants associated with seed or fertilizer suppliers appears to be a reason that some farmers do not use those consultants, or do not put much weight in what they say. It is possible that for some farmers, the added cost of hiring an independent crop consultant is a worthwhile alternative to these other sources. Cost might be a factor in deterring the use of independent crop consultants, or farmers may have sufficient trust in the specific retail affiliated consultants they use, as found by Stuart et al. (2018), to not feel the need to hire an independent consultant. 106 PSNT Variable rate N Cover crops Nutrient maps Variable rate P or K e s u e c i t c a r p f o y t i l i b a b o r P 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Never Once a year Once a month Once a week Once a day Crop consultant info source, frequency of use Figure 3.5. Graphical depiction of the probability of using BMPs at different frequencies of use of independent crop consultants as an information source, with all other variables at the average response level. Knowledge Self-reported knowledge on nutrient management and soil conservation practices was positive and significant for three of the five practices: nutrient maps, variable rate P or K and cover crops. In Figure 3.6, we can see the effect of knowledge level on practice use for all five practices, and how the slopes of the practices that are significantly influenced by knowledge are steeper than those of the other practices. We can also see the larger total effect that knowledge has on variable rate P or K (37 percent), compared to nutrient maps (25 percent) and cover crops (17 percent). These large total effects and relatively high predicted probabilities at the max, make knowledge one of the most important variables in the analysis. These results suggest that even 107 the non-specific measure of knowledge used in this analysis is an important contributor to the use of at least some practices and should be investigated further. The use of a non-practice specific measure of knowledge in this analysis may be the reason that the knowledge variable is not significant in all the models. If farmers associate some practices more strongly with “nutrient management and soil conservation practices” than others, then a lack of association wouldn’t be surprising. Using practice-specific measures of “how-to” or “procedural” knowledge is an important next step in this line of research and would be expected to have positive and significant effects on the use of the given practice. Vitale et al. (2011) find this to be the case with conservation tillage, where understanding of conservation tillage practices had a positive and significant effect on the use of both conservation tillage and reduced tillage, and a negative and significant effect on the use of conventional tillage. PSNT Nutrient maps* Variable rate N Variable rate P or K* Cover crops* e s u e c i t c a r p f o y t i l i b a b o r P 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1 2 3 4 5 Knowledge level Figure 3.6. Graphical depiction of the probability of using BMPs at the five reported levels of knowledge, with all other variables at the average response level. Note: Asterisks (*) indicates the models in which knowledge has a significant effect. 108 While using practice-specific measures of knowledge is a clear next step, so is better understanding what leads to knowledge creation/acquisition. Education, experience and use of and trust in specific information sources are obvious candidates as important predictors of practice knowledge. However, these may be practice specific and/or have a complex relationship due to the low levels of correlation (< 0.5) found between knowledge, education and information source use in the present data (see Table 3.20 in Appendix C). Future modeling with more specific data can investigate this further. Corn Yield Average corn yield is a significant positive predictor of variable rate N, and variable rate P or K, and a significant negative predictor of cover crop use (Figure 3.7). It is among the most important predictors for the use of variable rate nutrients in terms of total effect size and predicted probability at the max value, though it is of relatively minor importance for cover crop use. Corn yield is included in the model as a measure of soil quality, and as such is not a measure that has potential for encouraging practice use. There is also the potential for variable rate nutrient use to be influencing corn yield, either directly or as an indicator of general attentiveness of the operator. Future examination of this dynamic would benefit from additional data on the soil type and quality of the farm (not just the largest field) and the inclusion of data on fertilizer rate and the use of other management practices. 109 PSNT Nutrient maps Variable rate N* Variable rate P or K* Cover crops* e s u e c i t c a r p f o y t i l i b a b o r P 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 111 145 180 214 249 Corn yield (bushels per acre) Figure 3.7. Graphical depiction of the probability of using BMPs at the mean average corn yield, and at one and two standard deviations below and above the mean. Note: Asterisks (*) indicates the models in which corn yield has a significant effect. Similarities Between Practices As previously mentioned and as is clearly apparent in Table 3.9, there is remarkably little overlap in variable significance across the practice models, and especially between the practices that were expected to be the most similar. The only variables that are significant for both PSNT and nutrient map use are the use of independent crop consultants as is discussed above, and concern that agriculture contributes to water problems, which has a positive effect on both of about 6 percent. Variable rate N and variable rate P or K only have corn yield in common besides independent crop consultant use as previously discussed, though corn yield has a positive effect on both practices. 110 Variable rate N has more variables in common with cover crops than with variable rate P or K. Being organized as an LLC instead of a sole proprietorship and participation in a working land conservation program both increase the likelihood of both variable rate N and cover crops. The effect of working land conservation programs on the use of both practices is not surprising since these are both practices that would be supported by such programs, though it is interesting that it did not have a significant effect on PSNT use. It is possible that the reduction in personal liability of the LLC business structure encourages farmers to take more risk in the purchase of new equipment in the case of variable rate N and in the new use of cover crops, which can take time to return benefits. It is interesting that the working land programs, which are intended to help reduce the risk of practice adoption, do not appear to do so enough to remove the significant effect of being an LLC. The total effect of being an LLC is larger than that of working land program participation for variable rate N (12 percent versus 9 percent), but the reverse is true for cover crops, where the total effect of being an LLC is quite small (4 percent) while the effect of working lands programs is notably larger (14 percent). Having a dominantly silty-loam soil texture on the largest 2016 corn field compared to a clay-loam texture had a negative effect on the use of both variable rate N and cover crops. This might also be an indication of similarity between these practices, but it could also be spurious given that the whole farm may not share that dominant soil type. More knowledge of soil properties by the researcher would be necessary to hypothesize on the reasons for the negative effect of silty-loam on variable rate N and cover crops, and the negative effect of loam on PSNT use. The fact that any of the soil texture of the largest 2016 corn field variables is significant in any of the models suggests that soil texture is a potentially relevant variable for some practices and might be worth investigating with more precise data. This could be done at the field level 111 with detailed enough field-level data, or possibly at the farm level using remote sensing data to determine soil type and variability across a region that could be assigned to a farm. There are a few other variables that have a significant effect on more than one practice. Other farm type and use of local extension as an information source both have a negative effect on nutrient map use but a positive effect on cover crop use. The effect of other farm type (i.e., those having more than 10 percent of their revenue generated by neither field/grain crops nor livestock) on nutrient map and cover crop use is not immediately apparent, though it could be related to the (in)compatibility of these practices with their production practices. The effect of local extension on nutrient map and cover crop use would appear to be from local extension not promoting or supporting the use of nutrient maps while promoting or supporting the use of cover crops. Lastly, concern over erosion impacts on the farm is positively significant on the use of both PSNTs and cover crops. The relation between erosion and the use of cover crops is clear, but the connection between erosion and PSNT use requires further investigation. Finally, a few other variables are interesting in their lack of significance, since they are staple variables for this type of analysis. For example, farming experience, and cropland owned are not significant in any of the models, education level is only significant for nutrient maps, but has a negative effect on their use, and cropland area is only significant for variable rate N. Some of these variables were significant in smaller models but became insignificant when additional variables were added. Since most analysis of predictors of BMPs use many fewer variables than the present one, it is perhaps not so surprising that some common control variables are not significant. 112 CONCLUSION The goal of this analysis was to look for the common drivers of BMP use across the five practices considered, and to assess the relative importance of significant drivers. The results showed that these five practices have few drivers in common, even between practices that seemed to be similar. The use of independent crop consultants was the only variable of the 53 used in the model that had a significant effect on the use of all five practices. Practice knowledge and average corn yield each had significant effects on three of the five practices. These three variables were also consistently among the most important for predicting the use of each practice when considering both the total effect size and the probability of practice use at the maximum value of each variable. The first conclusion of these results is that there are no “families” or “types” or BMPs that share a common set of predictors. Without additional empirical testing we should not assume that any or all BMPs are meaningfully similar, and we should treat them separately in future analyses. This is goes against Reimer et al.’s (2014:58A) suggestion that “one approach might be to compare the significant drivers (or overall predictive power) of multivariate adoption models across categories of practices that share certain characteristics.” There are many dimensions in which practices could share characteristics, and the present analysis shows that these similarities may not be easily apparent—I find variable rate N and cover crops to be the most similar (and this is a relative measure as they only share three variables with the same sign in addition to independent crop consultant use), perhaps due to similarities in amount of cost and risk involved, rather than the pairs that seemed most similar based on the purpose of the practice, or pairs suggested by nutrient or practice timing. At the very least the formation of such categories should come with empirical support. 113 This conclusion is supported by the lack of consistent variable effects in the meta- analyses of BMP adoption studies (Baumgart-Getz et al. 2012; Prokopy et al. 2008; Wauters and Mathijs 2014) that consider all types of BMPs together. Prokopy et al. (2008) also include analyses broken down by general practice categories such as “soil management,” “nutrient management,” and “pest management,” which shows some potential trends in variables effects, but nearly all are not significant more than they are significant, suggesting that a fruitful avenue of future research could be meta-analyses of specific practices. The differences in significant predictors between the five practices could be due to differences between nutrients and timing of practice, as well as type of practice. For example, PSNTs and variable rate N are both nitrogen specific practices that would be used during the growing season, though they are different types of practices (decision aid vs application technology). Nutrient maps are based on soil tests that are done in the fall and would typically be used to inform P application placement, which is also done in the fall, though these practices are also of different types (decision aid vs application technology). The five practices considered in this analysis each have a unique combination of these three dimensions (practice type, nutrient, and season) that helps explain the lack of overlap in significant predictors. The second conclusion is that the larger effect sizes of knowledge and independent crop consultants on the probability of practice use, especially compared to the effect sizes of the significant attitudes and value variables and environmental awareness variables, suggests that the adoption-diffusion model may provide more predictive power in predicting farmers’ use of BMPs compared to the social-psych theories. This is not to say that attitudes, values and awareness are not important, but this analysis finds them to be substantively less important in a relative sense than knowledge, information sources and corn yield. This is at least partly due to 114 the already quite high levels of environmental attitudes, and awareness reported by farmers in the sample, which leaves little room for these variables to increase the probability of practice use beyond the current level, all else being equal. In contrast, the low use of independent crop consultants and modest average levels of knowledge reported provides greater potential for increasing practice use by increasing knowledge, and the use of independent crop consultants. This suggests that while analyses that focus on the role of attitudes and environmental awareness in BMP use are not incorrect, they may be missing important predictive variables from their models such as information sources, knowledge, and biophysical context. The adoption-diffusion model depends heavily on the transmission of information about innovations, and from this analysis we can see that it matters from whom farmers are receiving information. The large positive effects of both knowledge and the use of independent crop consultants suggests that there may be a relationship between the use of independent crop consultants and farmers’ knowledge of the practices, but also that some practices may be more dependent on knowledge than others. In the latter case it is possible that independent crop consultants are in some way directly facilitating the implementation of the practice, possibly why they are so important for the use of PSNTs. The relationships between information sources, practice knowledge, and practice use, warrant greater and more nuanced investigation to see how these relationships are similar or different across BMPs. While the present analysis highlights the important differences between the predictors of the use of different practices, and indicates the value of considering them individually, this in no way means that multiple practices should not be considered in combination. The use of multiple BMPs simultaneously has important environmental implications (Holmes et al. 2016; Lam et al. 2011; Liu et al. 2016; Sowa et al. 2016), and little social research has been done on how the use 115 of one BMP may lead to, facilitate or enhance the use of another (Weber and McCann (2015) are a notable example). For example, there is no reason for a farmer to use PSNTs if they do not intend to sidedress any of their nitrogen, but they very well might sidesdress nitrogen without using PSNTs. Arguably the use of both practices together would be the most effective at applying the optimal amount of nitrogen to the crop. It is these types of combined practice use that would be very interesting and useful to pursue. Future research could consider: 1) alternative measures of BMP use that better account for the spatial and temporal patterns of use, 2) practice specific measures of knowledge, 3) predictors of knowledge creation for different practices, 4) interaction effects of predictors on BMP use, 5) effects of the use of one BMP on another, and 6) the relative fit and explanatory power of the adoption-diffusion model compared to social-psych theories, using the same dataset and multiple BMPs—is the use of some practices better predicted by one approach vs the other? 116 APPENDICES 117 Cleaning APPENDIX A: Data Cleaning and Variable Calculation if f3=1924 then delete; *remarkably long time farming; if state='NA' then delete; *one respondent removed all identifying survey info; *a4 outliers; if a4_corn_low=4030 then delete; if a4_corn_avg=4355 then delete; if a4_corn_high=4482 then delete; Some questions were coded with missing values as zero. These variables were set to missing when all parts of the question were zero. **recoding "other" write in responses so not double counted when added; ***decided to put contract livestock into a7e--important part being livestock on the farm; if a7f_desc = '32 acres alfalfa' then a7a = 1 and a7f = 0; if a7f_desc = 'Alfalfa and Timothy Hay' then a7a = 1 and a7f = 0; if a7f_desc = 'Beef Cows 18 HD' then a7e = 1 and a7f = 0; if a7f_desc = 'Christmas Trees' then a7c = 1 and a7f = 0; if a7f_desc = 'Contract Pullets' then a7e = 1 and a7f = 0; if a7f_desc = 'Custom dairy heifer raising' then a7e = 1 and a7f = 0; if a7f_desc = 'Fish' then a7e = 1 and a7f = 0; if a7f_desc = 'Hay' then a7a = 1 and a7f = 0; if a7f_desc = 'Hay Alfalfa/ Grass' then a7a = 1 and a7f = 0; if a7f_desc = 'Haylage (73 Acres Alfafa)' then a7a = 1 and a7f = 0; if a7f_desc = 'I custom raise hiefens for another person' then a7 = 1 and a7f = 0; if a7f_desc = 'Layers' then a7e = 1 and a7f = 0; if a7f_desc = 'Organic Spelt' then a7a = 1 and a7f = 0; if a7f_desc = 'Processing Pumpkins' then a7b = 1 and a7f = 0; if a7f_desc = 'Seed Corn' then a7a = 1 and a7f = 0; if a7f_desc = 'Seed Production' then a7a = 1 and a7f = 0; if a7f_desc = 'Seed corn' then a7a = 1 and a7f = 0; if a7f_desc = 'Tobacco' then a7a = 1 and a7f = 0; if a7f_desc = 'alfalfa' then a7a = 1 and a7f = 0; if a7f_desc = 'alfalfa hay' then a7a = 1 and a7f = 0; if a7f_desc = 'alfalfa sales' then a7a = 1 and a7f = 0; if a7f_desc = 'contract canning crop' then a7b = 1 and a7f = 0; if a7f_desc = 'corn & soybean' then a7a = 1 and a7f = 0; if a7f_desc = 'corn & soybeans' then a7a = 1 and a7f = 0; if a7f_desc = 'corn and beans' then a7a = 1 and a7f = 0; if a7f_desc = 'corn+soybeans' then a7a = 1 and a7f = 0; if a7f_desc = 'corn,beans,wheat' then a7a = 1 and a7f = 0; if a7f_desc = 'cull cows and calves' then a7e = 1 and a7f = 0; if a7f_desc = 'eggs' then a7e = 1 and a7f = 0; if a7f_desc = 'greenhouse bedding' then a7c = 1 and a7f = 0; 118 if a7f_desc = 'hal-alf' then a7a = 1 and a7f = 0; *assuming should be 'hay-alf'; if a7f_desc = 'hay' then a7a = 1 and a7f = 0; if a7f_desc = 'hay + straw' then a7a = 1 and a7f = 0; if a7f_desc = 'hay&straw' then a7a = 1 and a7f = 0; if a7f_desc = 'popcorn' then a7a = 1 and a7f = 0; if a7f_desc = 'rent hay' then a7a = 1; if a7f_desc = 'seed corn' then a7a = 1 and a7f = 0; if a7f_desc = 'seed wheat' then a7a = 1 and a7f = 0; if a7f_desc = 'soybeans' then a7a = 1 and a7f = 0; if a7f_desc = 'soybeans=corn' then a7a = 1 and a7f = 0; if a7f_desc = 'straw bales sales' then a7a = 1 and a7f = 0; if a7f_desc = 'timothy hay clover hay' then a7a = 1 and a7f = 0; if a7f_desc = 'tobacco' then a7a = 1 and a7f = 0; if a7f_desc = 'turkeys' then a7e = 1 and a7f = 0; *a10--scale 1-6; **recodeing 6 responses that fit in listed categories; ***only a few clearly seem to be in a listed category; if a10_desc = 'Owner' then a10=1; if a10_desc = 'just me' then a10=1; Outcome variables PSNT if c1a in (4 5) then c1a_use2=1; else c1a_use2=0; if c1a=. then c1a_use2=.; Nutrient maps if c1d in (4 5) then c1d_use2=1; else c1d_use2=0; if c1d=. then c1d_use2=.; Variable rate N if c1b in (4 5) then c1b_use2=1; else c1b_use2=0; if c1b=. then c1b_use2=.; Variable rate P or K if c1c in (4 5) then c1c_use2=1; else c1c_use2=0; if c1c=. then c1c_use2=.; Cover crops if c1k in (4 5) then c1k_use2=1; else c1k_use2=0; if c1k=. then c1k_use2=.; Individual Characteristics yrsfrm=2017-f3; Years farming Education level College ag courses f5_rc=f5; f4; 119 if f5=. then f5_rc=0; if f4=. then f5_rc=.; if f5=0 and f4=. then f5_rc=0; *one respondent didn't answer f4 but did answer f5 so putting that response back in; f9; Knowledge c2; Value of sales Household income f10; Ag. income f11; Farm Characteristics—Physical crop2size=(a3_corn+a3_soy+a3_wheat+a3_other)/1000; Cropland area Corn-soy ratio Animal farm corn_soy=a3_corn/a3_soy; if a7d=1 then animal=1; else animal=0; if a7e=1 then animal=1; if a7d=. then animal=.; if a7e=. then animal=.; Grain farm a7a; Other farm type if a7b=1 then frm_oth=1; else frm_oth=0; if a7c=1 then frm_oth=1; if a7f=1 then frm_oth=1; if a7b=. then frm_oth=.; if a7c=. then frm_oth=.; if a7f=. then frm_oth=.; Field slope b6; Soy yield Corn yield Loam Clay a4_soy_avg; *renamed soy_avg; a4_corn_avg; *renamed corn_avg; if b7=3 then loam=1; else loam=0; if b7=. then loam=.; if b7=4 then clay=1; else clay=0; if b7=. then clay=.; Clay-loam if b7=5 then clay_loam=1; else clay_loam=0; 120 if b7=. then clay_loam=.; Silty-loam if b7=6 then silty_loam=1; else silty_loam=0; if b7=. then silty_loam=.; Sandy-loam if b7=7 then sandy_loam=1; else sandy_loam=0; if b7=. then sandy_loam=.; Other soil type if b7=1 then sand=1; else sand=0; if b7=. then sand=.; if b7=2 then silt=1; else silt=0; if b7=. then silt=.; if b7=8 then soil_other=1; else soil_other=0; if b7=. then soil_other=.; oth_soil=soil_other+sand+silt; Farm Characteristics—Social Cropland owned crpown_pc=(a2_crop_own/(a2_crop_own+a2_crop_in))*100; f8; Days worked off farm Number of employees Sole proprietorship employ=a8a+a8b; if a10=1 then sole_prop=1; else sole_prop=0; if a10=. then sole_prop=.; Partnership LLC if a10=2 then partnshp=1; else partnshp=0; if a10=. then partnshp=.; if a10=3 then llc=1; else llc=0; if a10=. then llc=.; S-corporation if a10=4 then s_corp=1; else s_corp=0; if a10=. then s_corp=.; C-corporation if a10=5 then c_corp=1; else c_corp=0; if a10=. then c_corp=.; Other business structure if a10=6 then other_bs=1; else other_bs=0; if a10=. then other_bs=.; a5; Conservation program participation Working land program participation if a5a>0 then wrkldprg=1; else wrkldprg=0; if a5a=. then wrkldprg=.; 121 Information Info index info_tot=d1cedu+d1tedu+d1ccamp+d1tcam+d1cchem+d1tchem+d1cseed+d1tseed+d1cin d+d1tind+d1cfam+d1tfam+d1cmag+d1tmag+d1cweb+d1tweb+d1cass+d1tass; d1cind; d1cass; d1cseed; d1cchem; Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values d1ccamp; d1cfam; d1cedu; ECON_AT BY d2inco d2weal d2prof; d2inco; d2envi; d2hunt; d2weal; Look after environment Hunting and fishing environment Economic attitude Income Wealth Profit Passing on good land Farming tradition Passing on farm Debt aversion d2debt; Crop insurance use d2land; d2prof; d2trad; f7; crop_ins=(a6/(a3_corn+a3_soy))*100; if crop_ins>100 then crop_ins=100; 122 *274 cases had crop insurance use over 100% suggesting they didn't read the question correctly and put all crop insurance acres, rather than just those for corn and soy, so I capped the values at 100 if they were larger than that; Environmental Awareness Concern ag. contributes to water problems WATERCON BY e1ccont e1calga e1chypo e1csoil; e1calga WITH e1chypo; e1calga; e1ccont; Groundwater contribution Algal bloom contribution Hypoxia contribution Erosion contribution e1csoil; Erosion impacts e1chypo; e1fsoil; e1fresi; Pest resistances impacts Extreme weather impacts Warming impacts e1fwarm; e1fextr; 123 Estimator Choice APPENDIX B: Modeling Decisions The choice of estimator was driven by the binary nature of the outcome variables of interest. In addition, the measures of the two latent exogenous variables are categorical with five response options. I had two options for estimators: maximum likelihood and categorical least squares. Maximum likelihood (ML)—often with robust standard errors (MLR)—is a full- information estimator and probably the most common estimator in SEM. ML(R) can be used with binary or categorical outcomes as the robustness correction adjusts for the non-normality of the categorical variable to make it compatible with the ML assumptions that apply to continuous variables (Li 2016; Rhemtulla, Brosseau-Liard, and Savalei 2012). This estimator uses the observed variable with a probit or logit link function to estimate the relationships between the specified variables (Rhemtulla et al. 2012). The alternative estimator is some form of limited information, categorical least squares (LS) estimator (usually weighted) that assumes a continuous latent variable underlying the categorical variable—threshold values are calculated that relate the values of the latent variable to the categories in the observed variable, and the latent variable is used in the rest of the model estimates (Li 2016; Rhemtulla et al. 2012). These estimators can also have robustness corrections, but these corrections are to adjust for using limited instead of full information rather than for data being non-normal as it is for ML(R) (Li 2016; Rhemtulla et al. 2012). In this analysis I use a robust, diagonally weighted least squares, means and variance adjusted estimator (specifically the WLSMV estimator in Mplus8). There are several reasons for using a limited information, categorical LS estimator, even though it is less common. 124 1. More recent weighted LS estimators perform better than older versions (Beauducel and Herzberg 2006; Li 2016). 2. Simulation studies have found categorical LS estimators to perform as well or better than ML estimators under a range of model conditions (Bandalos 2014; Beauducel and Herzberg 2006; Li 2016; Rhemtulla et al. 2012). 3. Weighted LS estimators, by being limited information, are less computationally intensive than ML(R), especially with large models (Muthén and Muthén 2017). 4. Though LS estimators are less efficient by virtue of being limited information, this has not been found to a problem in practice (Rhemtulla et al. 2012). 5. By using the underlying latent variables weighted LS estimators have additional benefits when computing more complex models, like models with categorical mediators, such as allowing for the estimation of indirect and total effects, and standardized effects, which can’t be done with ML(R), allowing for more comparisons between the results of different models (Muthén and Muthén 2017). Simulation studies of confirmatory factor analysis (CFA) models have found diagonally weighted LS to perform better than ML (Beauducel and Herzberg 2006), and MLR (Bandalos 2014; Li 2016), especially for binary outcomes (Beauducel and Herzberg 2006), as well as with model misspecification, as long as both sample size was sufficiently large and the data was not excessively asymmetric (Bandalos 2014; Li 2016). Bandalos (2014) and Li (2016) also consider the effects of different estimators on structural coefficients between latent variables with categorical measures, and they find that diagonally weighted LS performs at least as well as ML or MLR in most conditions. With a larger number of response categories, the differences between estimators diminished (Li 2016; Rhemtulla et al. 2012). While none of these studies examine the effects of estimators on estimates between two observed variables (I was not able to find any studies that did this), these findings are still relevant for my observed outcome variable, at least for the relationship between it and my latent 125 exogenous variables, since these are very similar to CFA dynamics. They are also relevant for the estimation of my exogenous latent variables themselves. Given my large sample size23 and not excessively skewed endogenous variables24 (see Table 3.15) there may not be a critical need for the WLSMV estimator, but it is conceptually the more appropriate choice and the simulation studies do not suggest that I face conditions where it would perform poorly. Table 3.15. Skewness and kurtosis for endogenous variables. PSNT Nutrient maps Variable rate N Variable rate P or K Cover crops Income Wealth Profit Groundwater contribution Algal bloom contribution Hypoxia contribution Erosion contribution Number of response categories N Skewness Kurtosis 2 2 2 2 2 5 5 5 5 5 5 5 814 814 814 814 814 813 812 814 802 803 799 805 1.653 -1.788 0.995 -0.596 0.607 -0.226 -0.886 -0.844 -0.454 -0.361 -0.246 -0.886 0.734 1.201 -1.013 -1.649 -1.636 -0.191 0.911 0.357 -0.321 -0.428 -0.765 0.726 Calculating Probabilities Using the WLSMV Estimator The WLSMV estimator uses a probit link function and thus estimates probit coefficients. However, if the outcome is considered to be the underlying latent variable then the coefficient can be treated as a linear coefficient—this is relevant for the discussion of standardized results below. 23 Sample sizes in the simulation studies ranged from 100 to 1,000, putting mine in the ‘large’ category. 24 In considering asymmetry in my endogenous observed variables, my variables fall in the low nonnormality category used by Bandalos (2014) by having skew smaller than ±2.4 and kurtosis less than ±3.8. 126 Probabilities of y =1 can be calculated from the intercept25, probit coefficients, and specified values for each variable using the following equation where F is the standard normal distribution function (Muthén, Muthén, and Asparouhov 2016:222-223). Using the above equation, a z score is calculated, then a left sided z table is used to determine the P(ui = 1 | xi) = F(β0 + β1 x1 + β2 x2 + … + βi xi) probability of F(z)—this is equal to the area under the curve to the left of the z value. Probabilities can be calculated from probit coefficients the same way whether the observed variable or the underlying latent variable is used if the residual variance is fixed at 1, which is typical for probit regression (Muthén, Muthén, and Asparouhov 2016:224-226). In cases where there is more than one categorical dependent variable, such as a mediation model, then the residual variance must be included in the probability calculation: P(ui = 1 | xi) = F[(β0 + β1 x1 + β2 x2 + … + βi xi)/√residual variance] If there is more than one mediator the formula is more complex (see Nguyen et al. 2016). Assessing the Nestedness of the Data Besides the model decisions based on the binary nature of the outcome variable, the potential for nestedness in the data drove many modeling decisions. When data is nested or clustered the OLS assumption of independence among observations is violated, and analyses that do not treat the data as clustered will tend to underestimate the standard errors, resulting in higher rates of type 1 errors (Bell and Jones 2015). Since the survey data used was collected in four states there is the potential for there to be significant clustering of respondents by state— that is, respondents from a given state may be more like each other than they are like farmers 25 The intercept for a binary outcome variable is the negative threshold value that Mplus reports. 127 from the other states. Examining the data for evidence of clustering is the key next step in determining how the model should be specified. The first step of assessing clustering is to run a one-way ANOVA model that divides the variance in the variable between levels, and to calculate the intraclass correlation coefficient (ICC), which is the between variance divided by the total variance, effectively providing the portion/percent of the variance that exists at the between level/level 2 (Snijders and Bosker 2012). Table 3.15 presents the ICCs and design effects for the full sample and the final sample using the WLSMV estimator with the variable specified as categorical, and for the full sample using MLR with the variable specified as continuous. When the variable is defined as categorical, regardless of what estimator is used, Mplus does not allow variances on the within level, thus making the between variances reported from these models suspect (none are significant). Therefore, I also present the results using MLR with the variable specified as continuous, to include the within level variance. The null model could not be computed for PSNT and variable rate N when using MLR for the final sample, due to the estimated between covariance matric not being positive definite. Design effect26 is included because there is no established threshold rule for ICC size, and even small ICCs can indicate important clustering with large cluster size (Snijders and Bosker 2012). The design effect, as a function of average cluster size and ICC, provides a way to evaluate the significance of the ICC based on cluster size (Lai and Kwok 2015). Lai and Kwok (2015) find that the ‘rule of thumb of being able to ignore clustering if the design effect is less than two is reasonable under certain conditions27. All but two of the design effects in Table 3.15 26 Design effect = 1 + (average cluster size – 1) x ICC (Lai and Kwok 2015) 27 When “(a) the cluster size … is at least 10, (b) the relations between level 1 predictors and the outcome are constant (i.e., no random coefficients are present), and (c) the predictors are group-mean centered” (Lai and Kwok 2015:434) 128 are substantially larger than 2, suggesting that there is important clustering in the data, at least for nutrient maps, variable rate P or K, and cover crops. Table 3.16. ICC comparison by sample and estimator. Nutrient Variable rate Variable rate maps N P or K Cover crops Final sample sample size average cluster size Full sample sample size average cluster size WLSMV final sample ICC Design Effect WLSMV full sample ICC Design Effect PSNT 814 203.5 2007 501.8 0.000 1.000 0.009 5.507 t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 814 203.5 814 203.5 814 814 203.5 203.5 2113 528.3 2061 515.3 2117 529.3 2098 524.5 0.039 8.898 0.008 5.218 0.000 1.000 0.007 4.600 0.025 6.0625 0.011 6.811 0.054 11.935 0.074 39.739 The conclusions from Table 3.16 are that there appears to be significant clustering for three of the five practices. For PSNT and variable rate N, the clustering appears to be less significant: the ICCs are zero for the final sample resulting in design effects of one, and these are the only two practices that do not have even marginally significant between level variances in the MLR model. This conclusion is also supported by considering the graphs of practice use by state (see Figure 3.8), which shows significant state-level differences as tested by logistic regression28 (see Table 3.17) for nutrient maps, variable rate P or K, and cover crops, but not PSNT and variable rate N. 28 Logistic regression used since ANOVA is inappropriate for binary variables. 129 Table 3.17. Wald Chi-square for joint test of logistic regression for state differences in practice use using final sample (n=814). State Degrees of Freedom 3 PSNT Nutrient maps Variable rate Variable rate N P or K Cover crops 2.474 17.274*** 3.733 16.235** 33.692*** t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% PSNT Nutrient maps Variable rate N Variable rate P Cover crops or K IL IN MI OH Figure 3.8. Practice use by state for final sample. Snijders and Bosker (2012) also suggest checking for group-by-covariate interaction effects, as clustering can manifest in relation to covariates. The presence of significant interaction terms for all five practices across a sample of covariates (using logistic regression given the inappropriateness of ANOVA for binary variables and the challenges of ANCOVA with binary outcomes) suggests that there are important group differences in the presence of covariates for all of the practices (see Table 3.18). Given the presence of group-by-covariate interactions and design effects >2 for the full sample, clustering is considered to be an important 130 consideration for all five practices considered. This assumption also facilitates comparisons across models, as it justifies using the same model specification. However, given the small number of clusters (there are only four) specifying a multilevel model is problematic and can produce downwardly biased standard errors for both coefficients and variances, though the coefficients and variances themselves have been found to be unbiased (Lai and Kwok 2015; Maas and Hox 2005). Use of the COMPLEX analysis type in Mplus8 should avoid this problem by adjusting the standard errors and chi-square test of model fit for clustering but without modeling the clustering as a multilevel model would (Muthén and Muthén 2017). 131 Table 3.18. Wald Chi-square for joint test of logistic regression for group-by-covariate interactions for select variables using final sample (n=814). Degrees of Freedom PSNT Nutrient maps Variable rate N Variable rate P or K Cover crops State Average corn yield Average corn yield x State State Average soy yield Average soy yield x State State Corn-soy ratio Corn-soy ratio x State State Percent cropland owned Percent cropland owned x State State Importance of looking after environment Importance of looking after environment x State State Knowledge of practices Knowledge of practices x State t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 3 1 3 3 1 3 3 1 3 3 1 3 3 1 3 3 1 3 1.255 12.173** 0.632 1.013 8.182** 31.273*** 15.908*** 42.957*** 1.624 12.358** 1.640 25.969*** 11.021*** 36.116*** 1.172 4.920* 1.598 2.248 0.006 2.744 3.628 4.672* 2.634 3.637 1.057 3.660 6.059 4.430* 5.357 15.836** 15.462*** 11.490** 2.075 5.912 1.625 3.602 3.581 3.702t 2.718 17.795*** 20.674*** 19.186*** 4.687 3.592t 4.006 15.116** 2.979t 14.255** 8.250* 3.869 7.636t 14.546** 0.050 12.863** 17.190*** 0.471 17.629*** 26.578*** 1.184 2.639 16.841*** 3.664t 4.143 1.426 23.756*** 2.031 6.976t 1.376 1.141 1.680 6.135 0.641 8.803* 9.314* 7.429** 1.372 1.159 5.960* 2.527 2.405 26.530*** 33.181*** 24.247*** 37.176*** 28.406*** 9.396* 1.901 8.520* 0.677 2.551 132 APPENDIX C: Additional Results A Note on R2 with Binary Outcomes The R-square values produced for the models are a bit different than the R-square for a regression model with a continuous variables, because, for binary outcomes, the R-square is based on the liner underlying latent variable and does not “involve a freely estimated residual variance” (Muthén et al. 2016:227). The proportion of variance explained is determined by the slope parameter alone which also describes the strength of relationship between x and u. Viewed from a probability curve perspective it is the steepness of a curve…that summarizes the situation: the flatter the curve, the lower the value of the slope, the lower the R2, and the higher the fixed residual variance relative to the total u* variance. (Muthén et al. 2016:227) For a binary outcome R2 = γ2V(x) / (γ2V(x) + c) The variance is V(u*) = γ2V(x) + c where c =1 for probit regression Additional Descriptives Table 3.19. Tolerance and variance inflation factor (VIF) for exogenous analysis variables (n=814). Individual Characteristics Years farming Education level College ag courses Knowledge Value of sales Household income Ag. income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Grain farm Other farm type Field slope Tolerance 0.785 0.580 0.594 0.770 0.422 0.792 0.607 0.399 0.873 0.827 0.635 0.641 0.893 VIF 1.274 1.724 1.685 1.299 2.372 1.262 1.647 2.509 1.146 1.209 1.576 1.561 1.120 133 Table 3.19. (cont’d). Soy yield Corn yield Loam Clay Silty-loam Sandy-loam Other soil type Farm Characteristics—Social Cropland owned Days worked off farm Number of employees Partnership LLC S-corporation C-corporation Other business structure Conservation program participation Working land program participation Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values Look after environment Hunting and fishing environment Passing on good land Farming tradition Passing on farm Debt aversion Crop insurance use Environmental Awareness Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts 0.514 0.491 0.856 0.851 0.787 0.824 0.858 0.842 0.556 0.711 0.823 0.788 0.882 0.823 0.939 0.681 0.671 0.362 0.689 0.522 0.526 0.757 0.668 0.556 0.573 0.695 0.785 0.786 0.673 0.713 0.856 0.858 0.577 0.600 0.488 0.488 1.944 2.035 1.169 1.176 1.270 1.213 1.166 1.188 1.798 1.406 1.215 1.270 1.134 1.215 1.065 1.468 1.491 2.765 1.451 1.915 1.902 1.320 1.497 1.798 1.746 1.439 1.274 1.272 1.487 1.402 1.169 1.166 1.734 1.667 2.051 2.050 134 Table 3.20. Select Pearson correlation coefficients for exogenous variables where correlation coefficient is < -0.5 or > 0.5 (n = 814). Ed. level Value of sales Ag. income Grain farm Soy yield College ag courses 0.580*** Cropland area Other farm type Corn yield Days worked off farm Working land program participation Seed dealers info source Local extension info source Pest resistances impacts Warming impacts * p < 0.05, ** p < 0.01, *** p < 0.001 0.645*** -0.518*** -0.553*** Conservation Chemical program dealers participation info source Campus extension info source Erosion impacts Extreme weather impacts 0.507*** 0.621*** 0.584*** 0.568*** 0.681*** 0.651*** 135 Measurement Model Results Table 3.21 Average and range of measurement model results across the five practice analyses. Probit Standardized R-square Economic attitude Income Wealth Profit Average coefficient (min - max) Average standard error (min - max) 1 0 (1.000 - 1.000) (0.000 - 0.000) 0.6988 0.0436 (0.681 - 0.724) (0.042 - 0.045) 0.627 0.079 (0.613 - 0.648) (0.077 - 0.080) Concern ag. contributes to water problems Groundwater contribution Algal bloom contribution 1.000 0.000 (1.000 - 1.000) (0.000 - 0.000) 0.511 0.081 (0.508 - 0.514) (0.077 - 0.082) 0.394 0.048 Hypoxia contribution Erosion contribution (0.392 - 0.395) (0.046 - 0.048) 0.401 0.056 (0.399 - 0.403) (0.053 - 0.057) Average coefficient (min - max) Average standard error (min - max) Average coefficient (min - max) Average standard error (min - max) 0.8178 0.0084 0.669 0.014 (0.812 - 0.822) (0.008 - 0.009) 0.7044 0.015 (0.701 - 0.709) (0.015 - 0.015) 0.665 0.037 (0.663 - 0.669) (0.036 - 0.037) 0.918 0.017 (0.917 - 0.919) (0.016 - 0.017) 0.763 0.015 (0.763 - 0.764) (0.014 - 0.015) 0.673 0.007 (0.673 - 0.674) (0.007 - 0.007) 0.680 0.016 (0.679 - 0.681) (0.016 - 0.016) (0.659 - 0.676) (0.012 - 0.014) 0.496 0.021 (0.491 - 0.503) (0.021 - 0.021) 0.442 0.049 (0.439 - 0.448) (0.048 - 0.049) 0.843 0.031 (0.841 - 0.844) (0.030 - 0.032) 0.583 0.023 (0.582 - 0.583) (0.022 - 0.023) 0.454 0.009 (0.453 - 0.455) (0.009 - 0.010) 0.462 0.022 (0.461 - 0.463) (0.022 - 0.022) Note: All factors are significant at < 0.001 136 Full Probit Regression Results Table 3.22. Probit regression coefficients and standard errors for PSNT and nutrient maps (n=814). Individual Characteristics Years farming Education level College ag courses Knowledge Value of sales Household income Ag. income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Grain farm Other farm type Field slope Soy yield Corn yield Loam Clay Silty-loam Sandy-loam Other soil type Farm Characteristics—Social Cropland owned Days worked off farm Number of employees Partnership LLC S-corporation C-corporation Other business structure Conservation program participation Working land program participation PSNT Estimate (S.E.) 0.004 (0.005) -0.086 (0.112) 0.125 (0.070)t 0.178 (0.222) -0.065 (0.128) 0.083 (0.054) 0.039 (0.049) 0.041 (0.083) 0.050 (0.047) -0.114 (0.306) -0.197 (0.437) -0.458 (0.285) -0.165 (0.127) 0.000 (0.008) 0.002 (0.002) -0.558 (0.149)*** 0.079 (0.100) -0.319 (0.216) 0.101 (0.272) -0.120 (0.256) -0.002 (0.002) 0.013 (0.025) 0.005 (0.026) 0.31 (0.121)* -0.025 (0.077) -0.126 (0.465) 0.001 (0.318) 0.987 (0.584)t 0.053 (0.235) 0.290 (0.201) 137 Nutrient maps Estimate (S.E.) -0.001 (0.007) -0.084 (0.016)*** 0.271 (0.141)t 0.204 (0.097)* 0.149 (0.131) 0.060 (0.031)* 0.008 (0.034) -0.097 (0.192) 0.021 (0.052) -0.038 (0.145) -0.093 (0.559) -0.671 (0.298)* 0.050 (0.190) 0.010 (0.006) 0.006 (0.008) 0.265 (0.280) 0.153 (0.124) -0.026 (0.126) 0.135 (0.232) 0.529 (0.418) -0.003 (0.005) 0.015 (0.068) 0.028 (0.109) 0.206 (0.374) 0.08 (0.0970) -0.109 (0.378) -0.035 (0.480) 0.127 (1.008) 0.277 (0.192) 0.056 (0.194) Table 3.22. (cont’d). Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values Look after environment Hunting and fishing environment Economic attitude Passing on good land Farming tradition Passing on farm Debt aversion Crop insurance use Environmental Awareness Concern ag. contributes to water problems Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts Intercept R-Square Chi-Square RMSEA CFI TLI t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 -0.005 (0.004) 0.161 (0.176) 0.143 (0.155) 0.035 (0.101) 0.295 (0.037)*** 0.018 (0.041) 0.017 (0.086) 0.067 (0.074) 0.217 (0.271) 0.007 (0.053) 0.052 (0.070) -0.091 (0.051)t 0.290 (0.185) 0.046 (0.145) -0.107 (0.100) 0.000 (0.002) 0.060 (0.006)*** 0.094 (0.035)** -0.084 (0.055) -0.040 (0.026) -0.035 (0.075) -4.325 (1.351)** 0.349 (0.073)*** 679.695*** 0.033 0.839 0.812 0.021 (0.025) 0.064 (0.066) 0.171 (0.232) -0.155 (0.072)* 0.174 (0.056)** 0.033 (0.093) 0.109 (0.078) -0.129 (0.054)* -0.014 (0.121) -0.001 (0.060) -0.003 (0.024) -0.041 (0.040) -0.082 (0.034)* 0.015 (0.153) -0.056 (0.078) 0.003 (0.001)t 0.051 (0.015)*** 0.013 (0.077) 0.091 (0.098) 0.011 (0.054) -0.098 (0.040)* -2.937 (0.695)*** 0.371 (0.121)** 678.584*** 0.033 0.839 0.812 138 Table 3.23. Probit regression coefficients and standard errors for variable rate N and variable rate P or K (n=814). Variable rate N Estimate (S.E.) Variable rate P or K Estimate (S.E.) 0.001 (0.003) 0.034 (0.070) -0.187 (0.171) 0.123 (0.105) 0.045 (0.068) 0.019 (0.071) -0.093 (0.057) 0.156 (0.060)* -0.002 (0.051) -0.231 (0.169) -0.040 (0.298) -0.415 (0.314) -0.063 (0.097) 0.002 (0.009) 0.004 (0.000)*** 0.030 (0.233) -0.173 (0.124) -0.375 (0.148)* 0.091 (0.132) -0.213 (0.157) 0.000 (0.002) 0.000 (0.041) 0.001 (0.025) -0.014 (0.367) 0.306 (0.074)*** 0.064 (0.100) -0.050 (0.169) -0.403 (0.869) -0.029 (0.131) 0.000 (0.007) -0.163 (0.094)t 0.130 (0.165) 0.244 (0.062)*** 0.198 (0.077)* 0.004 (0.064) -0.062 (0.041) 0.026 (0.102) -0.009 (0.107) -0.132 (0.110) -0.119 (0.144) -0.456 (0.260)t 0.023 (0.080) 0.010 (0.010) 0.006 (0.003)* 0.217 (0.150) -0.261 (0.241) -0.100 (0.099) 0.328 (0.174)t 0.242 (0.329) 0.000 (0.001) -0.019 (0.006)** -0.007 (0.010) -0.127 (0.160) 0.117 (0.118) -0.087 (0.209) -0.030 (0.187) -0.077 (0.456) 0.279 (0.073)*** -0.108 (0.188) Individual Characteristics Years farming Education level College ag courses Knowledge Value of sales Household income Ag. income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Grain farm Other farm type Field slope Soy yield Corn yield Loam Clay Silty-loam Sandy-loam Other soil type Farm Characteristics—Social Cropland owned Days worked off farm Number of employees Partnership LLC S-corporation C-corporation Other business structure Conservation program participation Working land program participation 0.236 (0.063)*** 139 Table 3.23. (cont’d). Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values Look after environment Hunting and fishing environment Economic attitude Passing on good land Farming tradition Passing on farm Debt aversion Crop insurance use Environmental Awareness Concern ag. contributes to water problems Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts Intercept R-Square Chi-Square RMSEA CFI TLI t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 0.011 (0.002)*** 0.051 (0.094) -0.013 (0.124) 0.072 (0.122) 0.134 (0.059)* 0.018 (0.107) -0.108 (0.079) 0.157 (0.124) 0.004 (0.161) -0.037 (0.031) 0.085 (0.035)* -0.048 (0.030) -0.001 (0.050) 0.251 (0.082)** 0.009 (0.054) -0.001 (0.001) -0.001 (0.021) 0.060 (0.056) -0.027 (0.108) 0.039 (0.012)** -0.069 (0.096) -2.878 (0.348)*** 0.237 (0.049)*** 679.410*** 0.033 0.840 0.813 0.019 (0.021) -0.121 (0.112) 0.226 (0.044)*** -0.087 (0.081) 0.106 (0.039)** -0.080 (0.037)* 0.044 (0.082) -0.059 (0.059) 0.035 (0.106) -0.002 (0.018) 0.008 (0.007) -0.132 (0.106) -0.010 (0.081) 0.070 (0.090) -0.144 (0.076)t 0.003 (0.001)*** 0.022 (0.013) 0.029 (0.071) 0.057 (0.084) 0.002 (0.032) 0.001 (0.019) -2.758 (0.967)** 0.310 (0.028)*** 677.040*** 0.033 0.840 0.813 140 Table 3.24. Probit regression coefficients and standard errors for cover crops (n=814). Individual Characteristics Years farming Education level College ag courses Knowledge Value of sales Household income Ag. income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Grain farm Other farm type Field slope Soy yield Corn yield Loam Clay Silty-loam Sandy-loam Other soil type Farm Characteristics—Social Cropland owned Days worked off farm Number of employees Partnership LLC S-corporation C-corporation Other business structure Conservation program participation Working land program participation 141 Cover crops Estimate (S.E.) 0.002 (0.006) 0.042 (0.071) 0.111 (0.064)t 0.257 (0.088)** 0.031 (0.038) 0.038 (0.048) 0.041 (0.074) -0.086 (0.044)t -0.035 (0.016)* 0.420 (0.164)* 0.130 (0.286) 0.338 (0.144)* 0.065 (0.022)** -0.003 (0.010) -0.006 (0.002)* 0.091 (0.140) 0.112 (0.132) -0.227 (0.048)*** 0.040 (0.105) 0.343 (0.245) -0.001 (0.002) 0.003 (0.071) -0.005 (0.018) 0.144 (0.128) 0.208 (0.059)*** -0.062 (0.225) 0.166 (0.231) 0.287 (0.843) 0.122 (0.101) 0.631 (0.127)*** -0.009 (0.010) 0.136 (0.033)*** -0.001 (0.067) -0.072 (0.071) 0.129 (0.049)** 0.074 (0.050) -0.060 (0.079) 0.292 (0.039)*** 0.248 (0.097)* 0.025 (0.078) -0.126 (0.046)** 0.082 (0.069) -0.040 (0.074) 0.125 (0.108) -0.034 (0.042) -0.002 (0.001) 0.040 (0.024)t 0.227 (0.017)*** -0.138 (0.051)** 0.010 (0.024) -0.105 (0.064) -2.726 (1.106)* 0.338 (0.020)*** 681.087*** 0.033 0.841 0.814 Table 3.24. (cont’d). Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values Look after environment Hunting and fishing environment Economic attitude Passing on good land Farming tradition Passing on farm Debt aversion Crop insurance use Environmental Awareness Concern ag. contributes to water problems Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts Intercept R-Square Chi-Square RMSEA CFI TLI t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 142 Standardized Results Standardized results are included to satisfy expectations for the presentation of regression results. However, standardization is of little usefulness when dealing with probit results, and especially with probabilities, as the probability effect of a given variable is highly dependent on the range of possible values the variable can take and the average variable value as is discussed in the chapter. This means that the standardized coefficients are no more useful in comparing the relative effects of different variables than the unstandardized coefficients; one still has to consider the value range for the variable, the size of the coefficient/slope, and the variable average. The reason for this is that the standardization is done on the latent variable underlying the observed binary variable, not on the binary variable itself; the standardized value would still need to be related to the binary outcome via the threshold values. As Muthén, Muthén, and Asparouhov (2016) explain it: Although the need for comparing effects of covariates having different metrics may arise, it should be emphasized that these are effects in the metric of the standardized u*. In contrast, the effects on the probability of u = 1 vary depending on where along the x-axis it is evaluated…. With two or more covariates, standardization refers to partial regression coefficients where the effect on the u probability depends on the location of the combination of covariates. This complicates the comparison of standardized coefficients. A possible solution is instead to study the change in probability as one covariate changes around a value of importance while holding the other covariates fixed at values of importance. (P. 227-228) In the below results binary variables were standardized based on the x variable changing from 0 to 1 (the STDY command in Mplus), while the rest of the variables were standardized based on a one standard deviation change in the x variable (STDYX command in Mplus), as is recommended by Muthén, Muthén, and Asparouhov (2016). 143 Table 3.25. Standardized regression coefficients and standard errors for PSNT and nutrient maps (n=814). Individual Characteristics Years farming Education level College ag courses Knowledge Value of sales Household income Ag. income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Grain farm Other farm type Field slope Soy yield Corn yield Loam Clay Silty-loam Sandy-loam Other soil type Farm Characteristics—Social Cropland owned Days worked off farm Number of employees Partnership LLC S-corporation C-corporation Other business structure Conservation program participation Working land program participation PSNT Estimate (S.E.) Nutrient maps Estimate (S.E.) 0.050 (0.056) -0.059 (0.075) 0.101 (0.055)t 0.115 (0.141) -0.046 (0.090) 0.069 (0.043) 0.035 (0.042) 0.039 (0.080) 0.063 (0.057) -0.092 (0.243) -0.159 (0.361) -0.369 (0.232) -0.086 (0.069) 0.002 (0.056) 0.048 (0.046) -0.450 (0.140)** 0.064 (0.082) -0.257 (0.171) 0.081 (0.222) -0.097 (0.209) -0.046 (0.056) 0.016 (0.031) 0.016 (0.088) 0.250 (0.091)** -0.020 (0.063) -0.101 (0.377) 0.001 (0.256) 0.796 (0.498) 0.043 (0.191) 0.234 (0.155) 144 -0.011 (0.073) -0.057 (0.013)*** 0.215 (0.101)* 0.130 (0.057)* 0.103 (0.097) 0.049 (0.029)t 0.007 (0.031) -0.092 (0.178) 0.027 (0.064) -0.030 (0.113) -0.074 (0.444) -0.532 (0.241)* 0.026 (0.097) 0.071 (0.053) 0.178 (0.199) 0.210 (0.211) 0.121 (0.092) -0.021 (0.101) 0.107 (0.179) 0.420 (0.337) -0.091 (0.122) 0.019 (0.082) 0.095 (0.358) 0.163 (0.286) 0.064 (0.076) -0.086 (0.308) -0.028 (0.379) 0.101 (0.805) 0.220 (0.149) 0.045 (0.158) Table 3.25. (cont’d). Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values Look after environment Hunting and fishing environment Economic attitude Passing on good land Farming tradition Passing on farm Debt aversion Crop insurance use Environmental Awareness Concern ag. contributes to water problems Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts Intercept R-Square Chi-Square RMSEA CFI TLI -0.030 (0.022) 0.098 (0.109) 0.082 (0.085) 0.021 (0.061) 0.233 (0.019)*** 0.013 (0.031) 0.010 (0.052) 0.037 (0.042) 0.130 (0.157) 0.007 (0.054) 0.060 (0.079) -0.051 (0.030)t 0.184 (0.111)t 0.037 (0.117) -0.071 (0.064) -0.005 (0.069) 0.112 (0.019)*** 0.080 (0.027)** -0.061 (0.037) -0.040 (0.028) -0.034 (0.073) -3.489 (0.903)*** 0.349 (0.073)*** 679.695*** 0.033 0.839 0.812 0.128 (0.166) 0.039 (0.041) 0.096 (0.122) -0.092 (0.040)* 0.135 (0.055)* 0.025 (0.068) 0.064 (0.043) -0.069 (0.031)* -0.008 (0.072) -0.001 (0.059) -0.004 (0.028) -0.022 (0.024) -0.051 (0.016)** 0.012 (0.121) -0.036 (0.048) 0.073 (0.037)* 0.094 (0.013)*** 0.011 (0.066) 0.064 (0.071) 0.011 (0.054) -0.094 (0.042)* -2.329 (0.591)*** 0.371 (0.121)** 678.584*** 0.033 0.839 0.812 t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 Note: Binary variable standardization is based on a change from 0 to 1, rather than a one standard deviation change as is used for the non-binary variables. 145 Table 3.26. Standardized regression coefficients and standard errors for variable rate N and variable rate P or K (n=814). Variable rate N Estimate (S.E.) Variable rate P or K Estimate (S.E.) 0.012 (0.035) 0.025 (0.052) -0.164 (0.147) 0.086 (0.073) 0.034 (0.052) 0.018 (0.063) -0.091 (0.053)t 0.163 (0.064)* -0.003 (0.070) -0.202 (0.151) -0.035 (0.261) -0.363 (0.265) -0.035 (0.054) 0.016 (0.073) 0.110 (0.013)*** 0.026 (0.203) -0.152 (0.105) -0.328 (0.134)* 0.080 (0.113) -0.186 (0.137) -0.002 (0.064) 0.000 (0.055) 0.004 (0.093) -0.012 (0.321) 0.268 (0.060)*** 0.056 (0.086) -0.044 (0.148) -0.352 (0.768) -0.026 (0.115) -0.002 (0.077) -0.116 (0.068)t 0.108 (0.138) 0.164 (0.044)*** 0.143 (0.056)* 0.004 (0.055) -0.058 (0.039) 0.026 (0.101) -0.011 (0.139) -0.110 (0.091) -0.099 (0.118) -0.379 (0.214)t 0.012 (0.043) 0.076 (0.076) 0.178 (0.077)* 0.180 (0.122) -0.217 (0.196) -0.083 (0.081) 0.272 (0.148)t 0.201 (0.277) 0.005 (0.028) -0.025 (0.008)** -0.024 (0.035) -0.106 (0.133) 0.097 (0.098) -0.072 (0.173) -0.025 (0.155) -0.064 (0.380) 0.232 (0.065)*** -0.089 (0.156) Individual Characteristics Years farming Education level College ag courses Knowledge Value of sales Household income Ag. income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Grain farm Other farm type Field slope Soy yield Corn yield Loam Clay Silty-loam Sandy-loam Other soil type Farm Characteristics—Social Cropland owned Days worked off farm Number of employees Partnership LLC S-corporation C-corporation Other business structure Conservation program participation Working land program participation 0.206 (0.053)*** 146 Table 3.26. (cont’d). Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values Look after environment Hunting and fishing environment Economic attitude Passing on good land Farming tradition Passing on farm Debt aversion Crop insurance use Environmental Awareness Concern ag. contributes to water problems Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts Intercept R-Square Chi-Square RMSEA CFI TLI 0.074 (0.011)*** 0.034 (0.062) -0.008 (0.077) 0.047 (0.078) 0.114 (0.048)* 0.015 (0.089) -0.070 (0.050) 0.093 (0.074) 0.002 (0.104) -0.040 (0.035) 0.105 (0.045)* -0.029 (0.019) -0.001 (0.034) 0.220 (0.065)** 0.006 (0.039) -0.038 (0.044) -0.003 (0.043) 0.055 (0.053) -0.021 (0.084) 0.043 (0.013)** -0.073 (0.103) -2.514 (0.323)*** 0.237 (0.049)*** 679.410*** 0.033 0.840 0.813 0.122 (0.129) -0.076 (0.070) 0.133 (0.028)*** -0.054 (0.050) 0.086 (0.03)** -0.064 (0.03)* 0.027 (0.051) -0.033 (0.033) 0.022 (0.066) -0.003 (0.018) 0.009 (0.007) -0.076 (0.061) -0.006 (0.053) 0.058 (0.076) -0.098 (0.051)t 0.102 (0.029)*** 0.042 (0.028) 0.025 (0.062) 0.042 (0.062) 0.002 (0.033) 0.001 (0.019) -2.292 (0.779)** 0.31 (0.028)*** 677.040*** 0.033 0.840 0.813 t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 Note: Binary variable standardization is based on a change from 0 to 1, rather than a one standard deviation change as is used for the non-binary variables. 147 Table 3.27. Standardized regression coefficients and standard errors for cover crops (n=814). Individual Characteristics Years farming Education level College ag courses Knowledge Value of sales Household income Ag. income Farm Characteristics—Physical Cropland area Corn-soy ratio Animal farm Grain farm Other farm type Field slope Soy yield Corn yield Loam Clay Silty-loam Sandy-loam Other soil type Farm Characteristics—Social Cropland owned Days worked off farm Number of employees Partnership LLC S-corporation C-corporation Other business structure Conservation program participation Working land program participation 148 Cover crops Estimate (S.E.) 0.020 (0.069) 0.029 (0.049) 0.090 (0.053)t 0.168 (0.059)** 0.022 (0.027) 0.032 (0.040) 0.037 (0.067) -0.084 (0.044)t -0.045 (0.021)* 0.342 (0.130)** 0.106 (0.234) 0.275 (0.120)* 0.034 (0.012)** -0.021 (0.074) -0.155 (0.060)* 0.074 (0.114) 0.091 (0.109) -0.185 (0.042)*** 0.033 (0.086) 0.279 (0.201) -0.040 (0.043) 0.004 (0.090) -0.017 (0.063) 0.117 (0.105) 0.169 (0.048)*** -0.050 (0.183) 0.135 (0.189) 0.234 (0.685) 0.099 (0.083) 0.513 (0.099)*** Table 3.27. (cont’d). Information Info index Associations info source Chemical dealers info source Seed dealers info source Crop consultants info source Other farmers info source Campus extension info source Local extension info source Attitudes and Values Look after environment Hunting and fishing environment Economic attitude Passing on good land Farming tradition Passing on farm Debt aversion Crop insurance use Environmental Awareness Concern ag. contributes to water problems Erosion impacts Pest resistances impacts Extreme weather impacts Warming impacts Intercept R-Square Chi-Square RMSEA CFI TLI -0.054 (0.064) 0.084 (0.021)*** 0.000 (0.038) -0.044 (0.044) 0.102 (0.040)* 0.057 (0.039) -0.036 (0.047) 0.160 (0.021)*** 0.150 (0.058)* 0.026 (0.080) -0.142 (0.050)** 0.046 (0.039) -0.026 (0.048) 0.102 (0.088) -0.022 (0.028) -0.056 (0.039) 0.075 (0.047) 0.196 (0.017)*** -0.100 (0.038)** 0.010 (0.024) -0.104 (0.063)t -2.218 (0.896)* 0.338 (0.020)*** 681.087*** 0.033 0.841 0.814 t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 Note: Binary variable standardization is based on a change from 0 to 1, rather than a one standard deviation change as is used for the non-binary variables. 149 ICC Code APPENDIX D: Mplus Code Using WLSMV estimator with binary variable USEVARIABLES ARE c1a_use2; MISSING ARE ALL (-99); CATEGORICAL ARE c1a_use2; CLUSTER IS stateansi; ANALYSIS: ESTIMATOR=WLSMV; TYPE IS TWOLEVEL; MODEL: %WITHIN% %BETWEEN% Using MLR estimator with continuous variable USEVARIABLES ARE c1a_use2; MISSING ARE ALL (-99); CLUSTER IS stateansi; ANALYSIS: ESTIMATOR=MLR; TYPE IS TWOLEVEL; MODEL: %WITHIN% %BETWEEN% Model Code *example using c1d_use2; *these models are called _final7_v6 in my dissertation folder for each practice; DATA: FILE IS "C:\Users\rchdenny\Dropbox\Riva's Dissertation\mplus\mplus8.dat"; NOBSERVATIONS ARE 814; VARIABLE: !variable names in NAMES ARE section must match the order of the variables in the dataset, in USEVARIABLES ARE they can be in any order; NAMES ARE a5 a7a b6 c2 d1cedu d1ccamp d1cchem d1cseed d1cind d1cfam d1cass d2debt d2weal d2inco d2hunt d2envi d2prof d2land d2trad e1fsoil e1fresi e1fextr e1fwarm e1ccont e1calga e1chypo e1csoil f4 f7 f8 f9 f10 f11 stateansi IL IN MI OH crpown_pc crop2size corn_soy av_corn av_soy wrkldprg crop_ins animal frm_oth employ sole_prop partnshp llc s_corp c_corp other_bs loam clay clay_loam silty_loam sandy_loam oth_soil info_tot f5_rc yrsfrm c1a_use2 c1b_use2 c1c_use2 c1d_use2 c1e_use2 c1k_use2; USEVARIABLES ARE c1d_use2 yrsfrm f4 f5_rc c2 f9 f10 f11 crop2size corn_soy animal a7a frm_oth b6 av_soy av_corn loam clay silty_loam sandy_loam oth_soil crpown_pc f8 employ partnshp llc s_corp c_corp other_bs a5 wrkldprg info_tot d1cass d1cchem 150 d1cseed d1cind d1cfam d1ccamp d1cedu d2envi d2hunt d2weal d2prof d2inco d2land d2trad f7 d2debt crop_ins e1ccont e1calga e1chypo e1csoil e1fsoil e1fresi e1fextr e1fwarm; MISSING ARE ALL (-99); CATEGORICAL ARE c1d_use2 e1ccont e1calga e1chypo e1csoil d2weal d2prof d2inco; CLUSTER IS stateansi; ANALYSIS: ESTIMATOR = WLMSV; ITERATIONS = 1000; CONVERGENCE = .00005; TYPE IS COMPLEX; !this accounts for the clustering of the data in the standard errors; PARAMETERIZATION = THETA; MODEL: ECON_AT BY d2inco d2weal d2prof; WATERCON BY e1ccont e1calga e1chypo e1csoil; e1calga WITH e1chypo; c1d_use2 ON yrsfrm f4 f5_rc c2 f9 f10 f11 crop2size corn_soy animal a7a frm_oth b6 av_soy av_corn loam clay silty_loam sandy_loam oth_soil crpown_pc f8 employ partnshp llc s_corp c_corp other_bs a5 wrkldprg info_tot d1cass d1cchem d1cseed d1cind d1cfam d1ccamp d1cedu d2envi d2hunt ECON_AT d2land d2trad f7 d2debt crop_ins WATERCON e1fsoil e1fresi e1fextr e1fwarm; c1d_use2@1; !fixes variance to 1; OUTPUT: STDY; STDYX; TECH4; 151 REFERENCES 152 REFERENCES Ajzen, Icek. 1991. “The Theory of Planned Behavior.” Orgnizational Behavior and Human Decision Processes 50:179–211. Arbuckle, J. G. and G. Roesch-McNally. 2015. “Cover Crop Adoption in Iowa: The Role of Perceived Practice Characteristics.” Journal of Soil and Water Conservation 70(6):418–29. Bandalos, Deborah L. 2014. “Relative Performance of Categorical Diagonally Weighted Least Squares and Robust Maximum Likelihood Estimation.” Structural Equation Modeling 21(1):102–16. Baumgart-Getz, Adam, Linda Stalker Prokopy, and Kristin Floress. 2012. “Why Farmers Adopt Best Management Practice in the United States: A Meta-Analysis of the Adoption Literature.” Journal of Environmental Management 96(1):17–25. Beauducel, Andre and Philipp Yorck Herzberg. 2006. “On the Performance of Maximum Likelihood Versus Means and Variance Adjusted Weighted Least Squares Estimation in CFA.” Structural Equation Modeling 13(2):186–203. Bell, Andrew and Kelvyn Jones. 2015. “Explaining Fixed Effects: Random Effects Modeling of Time-Series Cross-Sectional and Panel Data.” Political Science Research and Methods 3(1):133–53. Carolan, Michael S. 2006. “Do You See What I See? Examining the Epistemic Barriers to Sustainable Agriculture.” Rural Sociology 71(2):232–60. Dillman, D. A., J. D. Smyth, and L. M. Christian. 2014. Internet, Phone, Mail, and Mixed-Mode Surveys: The Tailored Design Method. Hoboken, NJ: John Wiley & Sons. Duff, Scott N., D. Pete. Stonehouse, Don J. Blackburn, and Stewart G. Hilts. 1992. “A Framework for Targeting Soil Conservation Policy.” Journal of Rural Studies 8(4):399– 410. Dunn, M., J. D Ulrich-Schad, L. S. Prokopy, R. L. Myers, C. R. Watts, and K. Scanlon. 2016. “Perceptions and Use of Cover Crops among Early Adopters: Findings from a National Survey.” Journal of Soil and Water Conservation 71(1):29–40. Ervin, Christine A. and David E. Ervin. 1982. “Factors Affecting the Use of Soil Conservation Practices: Hypotheses, Evidence and Policy Implications.” Land Economics 58(3):277–92. Fishbein, Martin and Icek Ajzen. 2015. Predicting and Changing Behavior: The Reasoned Action Approach. New York, NY: Routledge. Holmes, Roger, David G. Armanini, and Adam G. Yates. 2016. “Effects of Best Management Practice on Ecological Condition: Does Location Matter?” Environmental Management 153 57(5):1062–76. Kaiser, F. G., S. Wölfing, and U. Fuhrer. 1999. “Environmental Attitude and Ecological Behavior.” Journal of Environmental Psychology 19:1–19. Kaiser, Florian G. and Urs Fuhrer. 2003. “Ecological Behavior’s Dependency on Different Forms of Knowledge.” Applied Psychology 52(4):598–613. Knowler, Duncan and Ben Bradshaw. 2007. “Farmers’ Adoption of Conservation Agriculture: A Review and Synthesis of Recent Research.” Food Policy 32(1):25–48. Lai, Mark H. C. and Oi-man Kwok. 2015. “Examining the Rule of Thumb of Not Using Multilevel Modeling: The ‘Design Effect Smaller Than Two’ Rule.” The Journal of Experimental Education 83(3):423–38. Lam, Q. D., B. Schmalz, and H. Fohrer. 2011. “The Impact of Agricultural Best Management Practices on Water Quality in a North German Lowland Catchment.” Environmental Monitoring and Assessment 183(1–4):351–79. Li, Cheng Hsien. 2016. “The Performance of ML, DWLS, and ULS Estimation with Robust Corrections in Structural Equation Models with Ordinal Variables.” Psychological Methods 21(3):369–87. Liu, Tingting, Randall Bruins, and Matthew Heberling. 2018. “Factors Influencing Farmers’ Adoption of Best Management Practices: A Review and Synthesis.” Sustainability 10(2):432. Liu, Yongbo, Wanhong Yang, Luis Leon, Isaac Wong, Craig McCrimmon, Alice Dove, and Phil Fong. 2016. “Hydrologic Modeling and Evaluation of Best Management Practice Scenarios for the Grand River Watershed in Southern Ontario.” Journal of Great Lakes Research 42(6):1289–1301. Maas, Cora J. and Joop Hox. 2005. “Sufficient Sample Sizes for Multilevel Modeling.” Journal of Research Methods for the Behavioral and Social Sciences 1:86–92. Madden, Thomas J., Pamela Scholder Ellen, and Icek Ajzen. 1992. “A Comparison of the Theory of Planned Behavior and the Theory of Reasoned Action.” Personality and Social Psychology Bulletin 18(1):3–9. Muthén, B. O., L. K. Muthén, and T. Asparouhov. 2016. Regression and Mediation Analysis Using Mplus. Los Angeles, CA: Muthén & Muthén. Muthén, L. K. and B. O. Muthén. 2017. Mplus User’s Guide. 8th ed. Los Angeles, CA: Muthén & Muthén. (https://www.statmodel.com/download/usersguide/MplusUserGuideVer_8.pdf). Napier, TL, CS Thraen, and A. Gore. 1984. “Factors Affecting Adoption of Conventional and Conservation Tillage Practices in Ohio.” Journal of Soil and Water 39(3):205–9. 154 Nguyen, Trang Quynh, Yenny Webb-Vargas, Ina M. Koning, and Elizabeth A. Stuart. 2016. “Causal Mediation Analysis With a Binary Outcome and Multiple Continuous or Ordinal Mediators: Simulations and Application to an Alcohol Intervention.” Structural Equation Modeling 23(3):368–83. Nowak, Peter J. 1987. “The Adoption of Agricultural Conservation Technologies: Economic and Diffusion Explanations.” Rural Sociology 52(2):208–20. Nowak, Peter J. and Peter F. Korsching. 1985. “Conservation Tillage: Revolution or Evolution?” Journal of Soil and Water Conservation 40(2):199–201. Price, Jennifer C. and Zoe Leviston. 2014. “Predicting Pro-Environmental Agricultural Practices: The Social, Psychological and Contextual Influences on Land Management.” Journal of Rural Studies 34(0):65–78. Prokopy, L. S., K. Floress, D. Klotthor-Weinkauf, and A. Baumgart-Getz. 2008. “Determinants of Agricultural BMP Adoption: Evidence from the Literature.” Journal of Soil and Water Conservation 63(5):300–311. Reimer, Adam P. Aaron Thompson, Linda Stalker Prokopy, J. Gordon Arbuckle, Ken Genskow, Douglas Jackson-Smith, Gary Lynne, Laura McCann, Lois Wright Morton, and Pete Nowak. 2014. “People, Place, Behavior, and Context: A Research Agenda for Expanding Our Understanding of What Motivates Farmers’ Conservation Behaviors.” Journal of Soil and Water Conservation 69(2):57A–61A. Reimer, Adam P. and Linda S. Prokopy. 2012. “Environmental Attitudes and Drift Reduction Behavior among Commercial Pesticide Applicators in a US Agricultural Landscape.” Journal of Environmental Management 113:361–69. Reimer, Adam P. and Linda S. Prokopy. 2014. “Farmer Participation in US Farm Bill Conservation Programs.” Environmental Management 53(2):318–32. Reimer, Adam P., Denise K. Weinkauf, and Linda S. Prokopy. 2012. “The Influence of Perceptions of Practice Characteristics: An Examination of Agricultural Best Management Practice Adoption in Two Indiana Watersheds.” Journal of Rural Studies 28(1):118–28. Rhemtulla, Mijke, Patricia É. Brosseau-Liard, and Victoria Savalei. 2012. “When Can Categorical Variables Be Treated as Continuous? A Comparison of Robust Continuous and Categorical SEM Estimation Methods under Suboptimal Conditions.” Psychological Methods 17(3):354–73. Rogers, Everett M. 1958. “Categorizing the Adopters of Agricultural Practices.” Rural Sociology 23(4):346–54. Rogers, Everett M. 2003. Diffusion of Innovations. Fifth. New York, NY: Free Press. Rollins, Timothy. 1993. “Using The Innovation Adoption Diffusion Model To Target Educational Programming.” Journal of Agricultural Education 34(4):46–54. 155 Ruttan, Vernon W. 1996. “What Happened to Technology Adoption-Diffusion Research?” Sociologia Ruralis 36(1):51–73. Ryan, Bryce. 1948. “A Study in Technological Diffusion.” Rural Sociology 13(3):273–85. Ryan, Bryce and Neal C. Gross. 1943. “The Diffusion of Hybrid Seed Corn in Two Iowa Communities.” Rural Sociology 8(1):15–24. Snijders, Tom A. B. and Roel Bosker. 2012. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. 2nd ed. Thousand Oaks, CA: SAGE. Sowa, Scott P. Matthew Herbert, Sagar Mysorekar, Gust M. Annis, Kimberly Hall, A. Pouyan Nejadhashemi, Sean A. Woznicki, Lizhu Wang, and Patrick J. Doran. 2016. “How Much Conservation Is Enough? Defining Implementation Goals for Healthy Fish Communities in Agricultural Rivers.” Journal of Great Lakes Research 42(6):1302–21. Stuart, D., R. C. H. Denny, M. Houser, A. P. Reimer, and S. Marquart-Pyatt. 2018. “Farmer Selection of Sources of Information for Nitrogen Management in the US Midwest: Implications for Environmental Programs.” Land Use Policy 70:289–97. Ulrich-Schad, J. D., S. García de Jalón, N. Babin, A. Pape, and L. S. Prokopy. 2017. “Measuring and Understanding Agricultural Producers’ Adoption of Nutrient Best Management Practices.” Journal of Soil and Water Conservation 72(5):506–18. Vitale, J. D., C. Godsey, J. Edwards, and R. Taylor. 2011. “The Adoption of Conservation Tillage Practices in Oklahoma: Findings from a Producer Survey.” Journal of Soil and Water Conservation 66(4):250–64. Wauters, Erwin and Erik Mathijs. 2014. “The Adoption of Farm Level Soil Conservation Practices in Developed Countries: A Meta-Analytic Review.” International Journal of Agricultural Resources, Governance and Ecology 10(1):78–102. Weber, Catharine and Laura McCann. 2015. “Adoption of Nitrogen-Efficient Technologies by U.S. Corn Farmers.” Journal of Environment Quality 44(2):391–401. Werner, Magdalena, Erwin Wauters, Jo Bijttebier, Horst-Henning Steinmann, Greet Ruysschaert, and Andrea Knierim. 2017. “Farm Level Implementation of Soil Conservation Measures: Farmers’ Beliefs and Intentions.” Renewable Agriculture and Food Systems 1– 14. 156 CHAPTER 4: QUESTIONING THE FERTILIZER SAVINGS INCENTIVE FOR BMP ADOPTION IN THE NORTH CENTRAL REGION Agriculture is of the utmost importance to human societies for its production of food and fiber (and increasingly fuel), but it is also a source of environmental harm as a producer of non- point-source pollution to the air and water (Aneja, Schlesinger, and Erisman 2009; Carpenter et al. 1998; Drinkwater and Snapp 2007; Lal 1993; Robertson et al. 2013; Robertson and Vitousek 2009; Uri and Lewis 1998). Nutrients like nitrogen (N) and phosphorus (P) are vital to crop production but also create great environmental costs when they leave the farm and enter surrounding waterways or the atmosphere, where they contribute to eutrophication and hypoxia (“dead zones”) and global climate change respectively (Blesh and Drinkwater 2013; Robertson and Vitousek 2009). Efforts to reduce non-point-source nutrient pollution from agriculture have long focused on getting farmers to voluntarily adopt a range of best management practices (BMPs) either in the field to help retain nutrients or outside of fields to intercept nutrients. The retention of nutrients by in-field practices means that they have the potential to reduce the amount of fertilizer that a farmer needs to apply to get a good crop (Robertson and Vitousek 2009), thereby saving farmers money and reducing non-point-source nutrient pollution from agriculture. These nutrient savings are commonly assumed to act as an incentive for farmers to use these in-field practices and as an off-set for adoption costs, thus reducing the need for outside, particularly monetary, incentives (Odum 1984; Robertson and Harwood 2013; Robertson and Vitousek 2009). However, it is not clear to just what extent farmers actually do reduce their fertilizer use when they adopt any of these practices. If farmers are not experiencing tangible savings from the use of these practices, then it is likely that the present incentive system for their use is insufficient for encouraging their adoption. The present study examines the effects of three in-field practices (cover crops, conservation tillage, and no-till) on fertilizer use in the North Central Region of the US using county and state-level data. The use of this type of aggregated data, rather than farm-level or field-level data, represents an important move towards considering the overall environmental effects of agriculture and BMPs. While it is important to understand the farm-level dynamics surrounding fertilizer and BMP use, larger scale data lets us evaluate the cumulative effects of the agricultural practices of individual farmers. In this case: are enough farmers reducing their fertilizer use on their farms by a large enough amount when they use these BMPs to all together have a measurable effect on the environment? The use of aggregated data also moves the analysis in a different direction than the theoretical approaches commonly used to explain the drivers of BMP adoption: the Theory of Planned Behavior and related theories (Ajzen 1991; Fishbein and Ajzen 2015), and the Diffusion of Innovations Model (Rogers 2003). Variables these approaches have in common, such as attitudes towards a practice and personal values, are not easily attainable at the county level. Rather, biophysical characteristics become important contextual variables for the individual decisions that are aggregated, and important controls for county differences over a multi-state area. A coupled natural and human systems approach is used instead to inform the model design, which includes both social and biophysical variables. 158 BACKGROUND Fertilizer and BMPs As previously mentioned, the use of large amounts of concentrated chemical fertilizers, a hallmark of contemporary agriculture in much of the “developed world,” has greatly increased crop yields, but has also exacerbated the environmental problems from non-point source nutrient pollution (Robertson and Vitousek 2009). Nitrogen (N), and Phosphorus (P) are two of the most important nutrients that plants need and are the two nutrients that contribute most to non-point source nutrient pollution from agriculture, making them the focus of most nutrient management efforts (Kleinman et al. 2011; Robertson and Vitousek 2009). Nitrogen in particular is a critical nutrient in determining corn grain yields (Mahama et al. 2016), and its many, highly mobile forms make it especially challenging to manage (Robertson and Groffman 2015; Robertson and Vitousek 2009). To maximize profit and minimize environmental harm farmers are currently being encouraged to apply fertilizer of all kinds according to the 4R’s: right time, right place, right source, and right rate (Bruulsema, Lemunyon, and Herz 2009; Roberts and Johnston 2015; Snyder 2017). Applying fertilizer at the right time means applying it in accordance with the seasonal needs of the crop to ensure that the nutrients are available when the crop needs it and to reduce loss; this can include applying fertilizer in the spring instead of the fall, applying it during or after planting, or applying it multiple times. The right place means applying nutrients as close as possible to the plants’ roots, and particularly under the soil surface. The right source of nutrients refers to the formulation or product used, and particularly for N can include using a slow release product or a nitrification inhibitor (Snyder 2017). The right rate refers to applying the correct amount of the product to meet but not exceed crop needs, which is facilitated by the 159 use of decision-aids like soil or plant tissue tests that indicate the amount of nutrients available and variable rate technology. Some of these, primarily technology-based, practices serve more than one purpose, like slow-release products that relate to source and timing, and variable rate that relates to place and rate. Due to the interconnectedness of the 4Rs, changes to one means that the others should be re-assessed and also perhaps changed accordingly (Bruulsema et al. 2009). In addition to the 4R BMPs discussed above, which are primarily technology-based, biologically-based practices function by supporting and enhancing the composition of, and biological functions in, the soil to increase nutrient availability and at the same time reduce nutrient losses (Blesh and Drinkwater 2013; Drinkwater and Snapp 2007; Robertson et al. 2014; Robertson and Vitousek 2009). These practices include crop rotation, off-season cover crops, and reduced/conservation tillage or no-till, which all have the potential to increase the amount of organic matter in the soil, which in turn has an important function in retaining nitrogen and phosphorus and making them available to later crops (Blesh and Drinkwater 2013; Drinkwater and Snapp 2007; Robertson and Vitousek 2009; Vandermeer 2011). Diverse, multi-year crop rotations, particularly those that include legumes, have been shown to increase grain yields over time and to produce high yields with lower external N inputs (Coulter et al. 2011; Osterholz, Liebman, and Castellano 2018; Stanger and Lauer 2008), as well as reduce N losses (Gardner and Drinkwater 2009). Cover crops, non-harvested crops grown between the harvest and planting of cash-crops, not only provide crop rotation benefits, but they also add the benefit of living soil cover during the time that the field would otherwise be bare. The cover crop takes up available N remaining in the field after the cash crop is harvested, thus preventing it from being washed out of the soil over the winter; when the cover crop is killed 160 prior to planting the next cash-crop the N remains in the field and is slowly released as the cover crop residue decomposes (Robertson and Vitousek 2009). Similarly to diverse crop rotations, cover crops have been shown to reduce if not fully replace the need for additions N inputs (Doane et al. 2009; Mahama et al. 2016; Robertson and Vitousek 2009). In addition, the soil cover provided by living or dead cover crops also helps reduce soil erosion (Dabney, Delgado, and Reeves 2001), and thus retain P (Gentry et al. 2007; Yuan et al. 2018). Conservation tillage and no-till represent additional in-field practices29 that can retain nutrients and thus reduce the amount of additional nutrients that need to be applied. Conservation tillage is typically defined as leaving at least 30 percent residue cover on the soil after planting (CTIC 2002) often through the use of a non-inversion plows or harrows (see Chapter 2). No-till involves the minimal disturbance of the soil surface and crop residue between harvest and planting (CTIC 2002)30. Conservation tillage and no-till have long been shown to reduce soil erosion by increasing soil surface roughness and retaining more crop residue on the soil surface (Allmaras, Unger, and Wilkins 1985). Reducing soil erosion is a key mechanism for reducing P loss, though dissolved P can also be lost through tile drainage (Gentry et al. 2007; Yuan et al. 2018). Conservation tillage and no-till have also been shown to increase soil organic carbon, which increases the amount of N that can be stored in the soil (Grandy et al. 2006; Mazzoncini et al. 2011), at least in the upper soil layers (Dolan et al. 2006; Du, Ren, and Hu 2010). Conservation tillage and no-till can have additional soil benefits related to soil properties, including better water infiltration and retention, and larger and more diverse microbial communities that also contribute to soil fertility (Busari et al. 2015; Robertson et al. 2014). 29 There is also a significant technological aspect to tillage, but the primary feature that relates to nutrient retention is based in the soil qualities that result, rather than the direct effect of the specific tool. 30 A related practice commonly called “strip-till” is often included in the no-till category, where strips of up to one- third of the row width may be disturbed to facilitate planting and fertilizer application (CTIC 2002; NRCS 2012). 161 The cost of implementing BMPs is a potential barrier to their use. Conservation tillage and no-till offer direct savings through reduced fuel use and equipment wear, in addition to soil and nutrient benefits, though these benefits are not always clear enough to farmers to entice them to use those practices (Andrews et al. 2013). However, cover crops have the up-front cost of the seed, as well as adding an additional activities and management to plant and kill them (Plastina et al. 2018; Robertson and Vitousek 2009; Singer, Nusser, and Alf 2007) and farmers have been found to be uncertain about the benefits they may gain from using them (Arbuckle and Roesch- McNally 2015). The high potential costs, both up-front and indirect, especially for cover crops but also for conservation tillage and no-till, combined with uncertainty in benefits indicates the importance of having an accurate understanding of the benefits that these practices may or may not provide to farmers in the form of fertilizer savings. While the use of 4R practices and biological in-field BMPs have been a common topic of inquiry in the BMP adoption literature, the actual rates of fertilizer application are rarely considered as an outcome variable of interest (Houser et al., in preparation). This is an important omission from the body of BMP adoption literature, as fertilizer rate is just as important (economically and environmentally) as the other 4Rs and is highly related to non-point-source nutrient loss from agriculture (Ribaudo et al. 2011). The key assumption is that farmers want to use these tools to reduce their fertilizer rate, but this assumption has not been systematically tested. Weber and McCann (2015:392) report: “Previous 2001 and 2005 ARMS data showed that corn farmers who use soil testing can reduce overall commercial fertilizer application compared with nonadopters … If N soil testing similarly reduced N applications in excess of crop needs, this would improve water quality” (italics added). They do not, however, discuss 162 whether farmers actually do reduce their nitrogen applications when using soil testing. Henke (2006) reports that very few California vegetable growers even adopted a quick test for nitrogen that was demonstrated to allow for reductions in fertilizer without reducing yields, let alone actually reduced their fertilizer application because of it. However, Reimer, Thompson, et al. (2012) report reduced input costs as a motivating factor for conservation adoption in their study of an Indiana watershed, suggesting that at least some farmers do adjust their fertilizer use as a result of using conservation practices. The question of whether or not BMPs reduce actual fertilizer use is an important one since an assumption of cost reduction is highly involved in assumptions of farmer motivations for practice adoption, and thus what other incentives might be needed to encourage a farmer to adopt them. The potentially false assumption is that farmers will save money on fertilizer when they use conservation practices and that these savings will substantially offset the cost of adopting the practice (Odum 1984; Robertson and Harwood 2013; Robertson and Vitousek 2009), thereby assuming a reduction in the incentive needed. To be clear, agronomically the conservation practices do reduce fertilizer needs, but the question is how much do farmers believe this to be true and to what extent does that offset their perceived need for applying extra fertilizer. Coupled Natural and Human Systems A coupled human and natural system (CHANS), also often called a social-ecological system (SES), is most basically a “systems in which human and natural components interact” (Liu, Dietz, Carpenter, Folke, et al. 2007:639). While the two terms have somewhat different origins and disciplinary tendencies, the common defining characteristics of both are reciprocal effects and relationships across time and space, and the terms are frequently used 163 interchangeably (An and López-Carr 2012; Liu, Dietz, Carpenter, Alberti, et al. 2007; Liu, Dietz, Carpenter, Folke, et al. 2007). In studies that take this broadly defined CHANS/SES approach (henceforth referred to a coupled natural and human [CNH] system), the researcher(s) bring ecological variables into the heart of the study, especially allowing them to influence social actions and systems, either as exogenous variables or as endogenous variables as part of a feedback loop (see Figure 4.1). So far, few researchers have explicitly taken a CNH approach to understanding BMP use in industrialized agriculture, though Stuart et al. (2015) have made the case for doing so, and other studies have incorporated social and biophysical factors as drivers of human actions without referring to CHANS and without considering feedback effects (Blesh and Wolf 2014; Duff et al. 1992; Halbrendt et al. 2014; Houser, Stuart, and Carolan 2017; Stonehouse 1996). Figure 4.1. Diagram of feedback of CHN feedback in an agricultural system from Stuart et al. (2015:573). 164 For example Duff et al. (1992) and Stonehouse (1996) conceptualize a linked cultural and biophysical system where individual farmers and their farms are embedded in cultural and institutional settings and regional biophysical settings respectively. Stonehouse, more than Duff et al., conceptualizes farmer decision making as being dynamically embedded in this larger setting and as being both influenced by and influencing this larger context, that is, he recognizes the presence of feedbacks. Neither study makes any attempt to empirically test the conceptual models. More recently, Halbrendt et al. (2014), Blesh and Wolf (2014), and Houser et al. (2017) have also integrated social and biophysical factors in empirical studies, using mental models and qualitative interviews respectively. All three consider how the biophysical context of the agricultural system of interest influences farmers’ perceptions and farming practices. In the present study, I follow this tradition, but use a quantitative analysis. I include the environmental context in which farmers are operating and making decisions as an important control for their fertilizer use—both directly and indirectly through influencing corn yield. RESEARCH PROBLEM AND QUESTIONS The core question of this research is: Does the use of BMPs reduce fertilizer use? Anecdotally there are examples of this happening (Reimer, Thompson, et al. 2012), but it does not appear to have been studied systematically, so it is unclear how common or widespread this result is. More specifically: Are enough farmers reducing their fertilizer use enough for it to show up at a great-than-farm-size scale? The implications of this question are two-fold: 1) are the fertilizer reductions great enough to function as an incentive for farmers to adopt the practice and/or offset the cost of adoption?, and 2) are the fertilizer reductions enough to reduce nutrient loads to local waterbodies? 165 Fertilizer rate is an important aspect of the potential for nutrient losses from agricultural fields to local waterbodies, but has rarely been the outcome of interest in analyses (Houser et al. in preparation). Reducing fertilizer rates to just meet crop needs is an important method of reducing nutrient losses to the environment and can potentially be done without reducing crop yields (Davis et al. 2000; Hoben et al. 2011; McSwiney and Robertson 2005; Moriasi et al. 2013; Rocha, Roebeling, and Rial-Rivas 2015; Teshager et al. 2017). Much agricultural research is done at the farmer scale, and this seems to be particularly true of research on the adoption of soil and water conservation practices where a focus on attitudes as explanations for conservation practice adoption or conservation program participation is common (cf. Arbuckle et al. 2013; Arbuckle and Roesch-McNally 2015; A. Reimer and Prokopy 2014; Reimer, Thompson, et al. 2012; Reimer, Weinkauf, et al. 2012; Roesch-McNally, Gordon Arbuckle, and Tyndall 2017). However, the individual scale of these studies limits their usefulness when it comes to understanding the potential for these practices to have an effect at larger scales such as watersheds. While the decisions to adopt one or more conservation practices and how much fertilizer to apply are individual decisions this analysis is at the county-level. There are two reasons for the choice of scale used: 1) data availability, and 2) interest in the prevalence and potential impact of the phenomena. Using county level data allows for a large geographic sample, and to study the cumulative patterns of farmer management decisions to see if reducing fertilizer due to adopting a conservation practice happens often enough to be detected at the county level. If conservation practice use is not found to be significantly related to fertilizer use at the county-level, then it is unlikely that the conservation practices considered are having a significant impact in reducing the overall nutrient loss from agriculture at a watershed scale at 166 their current rate of use. An insignificant result would not necessarily indicate a lack of relationship between the conservation practices and fertilizer use. County-level data cannot tell us what is happening at the individual-level—for example we cannot distinguish between farmers not reducing their fertilizer use when using the conservation practices, and farmers substantially reducing their fertilizer use when using the conservation practices but doing so on too small a portion of the acres in the county to have a significant effect on the county total. METHODS AND ANALYSIS Data and sample The data for this analysis come from the 2012 Census of Agriculture (“AgCensus” hereafter) and from the PRISM Climate Group at Oregon State University (http://prism.oregonstate.edu) as further described below. The 2012 AgCensus is the first one to included questions on specific conservation practices, providing national data on the number of farms using each practice, as well as the number of acres under each one. These much more specific measures are an important step forward in the study of conservation practice adoption and use by being more specific than a simple binary of adopt or did not adopt that ignores the extensiveness of practice use; a step that was called for by Reimer et al. (2014). The sample used is the counties of the twelve North Central US states: Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin. Thirteen counties had to be excluded from the analysis due to either a lack of fertilized cropland or missing data31, leaving a final sample of 1,042 counties. In its reporting of the AgCensus, NASS withholds values in situations where the variable value is too close to the value of a specific operation (NASS 2012a). These missing, but not truly 31 See Appendix C for description of the excluded counties and descriptive statistics by state. 167 missing, values were replaced in the exogenous variables using the agriculture district32 mean for that variable (see Appendix A for details). In the one case where there were no observations for an agriculture district, the average of the counties immediately adjacent to the agriculture district were used instead (see Appendix A for details). Variables Outcome Variables. The main outcome variable of interest, average fertilizer use, is measured in dollars spent on fertilizer, including lime and soil conditioners per acre of fertilized cropland. The use of this dollar-based proxy is necessitated by the nature of the data available, and the complexity inherent in the large number of fertilizer products available containing different combinations and concentrations of nutrients. Independent Variables. The key independent variables in this analysis are the use of conservation tillage, no-till and cover crops, operationalized in this analysis as the proportion of cropland acres where each practice was used33. Conservation tillage is defined in the AgCensus as leaving at least 30 percent residue cover (NASS 2012c) and excludes no-till (NASS 2012b). No-till is described as “using no-till or minimum till” (NASS 2012a:B-13). Cover crops are “a crop planted primarily to manage soil fertility, soil quality, water, weeds, pests, diseases, or wildlife” (NASS 2012a:B-13), and do not include acres enrolled in the Conservation Reserve Program (CRP). 32 Agriculture districts are sub-state designations made by USDA NASS. There are 9 agriculture districts in each state in the sample (https://www.nass.usda.gov/Charts_and_Maps/Crops_County/boundary_maps/indexgif.php). 33 This formulation of independent variables as ratios was necessitated by the need to control for differences in total county size and more importantly the total area of farmland and cropland in each county without the introduction of the extreme multicollinearity that would result from including even one of these area measures as an independent variable along with total acres or counts. 168 Nineteen county-level control variables are included in the analysis and include farmer and farm characteristics, crop choice, and climate characteristics; all are group-mean centered34. The included variables were selected based on variables available in the AgCensus, literature on practice adoption, and agronomic principles. The farmer and farm characteristics are average years of farming experience, the portion of farmers who farm full time (full time famer ratio), median farm size, the ratio of rented county farmland to total county farmland (rented land ratio), the ratio of county acres under crop insurance to total county cropland acres (crop insurance acres ratio), and the ratio of irrigated county acres to total county cropland acres (irrigated acres ratio). Experience, farming full time, farm size, and amount of rented land are all common controls for BMP adoption analyses (Baumgart-Getz, Prokopy, and Floress 2012; Kara, Ribaudo, and Johansson 2008; Prokopy et al. 2008; Wauters and Mathijs 2014). Insurance and irrigation use are less common variables but both represent aspects of risk tolerance (Baumgart-Getz et al. 2012; Prokopy et al. 2008; Wauters and Mathijs 2014), and irrigation is an important aspect of growing conditions in conjunction with precipitation. Average farmland value is included as a measure of soil quality that is expected to have a positive effect on both corn yield and fertilizer spending, as high-quality soils have higher potential yields but only with sufficient fertilization. Because fertilization varies by crop (Kara et al. 2008; Wade, Claassen, and Wallander 2015), and some crops and livestock can contribute to soil fertility and thus influence fertilizer use (Robertson and Vitousek 2009), I include crop and livestock variables in the form of ratios of county cropland where they were grown and density of animals. Corn acres ratio is the 34 Corn yield is not centered because it is an endogenous variable, but controlling for state average corn yield (by including path between state corn yield and county fertilizer use) produces the same effect—effectively partitioning the effect of corn yield on fertilizer use between the county and the state (Bell and Jones 2015). See Appendix C for explanation of centering decisions. 169 proportion of cropland acres that was harvested for corn grain, and is considered on its own given that corn is a very common crop in the region and is given high rates of N fertilizer (Ribaudo et al. 2011). The legume acres ratio is the ratio of county acres in a legume, either soy beans or alfalfa, and is included as a control for fertilizer use, because legumes can contribute nitrogen to the cropping system (Robertson and Vitousek 2009) and because legume crops are an indication of a more diverse crop rotation, which can improve yield and reduce fertilizer needs (Coulter et al. 2011; Osterholz et al. 2018; Stanger and Lauer 2008). Three livestock variables, operationalized as density of animals, are included to control for manure use that could replace purchased fertilizers (Robertson and Vitousek 2009). The livestock variables are the average number of animals per acre of county cropland for poultry, including broilers, layers, and turkeys (poultry density), cattle, including beef cattle on feed and dairy cattle (cattle density), and hogs (hog density). The different kinds of animals are considered separately, since the volume and composition of manure varies by animal type. Lastly, three climate variables are included to control for the normal35 climatic conditions across the region and the effects of the 2012 drought. The normals are used because they represent the typical growing conditions. The normal annual corn growing degree days36 is the number in thousands for the county, the normal growing season precipitation (April-September) is the normal precipitation in inches for the county, and the 2012 growing season precipitation is the April-September 2012 precipitation for the county. 35 The “normal” is the 30-year average. For this analysis normals were calculated from daily temperature and precipitation data from 1982-2011, in accordance with NOAA description at https://www.ncdc.noaa.gov/data- access/land-based-station-data/land-based-datasets/climate-normals/1981-2010-normals-data. See Appendix A for more detail. 36 Corn growing degree days are a measure of the accumulated heat units in the temperature range in which plant growth occurs. For corn this temperature range is 50 °F to 86 °F (https://ndawn.ndsu.nodak.edu/help-corn-growing- degree-days.html). See Appendix A for more detail. 170 At the state level, average fertilizer use in dollars per acre is used as a control for county- level fertilizer use, and state average corn grain yield in bushels per acre is used as a control for county-level corn yield. Both the state-level variables are grand-mean centered as is standard in multilevel models (Enders and Tofighi 2007)37. Figure 4.2. Conceptual Multilevel SEM diagram of fertilizer use. Mediating variable. The mediating variable is average corn yield in bushels per acre. This variable, as a mediator, functions as both an outcome and an independent variable in the model. Expected yield is a guide that farmers use when determining how much fertilizer to apply. Corn yield is expected to be highly driven by soil quality, which is measured by the average farmland value. While corn yield is measured for the 2012 growing season, it is used in the model as a measure of the typical yield expected and which is thus the basis for the fertilizer 37 See Appendix C for explanation of centering decisions. 171 used. The growing season precipitation for 2012 is used as a control for corn yield to account for the effects of the 2012 drought that affected much of the study area. Analysis For this analysis I use a multilevel, random intercept, structural equation model (SEM) with county-level and state-level data. The state-level data is included in the intercept portion of the county-level equations. The key feature of a multilevel model is that it can include multiple random effects and accommodate data that includes variables at multiple levels (i.e., it is grouped or nested in a hierarchy) and not only control for but explain the variation that exists at those levels. Failing to account for the nested structure of one’s data and simply using an OLS model to analyze it typically results in underestimated standard errors, which increases the likelihood of type-1 errors (Bell and Jones 2015). The essential feature of nested, grouped or hierarchical data is that the observations are not all independent of each other; in other words, some are more alike than others. In the present case the grouping is related to counties sharing a state context (Jones and Duncan 1998). In this model, the multilevel features of the model that relate to the data structure are combined with the features of a structural equation model (SEM) to handle the specification of the desired conceptual relationships. SEM uses multiple, simultaneously estimated regression models and partitions variance among the different parts of the model. A notable feature of SEM is that it developed in-part from path modeling (Schumacker and Lomax 2010), so it has an integral visual element that makes it unique within the family of regression analysis. As the name suggests, SEM has a structural element and can accommodate multiple relationships between multiple variables including indirect effects, and reciprocal and feedback relationships, which makes it particularly well suited to modeling CNH systems as well as agricultural systems where many variables are 172 related to each other. SEM has not been used much to study agriculture or CNH systems, though its value for the latter purpose has been recognized (Fan et al. 2016; Marquart-Pyatt, Jorgenson, and Hamilton 2015), as well as its use as a communication tool in interdisciplinary research groups (Smith et al. 2014). SEM has also been recognized as being useful for modeling ecological systems (Fan et al. 2016; Grace et al. 2010). The effect of conservation practices on fertilizer use is the key outcome of interest, but the model includes corn yield as a mediator. The model also includes biophysical climate data among the control variables of corn yield and fertilizer use. Restricting the sample to corn-heavy states simplifies the analysis as the use of conservation practices and nitrogen fertilizer use and practices, especially rate and timing, are highly crop specific (Ribaudo et al. 2011; Ribaudo, Livingston, and Williamson 2012; Wade et al. 2015). RESULTS AND DISCUSSION Table 4.1 provides the descriptive statistics for the sample used in the analysis (see Table 4.5 in Appendix C. for select descriptives by state). Of particular interest in Table 4.1 are the average use of BMPs. On average, conservation tillage is used on 22 percent of cropland acres in a county but ranges up to just over 50 percent at the maximum. No-till has an even higher average rate of use at 28 percent, with a maximum close to 80 percent. Cover crops are only used on 3 percent of cropland acres on average, and 32 percent at most. Crop rotation is indicated by the distribution of crops grown, and a balanced corn-soy rotation appears to be quite common on average as indicated by the equal corn and legume acres ratios (32 percent). 173 Table 4.1. Uncentered descriptive statistics for analysis variables (n=1042 unless otherwise noted). Std Dev Minimum Maximum Mean Outcome Variables Average fertilizer use ($/acre) a Average corn yield (bu./acre) b Farmer and Farm Characteristics Rented land ratio Average farmland value ($1000/acre) Tile acres ratio Full-time farmer ratio Average years of farming experience Median farm size Irrigated acres ratio Crop insurance acres ratio Crops Corn acres ratio Legume acres ratio Poultry density Cattle density Hog density BMPs Conservation tillage ratio No-till ratio Cover crops ratio Climate Normal growing season precipitation 2012 growing season precipitation Normal annual growing degree days (in 1000s) State-level Variables Average state fertilizer use ($/acre) Average state corn yield (bu./acre) Notes: a. n=1008 b. n=1014 33.214 14.000 0.035 0.420 0.000 0.249 18.000 2.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 10.506 3.540 1853.850 72.346 68.545 122.994 103.479 0.446 3.868 0.189 0.510 26.827 44.118 39.112 0.136 2.057 0.233 0.096 1.917 229.014 325.380 0.141 0.221 0.181 0.133 8.782 0.063 0.345 0.114 0.185 0.029 3.851 4.202 663.122 24.870 24.586 0.067 0.617 0.315 0.321 1.589 0.054 0.194 0.221 0.277 0.026 21.622 16.042 3362.570 121.721 111.382 784.530 193.722 0.821 17.113 0.844 0.838 33.700 3900.000 0.875 2.253 0.771 0.638 193.220 0.343 3.730 0.512 0.787 0.318 30.396 32.960 4993.650 150.198 156.041 Results are presented in Table 4.2, organized by endogenous variable. The unstandardized coefficients of the ratio variables can be meaningfully interpreted by considering a tenth of a unit change (i.e., a 10 percent change in area) in the independent variable ratio and in the outcome variable unit. For example, a 0.1 unit increase in the irrigated acres ratio in a county is associated with a 12.3 bushel per acre higher average corn yields. Average corn yield is significantly and positively predicted by the average farmland value, the cattle density, the irrigated acres ratio, and the tile acres ratio, which are all consistent with expectations. Higher value land produces more corn, while ensuring sufficient water through irrigation and sufficient drainage through tile drainage also contributes to higher corn yields. Interestingly, cattle density is associated with higher corn yields, while hog density is associated with lower corn yields, though this could reflect the growing conditions where hog vs cattle production is favored. Cover crops are significantly associated with lower corn yields, though it should not be presumed that cover crops cause lower corn yields as they may be being grown more in areas that have more challenging growing conditions and thus lower corn yields to start with. Tillage practices did not have a significant effect on corn yield. The average state corn yield is a significant and positive predictor of county corn yield as is expected. Table 4.2. Direct effect coefficients and standard errors from multilevel structural equation model with random intercepts for 12 states, 2012 (n=1042). Coefficient (S.E.) Standardized Coefficient R2 Average corn yield (bu./acre)  Average farmland value ($1000/acre) Average years of farming experience Cattle density Conservation tillage ratio Cover crops ratio Full-time farmer ratio Hog density Irrigated acres ratio Legume acres ratio Normal annual growing degree days No-till ratio Poultry density Tile acres ratio 2.885 (1.395)* 0.629 (0.545) 70.701 (35.494)* 20.494 (15.496) -124.364 (49.378)* 29.096 (19.007) -3.607 (1.18)** 0.130 (0.062)* 0.035 (0.030) 0.112 (0.053)* 0.065 (0.048) -0.096 (0.038)* 0.073 (0.051) -0.033 (0.010)** 123.495 (24.67)*** 0.409 (0.068)*** -3.922 (16.759) -0.012 (0.009) -23.016 (14.962) 0.000 (0.071) -0.014 (0.061) -0.109 (0.080) -0.111 (0.075) 0.000 (0.020) 53.609 (15.053)*** 0.287 (0.075)*** 2012 growing season precipitation 0.506 (0.554) 0.053 (0.058) Residual variance (within) 531.025 (51.897)*** 0.561 (0.048)*** 175 0.439 Table 4.2. (cont’d). State-level effects Average state corn yield (bu./acre) 0.989 (0.040)*** Intercept Residual variance (between) Average fertilizer use ($/acres)  Average corn yield (bu./acre) Average farmland value ($1000/acre) 104.398 (1.651)*** 26.437 (11.975)* 0.385 (0.078)*** 13.235 (7.347)t Average years of farming experience 1.474 (1.154) 0.976 (0.014)*** 4.396 (0.868)*** 0.047 (0.028)t 0.321 (0.072)*** 0.498 (0.194)* 0.068 (0.044) -0.129 (0.057)* -0.074 (0.048) -0.006 (0.148) Cattle density Conservation tillage ratio Corn acres ratio Cover crops ratio Crop insurance acres ratio Full-time farmer ratio Hog density Legume acres ratio Median farm size Normal annual growing degree days Normal growing season precipitation No-till ratio Poultry density Rented land ratio -97.07 (49.332)* -27.929 (18.517) -1.828 (43.490) 219.191 (106.027)* 0.142 (0.066)* -0.108 (13.874) -21.743 (29.799) -5.396 (4.009) -53.518 (20.512)** 0.009 (0.006) 0.022 (0.009)* 1.567 (0.966) -26.011 (14.376)t 0.107 (0.086) 0.000 (0.064) -0.046 (0.059) -0.042 (0.026) -0.163 (0.068) 0.063 (0.044) 0.168 (0.071)* 0.099 (0.075) -0.105 (0.067) 0.025 (0.023) -94.080 (43.533)* -0.276 (0.091)** 0.953 0.342 0.926 Residual variance (within) 895.957 (294.11)** 0.658 (0.041)*** State-level effects Average state fertilizer use ($/acre) 0.956 (0.066)*** 1.007 (0.057)*** Average state corn yield (bu./acre) -0.194 (0.080)* -0.185 (0.093)* Intercept Residual variance (between) 119.765 (2.225)*** 44.724 (24.553)t 4.868 (0.956)*** 0.074 (0.052) Chi-square Chi-square p-value Root mean square error (RMSEA) Comparative fit index (CFI) Tucker-Lewis index (TLI) 13.567 0.139 0.022 0.980 0.901 t p < 0.1, * p < 0.05 , ** p < 0.01 , *** p < 0.001 Note: Estimator used is maximum likelihood estimation with robust standard errors in Mplus8. County-level exogenous variables are group-mean centered; state-level variables are grand-mean centered. Because the model includes a mediator, the interpretation of the fertilizer use coefficients (the direct effects) only provides part of the story, so I discuss these results briefly before discussing the total effects on fertilizer use. Average county-level fertilizer use is significantly 176 predicted by average county corn yield, cattle density, cover crop ratio, legume ratio, normal annual growing degree days, and the rented land ratio. Average corn yield, cover crops, and normal annual growing degree days are all positively and significantly associated with county fertilizer use, while cattle density, legume ratio, and the rented land ratio are all negatively and significantly associated with county fertilizer use. Average farmland value and the ratio of no-till acres are marginally significant with average farmland value having a positive effect and no-till having a negative effect. County-level fertilizer use is also positively predicted by the average state fertilizer use. The effect of state corn yield on county-level fertilizer use is included in the model to partition the effect of corn yield on fertilizer use between the county and the state levels, since the county corn yield variable cannot be centered like the exogenous variables in the model, and it is the difference between the county-level and state-level effects of corn yield on county fertilizer use (Bell and Jones 2015). The effect of the state corn yield on county fertilizer use can be calculated in this case by adding the county-level corn yield effect (0.385) and the state corn yield coefficient (-0.194) resulting in the actual effect of state corn yield on county fertilizer use being 0.191. Because it is a two-level model, variation in the outcome variables exists at both levels. The model explains 43.9 percent of the observed county-level variation in county-level corn yields. 37.5 percent of the variation in the corn yield data exists at the state-level38 and the R2 value for the state-level effects describes the portion of that variation that is explained by the state-level variables. In this case 95.3 percent of the 37.5 percent of variation in county corn yield is explained by the average state corn yield. The model explains 34.2 percent of the observed county-level variation in county-level fertilizer use. The ICC for fertilizer use is 0.326, 38 This measure is the intraclass correlation coefficient (ICC). 177 so the state-level predictors explain 92.6 percent of the 32.6 percent of variation in county fertilizer that is at the state level. Over-all model fit is very good with a non-significant chi- square, an RMSEA that is smaller than 0.05, and a CFI that is larger than 0.95, while the TLI is larger than 0.9 (West, Taylor, and Wu 2012). Given the mediating position of corn yield in the model, the direct effects on county fertilizer use given in Table 4.2 are of limited use, so the standardized total effects that combine the direct effect on fertilizer use with the indirect effect on fertilizer use that works through corn yield are presented in Table 4.3. The results show that use of conservation tillage has no significant total effect on fertilizer use, while cover crops and no-till have only a marginally significant total effect on fertilizer use. For cover crops the positive direct effect on fertilizer use is off-set by the negative effect on corn yield. The negative direction of the total effect of no-till on fertilizer use is the direction that the agronomic findings suggest should be there though and thus weakly supports the assumption that no-till can save farmers money, not only through saved time and fuel from less tillage but also through fertilizer use. However, the positive effect direction of cover crops on fertilizer use, does not lend any support to the assumption that cover crops can save farmers money on fertilizer. Among the control variables, average corn yield, average farmland value, the irrigated acres ratio, and the tile acres ratio all have significant and positive total effects on fertilizer use. The effects of corn yield and farmland value are consistent with expectations. The effects of irrigation and tile were less expected as they are coming entirely through their effect on corn yield, as their directs effects on fertilizer use were not included in the model. The total effects of farming experience, and normal annual corn growing degree days are positive but marginally significant. 178 Table 4.3. Standardized, county-level, total effects on fertilizer use per acre. Average corn yield (bu./acre) Average farmland value ($1000/acre) Average years of farming experience Cattle density Conservation tillage ratio Corn acres ratio Cover crops ratio Crop insurance acres ratio Full-time farmer ratio Hog density Irrigated acres ratio Legume acres ratio Median farm size Normal annual corn growing degree days Normal growing season precipitation No-till ratio Poultry density Rented land ratio Tile acres ratio 2012 growing season precipitation Coefficient (S.E.) 0.321 (0.072)*** 0.539 (0.190)** 0.079 (0.042)t -0.092 (0.050)t -0.053 (0.051) -0.006 (0.148) 0.111 (0.063)t 0.000 (0.064) -0.022 (0.062) -0.052 (0.024)* 0.131 (0.031)*** -0.168 (0.063)** 0.063 (0.044) 0.133 (0.080)t 0.099 (0.075) -0.141 (0.075)t 0.025 (0.028) -0.276 (0.091)** 0.092 (0.034)** 0.017 (0.019) t p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 Note: Estimator used is maximum likelihood estimation with robust standard errors in Mplus8. All variables listed are effectively group-mean centered. Hog density, legume ratio and rented land all have negative total effects on fertilizer use. That the legume ration decreases fertilizer spending is consistent with the agronomic literature on crop rotations and leguminous properties. Interestingly the total effect of hog density is negative and significant, while the negative effect of cattle density on fertilizer use is balanced by its positive effect on corn yield which in turn is related to increased fertilizer spending, resulting in cattle density having a non-significant total effect. A similar situation occurs for cover crops, where the positive direct effect on fertilizer use is off-set by the negative effect on corn yield to result in a marginally significant total effect on fertilizer use. Higher proportions of rented land have a negative and significant total effect on fertilizer use, which is generally 179 consistent with Prokopy et al. (2008), but inconsistent with Baumgart-Getz et al. (2012), though these studies are based on general BMP use, in which fertilizer rate is not typically included. Based on unpublished interviews I expect that this negative effect is the result of farmers not wanting to “build up” the land they rent through higher fertilizer applications. The lack of significant total effect of conservation tillage, no-till and cover crops, is not as expected, but important lessons can still be learned from this analysis. The big one is the potential, indicated by the significant direct effects and the marginally significant total effects, for cover crops to increase fertilizer use. Dunn et al. (2016) reported that farmers in their national 2013 survey thought that costs of cover crops and their reduction in yield in the following cash crop were “somewhat limits” to the adoption of cover crops in their area. It is possible that in order to try to avoid reductions in the following crop that farmers are tending to apply more fertilizer than they otherwise would. Even a slight inclination by farmers to apply more fertilizer when using cover crops would have large implications for not only the actual but perceived costs of using cover crops and with that the incentives needed to effectively promote their use for their other benefits. CONCLUSION The primary goal of this analysis was to see if greater use of three BMPs was related to a decrease in the use of fertilizer at the county-level. However, I found no significant reduction in fertilizer use as a result of more prevalent use of either of the three practices. As previously mentioned, because this is an analysis of county-level averages care should be taken not to extend these findings to the individual-level (an ecological fallacy). As such, while the result is a bit disappointing from a sustainable agriculture standpoint, it does not rule-out the practices 180 having a fertilizer reducing effect at the individual farm-level. Care should also be taken not to over interpret the implication of causal effects in the model. For instance, while conservation practices precede both yield and fertilizer use in this model, this does not mean that the use of these conservation practices causes a reduction in yield, since we do not know what yield was prior to the use of the practices; it may be that farmers are more likely to use cover crops and no- till on lower yielding fields where they are trying to maximize yield by applying relatively more fertilizer. Figure 4.3. Yield response graph (Vanotti and Bundy 1994:184). The size of the effect of corn yield on fertilizer use is relatively small, which may be due to fertilizer and yield having a non-linear relationship, combined with average county fertilizer use falling near the optimum amount by virtue of being an average. At the arrows in the yield response curve in Figure 4.3, the slope of the yield to input graph is relatively shallow, and a 181 0.385 unit relationship seems very reasonable. In addition, the fertilizer use variable is not corn specific, thus diffusing some of the potential effect. Water is clearly a key piece of yield and fertilizer use, with normal growing season precipitation, irrigation and tile drainage all having important effects on fertilizer use. Irrigation and tile both have substantial positive effects on corn yield that carry through to significant total effects on fertilizer. Tile lines are a major way that both N and P leave fields, so it is notable that tile is associated with greater fertilizer spending and suggests that there might be a positive feedback between fertilizer use and tile drainage. The 2012 growing season precipitation is an important control for the effects of the 2012 drought, but the effects appear to be spotty given the lack of significant effect on county average corn yield in 2012 even controlling for irrigation. Future work along the lines of this analysis would be benefited by more precise data— such as fertilizer use on corn rather than general fertilizer use—and additional waves of data at the county level. The release of the 2017 AgCensus data in early 2019 will make it possible to consider temporal variation in addition to the spatial variation considered in this analysis. Using farmer survey data (either from a private survey or the USDA Agricultural Resource Management Survey [ARMS]) to consider the relationships from this model at the farmer-level would also be an avenue for future work, as would scaling up further to the watershed scale, where measures of water quality could be included as an outcome of interest to see how conservation practices and fertilizer use affect water quality singularly and together and how water quality might feedback to influence the use of conservation practices. The county-level used in this analysis could also be used to model the drivers of the three conservation practices. I attempted to incorporate the three conservation practices as mediators in the current model of fertilizer use but given the number of variables available there were not 182 enough degrees of freedom to estimate that large a number of parameters. The contexts that relate to the wider spread use of the practices could be estimated separately or as a seemingly unrelated regression model with the three practice-use ratio variables as correlated outcomes. The effect of corn yield on fertilizer use raises some questions of order and causality. The relationship between corn yield and fertilizer is potentially complicated by having a reciprocal effect, since logically yield is a function of fertilizer use. In this model yield is being used as a proxy for expected yield, and thus as a predictor of fertilizer use. However, a non-recursive model with reciprocal effects between corn yield and fertilizer use was considered to see if the effect in the model was being suppressed by a negative effect going in the opposite direction. The effect from fertilizer use to corn yield was negative but not significant, suggesting that the relationship between these variables used in the model is appropriate. In sum, the use of conservation tillage, no-till and cover crops does not appear to have an effect on fertilizer use at the county-level, while corn yield, land value, hog density, irrigation crop rotation, (i.e. use of legumes), rented land, and tile drainage all have a significant effect on fertilizer use. Farmers’ use and management of nutrients like nitrogen have very real environmental impacts (hypoxia, climate change) and very real money is spent trying to minimize these impacts. Not only does this analysis scale up from the individual to better assess the cumulative effects of many individual actions at a scale where the effects are more visible, but it also includes a mixture of ecological and economic and non-economic social factors. The combining of all of these types of variables in one analysis is novel in sociological studies, but is in keeping with calls for more integrated research on social-ecological systems and soil and water conservation behaviors (Liu, Dietz, Carpenter, Alberti, et al. 2007; Reimer et al. 2014; Stuart et al. 2015) as well as a few recent studies that have integrated social and ecological 183 elements in one study (Blesh and Wolf 2014; Halbrendt et al. 2014). This analysis provides an important step toward both better integration of social and biophysical systems, from a sociological stand-point, and also better understanding how conservation practices can play a role in reducing the environmental impact of US agriculture. 184 APPENDICES 185 Replacing Missing Values APPENDIX A: Variable Calculations For exogenous variables for which NASS had withheld values I replaced them using the agriculture district39 average. The process involved calculating agriculture district averages from the county data, calculating the ratio variables I use with the agriculture district averages, and then using the agriculture district ratios to replace the missing values in the county-level ratio variables. It is important that I use the ratio values in the replacement processes, rather than the counts, to avoid scale miss-matches resulting from counties of different sizes. I first calculated the agriculture district averages from the county data: mean(agdist farmland_12a corngrain_12b corngrain_12a cropland_12a proc sort data=yy; by agdist; run; proc summary data=yy; *means by ag district to replace missing values; var agdist farmland_12a corngrain_12b corngrain_12a cropland_12a alfalfa_12a soy_12a cropins_12a covcrop_12a constill_12a rentpartown_12a tenant_12a irrig_12a notill_12a hogs_12c catfeed_12c dairy_12c broilers_12c layers_12c turkeys_12c fertcrop_12a fert_12d tile_12a income_12df; by agdist; output out=adm_vars alfalfa_12a soy_12a cropins_12a covcrop_12a constill_12a rentpartown_12a tenant_12a irrig_12a notill_12a hogs_12c catfeed_12c dairy_12c broilers_12c layers_12c turkeys_12c fertcrop_12a fert_12d tile_12a income_12df) adm_alfalfa adm_soy adm_cropins adm_covcrop adm_constill adm_rentpartown adm_tenant adm_irrig adm_notill adm_hogs adm_cat adm_dairy adm_broil adm_lay adm_turk adm_fertcrop adm_fert adm_tile adm_income; run; =agdist adm_farmland adm_corn_b adm_corn_a adm_cropland Then I used those averages to calculate the agriculture district ratios for the variables I needed to replace: *calculate agdist averages for ratios that need to be replaced; data adm_r; set add; adm_alfalfa_r=adm_alfalfa/adm_cropland; adm_constill_r=adm_constill/adm_cropland; adm_corn_r=adm_corn_a/adm_cropland; adm_covcrop_r=adm_covcrop/adm_cropland; adm_cropins_r=adm_cropins/adm_cropland; adm_soy_r=adm_soy/adm_cropland; adm_rentpartown_r=adm_rentpartown/adm_farmland; adm_tenant_r=adm_tenant/adm_farmland; adm_irrig_r=adm_irrig/adm_cropland; 39 Agriculture districts are sub-state designations made by USDA NASS. There are 9 agriculture districts in each state in the sample (https://www.nass.usda.gov/Charts_and_Maps/Crops_County/boundary_maps/indexgif.php). 186 adm_notill_r=adm_notill/adm_cropland; adm_hogs_r=adm_hogs/adm_cropland; adm_dairy_r=adm_dairy/adm_cropland; adm_cat_r=adm_cat/adm_cropland; adm_broil_r=adm_broil/adm_cropland; adm_lay_r=adm_lay/adm_cropland; adm_turk_r=adm_turk/adm_cropland; adm_tile_r=adm_tile/adm_cropland; run; One district was missing dairy cattle data for all the counties in the district, so I calculated the average for the district from the surrounding counties. All other agriculture districts had data from at least two counties for each variable. proc means data=ratios; *calculating an alternate dairy ratio average for agdist 3820 which is missing observation on that variable for all counties. These 9 counties are adjacent to agdist 3820. Calculated mean for the 7 of these that are non-missing is 0.00009523; var dairy_12r; where fips in (38075 38101 38055 38083 38103 38027 38063 38071 38095) and dairy_12r~=.; run; data fill1; set ratios; *dropping a couple variables left over from calculating agdist means and replacing missing agdist with alternatively calculated value; drop _TYPE_ _FREQ_; if agdist=3820 then adm_dairy_r=0.00009523; run; I calculated the ratios for the county-level variables: *calculate county level ratio variables; data ratios; set adm_r; constill_12r=constill_12a/cropland_12a; corngrain_12ba=corngrain_12b/corngrain_12a; corngrain_12r=corngrain_12a/cropland_12a; covcrop_12r=covcrop_12a/cropland_12a; cropins_12r=cropins_12a/cropland_12a; cropland_12r=cropland_12a/county_12a; csp_12df=csp_12d/csp_12f; csp_12da1=csp_12d/cropland_12a; csp_12da2=csp_12d/farmland_12a; eqip_12df=eqip_12d/eqip_12f; eqip_12da1=eqip_12d/cropland_12a; eqip_12da2=eqip_12d/farmland_12a; farmland_12r=farmland_12a/county_12a; fert_12da=fert_12d/fertcrop_12a; fertcrop_12af=fertcrop_12a/fertcrop_12f; fertcrop_12r=fertcrop_12a/cropland_12a; fullowner_12r=fullowner_12a/farmland_12a; fulltime_12r=fulltime_12f/farms_12c; irrig_12r=irrig_12a/cropland_12a; 187 notill_12r=notill_12a/cropland_12a; rentpartown_12r=rentpartown_12a/farmland_12a; tenant_12r=tenant_12a/farmland_12a; soy_12r=soy_12a/cropland_12a; alfalfa_12r=alfalfa_12a/cropland_12a; tile_12r=tile_12a/cropland_12a; broilers_12r=broilers_12c/cropland_12a; layers_12r=layers_12c/cropland_12a; turkeys_12r=turkeys_12c/cropland_12a; cat_12r=catfeed_12c/cropland_12a; dairy_12r=dairy_12c/cropland_12a; hogs_12r=hogs_12c/cropland_12a; run; Then I replaced the missing county data in the ratio variables with the agriculture district average ratios: *replacing missing values with agdist average; data fill2; set fill1; corngrain_r=input(corngrain_12r,best32.); if corngrain_12r=. then corngrain_r=adm_corn_r; soy_r=input(soy_12r,best32.); if soy_12r=. then soy_r=adm_soy_r; alfalfa_r=input(alfalfa_12r,best32.); if alfalfa_12r=. then alfalfa_r=adm_alfalfa_r; cropins_r=input(cropins_12r,best32.); if cropins_12r=. then cropins_r=adm_cropins_r; covcrop_r=input(covcrop_12r,best32.); if covcrop_12r=. then covcrop_r=adm_covcrop_r; constill_r=input(constill_12r,best32.); if constill_12r=. then constill_r=adm_constill_r; rentpartown_r=input(rentpartown_12r,best32.); if rentpartown_12r=. then rentpartown_r=adm_rentpartown_r; tenant_r=input(tenant_12r,best32.); if tenant_12r=. then tenant_r=adm_tenant_r; irrig_r=input(irrig_12r,best32.); if irrig_12r=. then irrig_r=adm_irrig_r; notill_r=input(notill_12r,best32.); if notill_12r=. then notill_r=adm_notill_r; hogs_r=input(hogs_12r,best32.); if hogs_12r=. then hogs_r=adm_hogs_r; catfeed_r=input(catfeed_12r,best32.); if catfeed_12r=. then catfeed_r=adm_cat_r; dairy_r=input(dairy_12r,best32.); if dairy_12r=. then dairy_r=adm_dairy_r; broilers_r=input(broilers_12r,best32.); if broilers_12r=. then broilers_r=adm_broil_r; layers_r=input(layers_12r,best32.); if layers_12r=. then layers_r=adm_lay_r; turkeys_r=input(turkeys_12r,best32.); if turkeys_12r=. then turkeys_r=adm_turk_r; tile_r=input(tile_12r,best32.); 188 if tile_12r=. then tile_r=adm_tile_r; income_df=input(income_12df,best32.); if income_12df=.d then income_df=adm_income; run; I created the combined variables in a separate data step: poultry_r=broilers_r+layers_r+turkeys_r; rented_r=rentpartown_r+tenant_r; cattle_r=catfeed_r+dairy_r; legume_r=soy_r+alfalfa_r; Calculation of Normal Precipitation and Corn Growing Degree Days 30-year normals (e.g., averages) were calculated from daily from daily temperature and precipitation data from 1982-2011, in accordance with NOAA description at https://www.ncdc.noaa.gov/data-access/land-based-station-data/land-based-datasets/climate- normals/1981-2010-normals-data. For precipitation this was done by calculating the total precipitation for each of the 360 months, and then taking the average for each month of the year (e.g., the average of the 30 June totals) to get the normal precipitation for each month of the year. These normal totals for April through September were then added together to get the normal growing season precipitation. For corn growing degree days (GDD) the same process as above was followed except that prior to finding the total for each of the 360 months the corn GDD were calculated for each day according to the following formula and rules (https://ndawn.ndsu.nodak.edu/help-corn-growing- degree-days.html): Daily Corn GDD (°F) = [(Daily Max Temp °F + Daily Min Temp °F) / 2] - 50 °F If the daily max and/or min temperature was < 50 °F it was set equal to 50 °F, and if the daily max temperature > 86 °F, it was set equal to 86 °F. 189 APPENDIX B: Centering Analysis Variables There are two ways to center a variable: group mean centering and grand mean centering. Group-mean centering (also called centering within clusters) is what its name suggests— centering the variable around the mean for each group. Grand-mean centering is centering the variable around the mean for all the observations of that type—for instance the average farm size across all the counties sampled. Level-2 variables (e.g., state-level variables) can be grand mean centered or left in their raw form. The decision to center level-2 variables depends on the nature of the variable and is for ease of interpretation as it is in OLS regressions (Enders and Tofighi 2007). In this case, grand- mean centering the state-level variables is helpful because while having zero fertilizer use or zero corn grain yield can conceptually exist it does not in practice, thus the average is more interesting. Centering level-1 (e.g., county-level) variables has more important model implications than centering at level-2. Group-mean centering level-1 variables serves to separate the within- group effects from the between group effects by removing the between group variation from the data, and thus removing correlation between the level-1 variables and the level-2 variables (Enders and Tofighi 2007). This results in level-1 coefficients that only include level-1 effects, thus making their interpretation more precise when the level-1 effects are of primary interest, as they are in this analysis (Enders and Tofighi 2007). The removal of between-level correlation is critical in random-slope models, where this lack of correlation is an important assumption (Enders and Tofighi 2007). Group-mean centering level-1 variables means that the model results are interpreted as follows (Enders and Tofighi 2007): • Intercept is the average unadjusted cluster mean—ie the average intercept of all the clusters 190 • Slope is pooled within-cluster level-1 regression coefficient (ie the average slope of all the group slopes)—in this case since model has fixed slope it is the same for all the groups Intercept variance is the variance of the unadjusted cluster means • 191 APPENDIX C: Extra Sample Descriptives Table 4.4. Counties excluded from analysis due to low amount of agriculture or withheld data. County Notes Corn bushels Corn acres Fertilizer dollars State FIPS Fertilized acres Cook DuPage Keweenaw Iron Luce Ontonagon Schoolcraft Cook Lake IL IL MI MI MI MI MI MN MN 17031 17043 26083 26071 26095 26131 26153 27031 27075 St. Louis City MO 29510 Grant Hooker NE NE 31075 31091 Menominee WI 55078 375456 WH 0 0 0 WH WH 0 0 0 0 0 0 WH = Withheld (by NASS) WH WH 0 0 0 31 WH 0 0 0 0 0 0 765000 499000 0 473000 WH 72000 91000 1000 9000 0 WH WH 0 WH WH 0 WH 1023 WH WH WH WH 0 WH 1560 0 County contains majority of metro Chicago County contain west Chicago suburbs County in the Upper Peninsula County in the Upper Peninsula County in the Upper Peninsula County in the Upper Peninsula County in the Upper Peninsula Most of county is in Superior National Forest Most of county is in Superior National Forest or Finland State Forest County is metro St. Louis County essentially coterminous with Menominee Indian Reservation 192 Outcome Variables Average fertilizer use ($/acre) Average corn yield (bu./acre) Farmer and Farm Characteristics Rented land ratio Av. farmland value ($1000/acre) Tile acres ratio Full-time farmer ratio Average years of farming experience Median farm size Irrigated acres ratio Crop insurance acres ratio Crops Corn acres ratio Legume ratio Poultry density Cattle density Hog density BMPs Conservation tillage ratio No-till ratio Cover crops ratio Climate Normal growing season precipitation 2012 growing season precipitation Normal annual growing degree days (in 1000s) 3 Note: The data presented here (except for the climate variables) are the state-level values from the AgCensus, not an average of the county-level data used in the analysis. 3 4 4 3 4 3 4 3 3 2 193 Table 4.5. State-level averages for analysis variables. IL IN IA KS MI MN MO NE ND OH SD WI 134 102 150 99 135 134 74 85 147 131 133 156 124 69 113 131 72 117 141 120 90 91 128 120 0.6 6 0.4 0.5 27 100 0.02 0.74 0.52 0.39 0.19 0.01 0.17 0.32 0.25 0.01 0.5 5 0.5 0.4 26 57 0.03 0.67 0.48 0.43 2.51 0.02 0.25 0.24 0.39 0.05 0.5 6 0.5 0.5 28 136 0.01 0.80 0.52 0.38 1.91 0.06 0.67 0.33 0.26 0.01 0.5 2 0.0 0.5 27 200 0.10 0.63 0.14 0.15 0.01 0.05 0.04 0.21 0.37 0.01 0.4 4 0.3 0.5 26 60 0.08 0.53 0.31 0.34 1.61 0.05 0.11 0.23 0.20 0.06 0.4 4 0.3 0.5 28 142 0.02 0.75 0.39 0.36 1.42 0.04 0.29 0.28 0.04 0.02 0.3 3 0.1 0.4 26 120 0.08 0.51 0.22 0.36 2.22 0.01 0.10 0.18 0.26 0.03 0.4 2 0.0 0.6 27 280 0.38 0.72 0.42 0.27 0.23 0.06 0.07 0.23 0.43 0.02 0.5 1 0.0 0.6 28 480 0.01 0.76 0.13 0.22 0.01 0.00 0.00 0.23 0.29 0.01 0.4 5 0.5 0.4 26 69 0.00 0.60 0.34 0.46 3.05 0.03 0.15 0.22 0.40 0.03 0.4 2 0.0 0.6 27 337 0.02 0.82 0.28 0.32 0.11 0.01 0.03 0.18 0.37 0.01 23 17 24 18 25 16 21 14 20 17 21 19 26 18 19 11 14 11 23 19 17 13 0.3 4 0.1 0.5 26 96 0.04 0.50 0.33 0.28 1.15 0.11 0.02 0.26 0.18 0.06 22 18 3 REFERENCES 194 REFERENCES Ajzen, Icek. 1991. “The Theory of Planned Behavior.” Orgnizational Behavior and Human Decision Processes 50:179–211. Allmaras, R. R., P. W. Unger, and D. W. Wilkins. 1985. “Conservation Tillage Systems and Soil Productivity.” Pp. 357–412 in Soil Erosion and Crop Productivity, edited by R. F. Follett and B. A. Stewart. Madison, WI: American Society of Agronomy, Crop Science Society of America, Soil Science Society of America. An, Li and David López-Carr. 2012. “Understanding Human Decisions in Coupled Natural and Human Systems.” Ecological Modelling 229:1–4. Andrews, A. C., R. A. Clawson, B. M. Gramig, and L. Raymond. 2013. “Why Do Farmers Adopt Conservation Tillage? An Experimental Investigation of Framing Effects.” Journal of Soil and Water Conservation 68(6):501–11. Aneja, Viney P., William H. Schlesinger, and Jan Willem Erisman. 2009. “Effects of Agriculture upon the Air Quality and Climate: Research, Policy, and Regulations.” Environmental Science & Technology 43(12):4234. Arbuckle, J. G. and G. Roesch-McNally. 2015. “Cover Crop Adoption in Iowa: The Role of Perceived Practice Characteristics.” Journal of Soil and Water Conservation 70(6):418–29. Arbuckle, J. Gordon, Linda Stalker Prokopy, Tonya Haigh, Jon Hobbs, Tricia Knoot, Cody Knutson, Adam Loy, Amber Saylor Mase, Jean McGuire, Lois Wright Morton, John Tyndall, and Melissa Widhalm. 2013. “Climate Change Beliefs, Concerns, and Attitudes toward Adaptation and Mitigation among Farmers in the Midwestern United States.” Climatic Change 117(4):943–50. Baumgart-Getz, Adam, Linda Stalker Prokopy, and Kristin Floress. 2012. “Why Farmers Adopt Best Management Practice in the United States: A Meta-Analysis of the Adoption Literature.” Journal of Environmental Management 96(1):17–25. Bell, Andrew and Kelvyn Jones. 2015. “Explaining Fixed Effects: Random Effects Modeling of Time-Series Cross-Sectional and Panel Data.” Political Science Research and Methods 3(1):133–53. Blesh, J. and L. E. Drinkwater. 2013. “The Impact of Nitrogen Source and Crop Rotation on Nitrogen Mass Balances in the Mississippi River Basin.” Ecological Applications 23(5):1017–35. Blesh, Jennifer and Steven A. Wolf. 2014. “Transitions to Agroecological Farming Systems in the Mississippi River Basin: Toward an Integrated Socioecological Analysis.” Agriculture and Human Values 31(4):621–35. 195 Bollen, Kenneth A. and Judea Pearl. 2013. “Eight Myths about Causality and Structural Equation Models.” Pp. 301–28 in Handbook of Causal Analysis for Social Research, edited by S. L. Morgan. New York: Springer. Bruulsema, Tom, Jerry Lemunyon, and Bill Herz. 2009. “Know Your Fertilizer Rights.” Crops & Soils (March-April):13–18. Busari, Mutiu Abolanle, Surinder Singh Kukal, Amanpreet Kaur, Rajan Bhatt, and Ashura Ally Dulazi. 2015. “Conservation Tillage Impacts on Soil, Crop and the Environment.” International Soil and Water Conservation Research 3(2):119–29. Carpenter, S. R., N. F. Caraco, D. L. Correll, R. W. Howarth, A. N. Sharpley, and V. H. Smith. 1998. “Nonpoint Pollution of Surface Waters with Phosphorus and Nitrogen.” Ecological Applications 8(3):559–68. Coulter, Jeffrey A. Craig C. Sheaffer, Donald L. Wyse, Milton J. Haar, Paul M. Porter, Steven R. Quiring, and Lee D. Klossner. 2011. “Agronomic Performance of Cropping Systems with Contrasting Crop Rotations and External Inputs.” Agronomy Journal 103(1):182–92. CTIC. 2002. “Tillage Type Definitions.” Conservation Technology Information Center. Retrieved April 20, 2018 (http://www.ctic.purdue.edu/resourcedisplay/322/). Dabney, S. M., J. A. Delgado, and D. W. Reeves. 2001. “Using Winter Cover Crops to Improve Soil and Water Quality.” Communications in Soil Science and Plant Analysis 32(7– 8):1221–50. Davis, D. M., P. H. Gowda, D. J. Mulla, and G. W. Randall. 2000. “Modeling Nitrate Nitrogen Leaching in Response to Nitrogen Fertilizer Rate and Tile Drain Depth or Spacing for Southern Minnesota, USA.” Journal of Environment Quality 29(5):1568–81. Doane, Timothy A. William R. Horwath, Jeffrey P. Mitchell, Jim Jackson, Gene Miyao, and Kent Brittan. 2009. “Nitrogen Supply from Fertilizer and Legume Cover Crop in the Transition to No-Tillage for Irrigated Row Crops.” Nutrient Cycling in Agroecosystems 85(3):253–62. Dolan, M. S., C. E. Clapp, R. R. Allmaras, J. M. Baker, and J. A. E. Molina. 2006. “Soil Organic Carbon and Nitrogen in a Minnesota Soil as Related to Tillage, Residue and Nitrogen Management.” Soil and Tillage Research 89(2):221–31. Drinkwater, L. E. and S. S. Snapp. 2007. “Nutrients in Agroecosystems: Rethinking the Management Paradigm.” Advances in Agronomy 92:163–86. Du, Zhangliu, Tusheng Ren, and Chunsheng Hu. 2010. “Tillage and Residue Removal Effects on Soil Carbon and Nitrogen Storage in the North China Plain.” Soil Science Society of America Journal 74(1):196. Duff, Scott N., D. Pete. Stonehouse, Don J. Blackburn, and Stewart G. Hilts. 1992. “A Framework for Targeting Soil Conservation Policy.” Journal of Rural Studies 8(4):399– 196 410. Dunn, M., J. D Ulrich-Schad, L. S. Prokopy, R. L. Myers, C. R. Watts, and K. Scanlon. 2016. “Perceptions and Use of Cover Crops among Early Adopters: Findings from a National Survey.” Journal of Soil and Water Conservation 71(1):29–40. Enders, Craig K. and Davood Tofighi. 2007. “Centering Predictor Variables in Cross-Sectional Multilevel Models: A New Look at an Old Issue.” Psychological Methods 12(2):121–38. Fan, Yi, Jiquan Chen, Gabriela Shirkey, Ranjeet John, Susie R. Wu, Hogeun Park, and Changliang Shao. 2016. “Applications of Structural Equation Modeling (SEM) in Ecological Studies: An Updated Review.” Ecological Processes 5(1). Fishbein, Martin and Icek Ajzen. 2015. Predicting and Changing Behavior: The Reasoned Action Approach. New York, NY: Routledge. Gardner, Jennifer B. and Laurie E. Drinkwater. 2009. “The Fate of Nitrogen in Grain Cropping Systems: A Meta-Analysis of 15N Field Experiments.” Ecological Applications 19(8):2167–84. Gentry, L. E., M. B. David, T. V. Royer, C. A. Mitchell, and K. M. Starks. 2007. “Phosphorus Transport Pathways to Streams in Tile-Drained Agricultural Watersheds.” Journal of Environment Quality 36(2):408–15. Grace, J. B., T. M. Anderson, Han Olff, and Samuel M. Scheiner. 2010. “On the Specification of Structural Equation Models for Ecological Systems.” Ecological Monographs 80(1):67–87. Grandy, A. Stuart, Terrance D. Loecke, Sara Parr, and G. Philip Robertson. 2006. “Long-Term Trends in Nitrous Oxide Emissions, Soil Nitrogen, and Crop Yields of Till and No-Till Cropping Systems.” Journal of Environment Quality 35(4):1487. Halbrendt, Jacqueline, Steven A. Gray, Susan Crow, Theodore Radovich, Aya H. Kimura, and Bir Bahadur Tamang. 2014. “Differences in Farmer and Expert Beliefs and the Perceived Impacts of Conservation Agriculture.” Global Environmental Change 28:50–62. Henke, Christopher R. 2006. “Changing Ecologies: Science and Environmental Politics in Agriculture.” Pp. 215–43 in The New Political Sociology of Science: Institutions, Networks, and Power, edited by S. Frickel and K. Moore. Madison, WI: The University of Wisconsin Press. Hoben, J. P., R. J. Gehl, N. Millar, P. R. Grace, and G. P. Robertson. 2011. “Nonlinear Nitrous Oxide (N2O) Response to Nitrogen Fertilizer in on-Farm Corn Crops of the US Midwest.” Global Change Biology 17(2):1140–52. Houser, Matthew, Diana Stuart, and Michael Carolan. 2017. “Is Seeing Believing? Applying a Realist Framework to Examine Agriculture and Climate Change.” Environmental Sociology 3(4):1–16. 197 Houser, Matthew, Diana Stuart, Riva C. H. Denny, and Sandra T. Marquart-Pyatt. n.d. “Proposing and Operationalizing Synthetic Nitrogen Fertilizer Application Rate as a Measure for Nitrogen Best Management Practice Adoption Research.” Jones, Kelvyn and Craig Duncan. 1998. “Modelling Context and Heterogeneity: Applying Multilevel Models.” Pp. 95–125 in Research strategies in the social sciences, edited by E. Scarbrough and E. Tanenbaum. Oxford, UK: Oxford University Press. Kara, Erdal, Marc Ribaudo, and Robert C. Johansson. 2008. “On How Environmental Stringency Influences Adoption of Best Management Practices in Agriculture.” Journal of Environmental Management 88(4):1530–37. Kleinman, Peter J. A., Andrew N. Sharpley, Richard W. McDowell, Don N. Flaten, Anthony R. Buda, Liang Tao, Lars Bergstrom, and Qing Zhu. 2011. “Managing Agricultural Phosphorus for Water Quality Protection: Principles for Progress.” Plant and Soil 349(1– 2):169–82. Lal, Rattan. 1993. “Tillage Effects on Soil Degradation, Soil Resilience, Soil Quality, and Sustainability.” Soil and Tillage Research 27(1–4):1–8. Liu, Jianguo, Thomas Dietz, Stephen R. Carpenter, Marina Alberti, Carl Folke, Emilio Moran, Alice N. Pell, Peter Deadman, Timothy Kratz, Jane Lubchenco, Elinor Ostrom, Zhiyun Ouyang, William Provencher, Charles L. Redman, Stephen H. Schneider, and William W. Taylor. 2007. “Complexity of Coupled Human and Natural Systems.” Science 317(5844):1513–16. Liu, Jianguo, Thomas Dietz, Stephen R. Carpenter, Carl Folke, Marina Alberti, Charles L. Redman, Stephen H. Schneider, Elinor Ostrom, Alice N. Pell, Jane Lubchenco, William W. Taylor, Zhiyun Ouyang, Peter Deadman, Timothy Kratz, and William Provencher. 2007. “Coupled Human and Natural Systems.” AMBIO 36(8):639–49. Mahama, George Y., P. V. Vara Prasad, Kraig L. Roozeboom, Jesse B. Nippert, and Charles W. Rice. 2016. “Response of Maize to Cover Crops, Fertilizer Nitrogen Rates, and Economic Return.” Agronomy Journal 108(1):17–31. Marquart-Pyatt, Sandra, Andrew K. Jorgenson, and Lawrence C. Hamilton. 2015. “Methodological Approaches for Sociological Research on Climate Change.” Pp. 369–411 in Climate Change and Society: Sociological Perspectives, edited by R. E. Dunlap and R. J. Brulle. New York: Oxford University Press. Mazzoncini, Marco, Tek Bahadur Sapkota, Paolo Bàrberi, Daniele Antichi, and Rosalba Risaliti. 2011. “Long-Term Effect of Tillage, Nitrogen Fertilization and Cover Crops on Soil Organic Carbon and Total Nitrogen Content.” Soil and Tillage Research 114(2):165–74. McSwiney, Claire P. and G. Philip Robertson. 2005. “Nonlinear Response of N2O Flux to Incremental Fertilizer Addition in a Continuous Maize (Zea Mays L.) Cropping System.” Global Change Biology 11(10):1712–19. 198 Moriasi, Daniel N., Prasanna H. Gowda, Jeffrey G. Arnold, David J. Mulla, Srinivasulu Ale, and Jean L. Steiner. 2013. “Modeling the Impact of Nitrogen Fertilizer Application and Tile Drain Configuration on Nitrate Leaching Using SWAT.” Agricultural Water Management 130:36–43. NASS. 2012a. Appendix A. Census of Agriculture Methodology. National Agricultural Statistics Service, U.S. Department of Agriculture. Retrieved February 14, 2018 (https://www.agcensus.usda.gov/Publications/2012/Full_Report/Volume_1,_Chapter_1_US /usappxa.pdf). NASS. 2012b. Appendix B. General Explanation and Census of Agriculture Report Form. National Agricultural Statistics Service, U.S. Department of Agriculture. Retrieved June 29, 2018 (https://www.agcensus.usda.gov/Publications/2012/Full_Report/Volume_1,_Chapter_1_US /usappxb.pdf). NASS. 2012c. “Instruction Sheet United States 2012 Census of Agriculture.” National Agricultural Statistics Service, U.S. Department of Agriculture. Retrieved June 13, 2018 (https://www.agcensus.usda.gov/Help/Report_Form_and_Instructions/2012_Report_Form/1 2-A01(i).pdf). NRCS. 2012. Conservation Practice Standard Overview: Residue and Tillage Management (329) No-Till/Strip Till/Direct Seed. Natural Resources Conservation Service, U.S. Department of Agriculture. Retrieved (https://www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/stelprdb1254958.pdf). Odum, Eugene P. 1984. “Properties of Agroecosystems.” Pp. 5–11 in Agricultural Ecosystems: Unifying Concepts, edited by R. Lowrance, B. R. Stinner, and G. J. House. New York: John Wiley. Osterholz, William R., Matt Liebman, and Michael J. Castellano. 2018. “Can Soil Nitrogen Dynamics Explain the Yield Benefit of Crop Diversification?” Field Crops Research 219:33–42. Plastina, Alejandro, Fangge Liu, Fernando Miguez, and Sarah Carlson. 2018. “Cover Crops Use in Midwestern US Agriculture: Perceived Benefits and Net Returns.” Renewable Agriculture and Food Systems online. Prokopy, L. S., K. Floress, D. Klotthor-Weinkauf, and A. Baumgart-Getz. 2008. “Determinants of Agricultural BMP Adoption: Evidence from the Literature.” Journal of Soil and Water Conservation 63(5):300–311. Reimer, Adam P. Aaron Thompson, Linda Stalker Prokopy, J. Gordon Arbuckle, Ken Genskow, Douglas Jackson-Smith, Gary Lynne, Laura McCann, Lois Wright Morton, and Pete Nowak. 2014. “People, Place, Behavior, and Context: A Research Agenda for Expanding Our Understanding of What Motivates Farmers’ Conservation Behaviors.” Journal of Soil and Water Conservation 69(2):57A–61A. 199 Reimer, Adam P. and Linda S. Prokopy. 2014. “One Federal Policy, Four Different Policy Contexts: An Examination of Agri-Environmental Policy Implementation in the Midwestern United States.” Land Use Policy 38:605–14. Reimer, Adam P., Aaron W. Thompson, and Linda S. Prokopy. 2012. “The Multi-Dimensional Nature of Environmental Attitudes among Farmers in Indiana: Implications for Conservation Adoption.” Agriculture and Human Values 29(1):29–40. Reimer, Adam P., Denise K. Weinkauf, and Linda S. Prokopy. 2012. “The Influence of Perceptions of Practice Characteristics: An Examination of Agricultural Best Management Practice Adoption in Two Indiana Watersheds.” Journal of Rural Studies 28(1):118–28. Ribaudo, Marc, Jorge Delgado, Leroy Hansen, Michael Livingston, Roberto Mosheim, and James Williamson. 2011. Nitrogen In Agricultural Systems: Implications For Conservation Policy. Economic Research Service, U.S. Department of Agriculture. Retrieved November 22, 2016 (http://www.chesapeake.org/stac/presentations/63_Ribaudo M et al_2011_Nitrogen in agricultural systems_implications for conservation policy.pdf). Ribaudo, Marc, Michael Livingston, and James Williamson. 2012. Nitrogen Management on U.S. Corn Acres, 2001-10. Economic Research Service, U.S. Department of Agriculture. Retrieved September 6, 2016 (http://162.79.45.209/media/947769/eb20.pdf). Roberts, Terry L. and A. Edward Johnston. 2015. “Phosphorus Use Efficiency and Management in Agriculture.” Resources, Conservation and Recycling 105:275–81. Robertson, G. Philip Tom W. Bruulsema, Ron J. Gehl, David Kanter, Denise L. Mauzerall, C. Alan Rotz, and Candiss O. Williams. 2013. “Nitrogen–climate Interactions in US Agriculture.” Biogeochemistry 114(1–3):41–70. Robertson, G. Philip Katherine L. Gross, Stephen K. Hamilton, Douglas A. Landis, Thomas M. Schmidt, Sieglinde S. Snapp, and Scott M. Swinton. 2014. “Farming for Ecosystem Services: An Ecological Approach to Production Agriculture.” BioScience 64(5):404–15. Robertson, G. Philip and P. M. Groffman. 2015. “Nitrogen Transformations.” Pp. 421–46 in Soil Microbiology, Ecology and Biochemistry, edited by E. A. Paul. Burlington, MA: Academic Press. Robertson, G. Philip and Richard R. Harwood. 2013. “Agriculture, Sustainable.” Pp. 111–18 in Encyclopedia of Biodiversity, vol. 1, edited by S. A. Levin. Waltham, MA: Academic Press. Robertson, G. Philip and Peter M. Vitousek. 2009. “Nitrogen in Agriculture: Balancing the Cost of an Essential Resource.” Annual Review of Environment and Resources 34(1):97–125. Rocha, João, Peter Roebeling, and María Ermitas Rial-Rivas. 2015. “Assessing the Impacts of Sustainable Agricultural Practices for Water Quality Improvements in the Vouga Catchment (Portugal) Using the SWAT Model.” Science of the Total Environment 536:48–58. Roesch-McNally, Gabrielle E., J. Gordon Arbuckle, and John Charles Tyndall. 2017. “What 200 Would Farmers Do? Adaptation Intentions under a Corn Belt Climate Change Scenario.” Agriculture and Human Values 34(2):333–46. Rogers, Everett M. 2003. Diffusion of Innovations. Fifth. New York, NY: Free Press. Schumacker, Randall E. and Richard G. Lomax. 2010. A Beginner’s Guide to Structural Equation Modeling. 3rd Editio. Mahwah, NJ: Erlbaum and Associates. Singer, J. W., S. M. Nusser, and C. J. Alf. 2007. “Are Cover Crops Being Used in the US Corn Belt?” Journal of Soil and Water Conservation 62(5):353–58. Smith, R. G., A. S. Davis, N. R. Jordan, L. W. Atwood, A. B. Daly, A. Stuart Grandy, M. C. Hunter, R. T. Koide, D. A. Mortensen, P. Ewing, D. Kane, M. Li, Y. Lou, S. S. Snapp, K. A. Spokas, and A. C. Yannarell. 2014. “Structural Equation Modeling Facilitates Transdisciplinary Research on Agriculture and Climate Change.” Crop Science 54(2):475– 83. Snyder, Clifford S. 2017. “Enhanced Nitrogen Fertiliser Technologies Support the ‘ 4R ’ Concept to Optimise Crop Production and Minimise Environmental Losses.” Soil Research 55(6):463–72. Stanger, Trenton F. and Joseph G. Lauer. 2008. “Corn Grain Yield Response to Crop Rotation and Nitrogen over 35 Years.” Agronomy Journal 100(3):643–50. Stonehouse, D. Peter. 1996. “A Targeted Policy Approach to Inducing Improved Rates of Conservation Compliance in Agriculture.” Canadian Journal of Agricultural Economics 44(2):105–19. Stuart, Diana, Bruno Basso, Sandy Marquart-Pyatt, Adam Reimer, G. Philip Robertson, and Jinhua Zhao. 2015. “The Need for a Coupled Human and Natural Systems Understanding of Agricultural Nitrogen Loss.” BioScience 65(6):571–78. Teshager, Awoke Dagnew, Philip W. Gassman, Silvia Secchi, and Justin T. Schoof. 2017. “Simulation of Targeted Pollutant-Mitigation-Strategies to Reduce Nitrate and Sediment Hotspots in Agricultural Watershed.” Science of The Total Environment 607–608:1188– 1200. Uri, Noel D. and James A. Lewis. 1998. “The Dynamics of Soil Erosion in US Agriculture.” Science of the Total Environment 218(1):45–58. Vandermeer, John H. 2011. The Ecology of Agroecosystems. Sudbury, MA: Jones and Bartlett Publishers. Vanotti, M. B. and L. G. Bundy. 1994. “An Alternative Rationale for Corn Nitrogen Fertilizer Recommendations.” Journal of Production Agriculture 7(2):243–49. Wade, Tara, Roger Claassen, and Steven Wallander. 2015. Conservation-Practice Adoption Rates Vary Widely by Crop and Region. Washington DC. Retrieved May 16, 2016 201 (http://www.ers.usda.gov/media/1979972/eib147.pdf). Wauters, Erwin and Erik Mathijs. 2014. “The Adoption of Farm Level Soil Conservation Practices in Developed Countries: A Meta-Analytic Review.” International Journal of Agricultural Resources, Governance and Ecology 10(1):78–102. Weber, Catharine and Laura McCann. 2015. “Adoption of Nitrogen-Efficient Technologies by U.S. Corn Farmers.” Journal of Environment Quality 44(2):391–401. West, Stephen G., Aaron B. Taylor, and Wei Wu. 2012. “Model Fit and Model Selection in Structural Equation Modeling.” Pp. 209–31 in Handbook of Structural Equation Modeling, edited by R. H. Hoyle. New York: Guilford Press. Yuan, Mingwei, Fabián G. Fernández, Cameron M. Pittelkow, Kristin D. Greer, and Daniel Schaefer. 2018. “Tillage and Fertilizer Management Effects on Phosphorus Runoff from Minimal Slope Fields.” Journal of Environment Quality 47:462–70. 202 CHAPTER 5: CONCLUSION In the preceding three analyses I examine how farmers till and the major factors that influence their tillage decisions, the range of predictors of the use of cover crops and four nutrient management BMPs, and the aggregated effects of tillage, and cover crops on fertilizer use. While these analyses have been written up as separate papers for purposes of the dissertation, there are several ways that the findings intersect and several overall conclusions that can be drawn. MEASURES AND RESEARCH METHODS One of the other general conclusions that can be drawn from my analyses is the importance of how BMP use is measured. This should involve not just the incorporation of temporal and spatial variability, which is particularly important for practices like tillage and cover crops, but how practices are defined and grouped. The first analysis clearly demonstrates, with the number of versions of no-till found, just how important it is to consider the temporal and spatial use of a practice as well as clearly define the parameters of the practice of interest. This is not to pretend that doing so is easy. Indeed, a practice like tillage presents a substantial challenge given the number of dimensions that need to be considered. However, it is clear that the current method of using three or four categories provides at best a cursory representation of the actual practices being used and as a result much inconsistency is likely to exist in the data of such summary categories. The second analysis further demonstrates the potential trouble with using practice categories. The lack of overlap in significant variables across the practices, even ones like variable rate nitrogen (N) and variable rate phosphorus (P) or potassium (K), indicates that the meaningful features that could define a class of practices may not be readily apparent. There 203 is certainly great utility in using categories to simplify analyses, but it would seem that researchers in the social sciences have not yet spent enough time empirically establishing those categories for agricultural BMPs. This conclusion is not to disparage previous or current research, but rather to push us all to do more and to do it better. There are reasons why BMP use data has been collected as it has, and there are still useful things to be learned from analyzing it. Indeed, as an example, I fail to follow my own advice in my second and third analyses by using binary practice use measures in my second analysis and the same tillage categories in the third analysis that I call out in the first. Mostly this use is because the data had already been collected, as is often the case. I have, however, done my best to reduce the worst of the effects of these measurement issues. For example, spatial variation in the four nutrient management practices I consider in the second analysis should be relatively low, thereby making the use of the binary measures less problematic than it is for cover crops. Using aggregated data in the third analysis potentially smooths out some of the variation in the overly general tillage categories used by the US AgCensus, making it reasonable to think that on average they indicate a relative position along the tillage continuum. KNOWLEDGE AND THE BIOPYISCAL ENVIRONMENT The inclusion of the biophysical environment was a secondary goal of all three analyses, and the importance of it is a key finding, especially in the first two analyses. Soil qualities and conditions are an important aspect of farmers’ tillage decisions and also an influence on the use of PSNTs, variable rate N and cover crops. The biophysical context is a notable omission from both the adoption-diffusion model as well as the TPB/RAA though it potentially influences the appropriateness and function of BMPs and therefore their use. However, the finding of different 204 tillage decisions in response to ostensibly the same environmental conditions in the first analysis demonstrates a challenge to incorporating biophysical variables into social analyses: Are the environmental conditions the farmers are responding to actually the same? If they are then why are farmers responding differently? Do the farmers have the same understandings of how different tillage approaches work or are their differing responses based on different knowledge and expectations of the practices? How important then is the perception vs. the “reality”? These questions raise the possibility of the existence of substantial differences in how farmers perceive both their biophysical conditions and the practices themselves. The environmental implications of only considering farmers’ perceptions can be considerable. If two farmers in the same farming conditions perceive their conditions differently so that one thinks it is important to use conservation tillage to reduce erosion and the other does not think it is needed because the land is pretty flat, then the two farmers are both acting consistently with their perceptions but the environmental outcomes will be quite different. The TPB/RAA is fully based on the importance of individual perceptions that may or may not be accurate (Fishbein and Ajzen 2015), especially through the behavioral beliefs and perceived behavioral control (PBC) components, which make it poorly suited for connecting the social and the ecological, since it is unknown what signals farmers actually received from their environmental contexts and thus how similar or dissimilar their perceptions are. With this approach the environmental context remains a social idea, which can limit our ability to connect social and environmental conditions, understand the causal paths that create environmental problems and conduct research that can contribute to finding solutions to those problems. In contrast, the adoption-diffusion model perhaps does not include the importance of perceptions enough, especially in the knowledge and persuasion steps of the innovation-decision 205 process and in the relative advantage and compatibility innovation attributes. The innovation attributes are conceptualized as inherent characteristics of the innovation, meaning there is no conceptual space for these attributes to differ between farmers. The lack of conceptual interaction between the knowledge and persuasion decision steps, and the practice characteristics in the adoption-diffusion model is an additional limitation to those discussed in the introduction chapter. Critical realism (Carolan 2005) offers a conceptual framework for integrating the “real” and perceived and for connecting the adoption-diffusion decision process to the innovation attributes. Knowledge was of particular interest at the outset of these analyses, though it is only explicitly examined in the second analysis. In this analysis it is found to be an important predictor of the use of nutrient maps, variable rate P or K and cover crops, in spite of being a non-practice-specific measure. Knowledge of how to use or implement a practice is a basic threshold necessary for practice use (at least successful practice use). However, as is discussed in Chapter 3, there are conceptualized to be multiple types of knowledge, of which technical knowledge of the how-to variety is but one. Of particular interest is what Rogers’ principles- knowledge and Kaiser and Fuhrer’s declarative knowledge—knowledge of how the environmental system and the practice work. In the context of the present analyses, this might include knowledge of the nitrogen cycle, soil physics and chemistry, and crop physiology, and is hypothesized to be extremely important in providing farmers with an understanding of how and why a given practice works. Not only would this knowledge be very helpful in successfully implementing the practice in the farmer’s specific conditions, but it would also help in adapting to changing conditions. 206 Conceptually this system-type of knowledge goes beyond both education and information. Though education and information can contribute to the acquisition of system knowledge they are not equivalent to it. The potential importance of system knowledge in the agricultural context is that it includes aspects that cannot be observed directly, such as nitrogen in its elemental form, and thus connects the biophysical world with the social world. The less a process can be observed directly the more room exists for multiple social meanings and explanations to be attached to it, and thus for multiple actions seem appropriate (Carolan 2004, 2006). This type of relationship between the biophysical and social through knowledge can be understood through the lens of critical realism. Rosa (1998) and Carolan (2004) both offer very similar looking and compatible, but apparently independently developed40, conceptual frameworks for understanding of the interaction between social and biophysical components. Figure 5.1. Conceptual diagrams from Rosa (1998:35) on left and Carolan (2004:515) on right. Rosa developed his framework (shown on the left in Figure 5.1) in the context of studying and understanding risk in the face of meta-theoretical debates in the sociology of risk. He takes a position of ontological realism but recognizes and incorporates the potential 40 Carolan does not cite Rosa’s 1998 article. 207 usefulness of constructivist insights into his model. He does this by including epistemological hierarchicalism with his ontological realism (what he acronyms to OREH). What he means by epistemological hierarchicalism is that while all knowledge claims are potentially fallible, not “all knowledge claims are equally fallible” (Rosa 1998:34). His method of determining where a knowledge claim falls in the hierarchy is to ask where it falls in its measures of “ostensibility” and “repeatability” or “O&R.” “Ostensibility” is the degree to which observations align while “repeatability” is the degree to which repeated observations align. In Figure 5.1 above, Rosa places claims with high O&R in the corner of the axis and with these claims being highest in the epistemological hierarchy and thus being most appropriately understood by a “grounded realist” approach. On the other hand, claims with low O&R are at the other end of the scale and are best understood using a social constructivist approach. Claims between the two extremes easily take a middle approach, i.e., critical realism, or what Rosa calls “synthetic realism.” Carolan takes an ultimately similar but slightly different approach. He suspends belief in a single realist ontology not because he thinks that a physical reality doesn’t exist but because he recognizes that some environmental problems are more epistemologically distant and complex than others, and asks “What if multiple knowledges reflect not only varying positions but, in certain situations, a multiple ontology?” (Carolan 2004:498). Thus, he proposes that the more epistemologically distant and complex a phenomenon is the harder it is for us to know about it and the more ontologically fuzzy it is. He organizes the degree of fuzziness in his diagram of ontological orders (shown on the right in Figure 5.1). First order phenomena, like litter, are ones that are epistemologically closest to us—we can assess them with our own physical senses. Second order phenomena, like dioxin and nitrogen, area a step removed from our own senses; we can identify clearly what they are, but we need to use tools to measure or “see” them. Third order 208 phenomena, like global climate change, cannot be “seen” in their entirety, even with tools; we can only know them by observing other things (like temperature, size of ice caps etc.) that we then “translate” “into” the phenomena itself. Carolan (2004:505) says that this translation causes a shift (even if a small one) in our reality “because the object itself has shifted.” Not surprisingly, Carolan posits that at higher ontological orders trust in information and its sources becomes more important to hold together the complexity, and navigate the increased potential for multiple “realities.” If we were to lay Rosa and Carolan’s diagrams on top of each other we can easily see how similar and compatible they are. Ostensibility is highly related to epistemological distance—things that are easily observed are epistemologically closer and thus observations are more likely to be in agreement. Similarly, complex and epistemologically distant things are harder to observe and measurements are less likely to be repeatable, as differences in timing, place, instrument etc. are more likely to produce a variety of results. Agreement on these more complex and distant phenomena are therefore going to require more trust in those producing the results and the processes they use. Additionally, while realism certainly still applies to complex and epistemologically distant phenomena, the increased multiplicity that arises at higher ontological levels makes social constructions surrounding those empirical observations of greater political and research interest. As Carolan (2004:513) puts it: “if reality is social—and thus multiple—than it is also political” and we have seen this play out clearly in the case of global climate change where different groups have created different (and often political) narratives about the existence and potential impact of climate change and have expressed a lack of trust of the scientific observations that have been made of it (see for example: McCright and Dunlap 2003; McCright 2007; Steel, Lach, and Satyal 2006; Zehr 2000). Agricultural nitrogen would 209 likely fall somewhere in the middle of these models—it is not overly complex in that it can be fully measured, but it also cannot be observed directly, being only “visible” through a chemical test or plant growth—including algae in local water bodies. In relation to farmer decision-making, these critical realist frameworks and the idea of “epistemological distance” helps contextualize several of the notable results of my analyses and suggests future research directions. One is the potential for farmers to respond differently to the same biophysical situation, as was found in the tillage analysis, indicates either that the biophysical situations are not in fact the same, and/or that as seemingly tangible as soil is it still has enough ontological distance to complicate the relationship between the biophysical conditions and the farmers’ actions. More detailed data on farm conditions, such as researcher conducted soil tests and geospatial data, that can be coupled with more detailed information from farmers on how they till and what they are trying to accomplish with their tillage, would help tease out the sources of this variation. Second, the importance of information sources in Chapter 3 relates well to the need for greater trust in information sources that comes with greater epistemological distance. While I did not include trust in information sources in the analysis explicitly, the frequency of use should be related to both trust and mode of access (Stuart et al. 2018). The association of specific information sources with specific practices suggests that something beyond a general trust in the source is at play, which bears further investigation. Third, critical realism provides opportunity for developing an alternative approach to thinking about the relationship between attitudes, knowledge and action than that proposed by the TPB/RAA. The adoption-diffusion model, while certainly including the importance of trust 210 and communication could likely benefit by being considered within an explicitly critical realist framework as well. Lastly, epistemological distance is an important aspect to keep in mind when considering coupled natural and human (CNH) systems, and especially modeling feedbacks within them between the biophysical and social components. The critical realist approach suggests that these feedbacks will only be as good as the social system (or its relevant segment given the topic) is at recognizing the effects that society is having on the environment. Thus, the dominant narrative about the environment will mediate the feedback, rather than there being a direct relationship between environmental conditions. A very simplified example of this is shown in Figure 5.2 part B, where water quality influence the social narrative that in turn influences water policy to address it, rather than being conceptualized as having a more directed reciprocal relationship (part A.). Figure 5.2. Simple example of realist (part A.) versus critical realist (part B.) approach to CNH feedback model. 211 PLANS FOR FUTURE RESEARCH My plans for future research extend beyond what I suggest in the analysis chapters and includes dividing several of the chapters into multiple papers that will be a suitable length for publication in academic journals. Tillage My first analysis chapter will be divided into two parts. One that covers the variation in the ways that farmers are tilling and the implications for tillage categories, and one that covers the reasons farmers gave for why they till the way they do. Extensions of this research could use survey data to tease out tillage variation using a larger number of observations as well as more representative and systematic data. Potential research questions include: How common is the use of multiple practices? Are farmers who use multiple practices different from those who use only one type? Does it depend on which combinations of tillage types a farmer uses? A mixed methods approach using the interviews data could be very helpful in interpreting the results of the quantitative analyses. Given the measurement challenges discussed above, I think that it would be profitable to compare the results of the tillage questions asked on the four farmer surveys I have been involved with to date and compare how the questions have been formulated. This type of analysis could reveal aspects that are consistent across samples and measures and provide guidance in the design of future surveys on tillage. In the longer term, I would like to work with soil scientists to better understand the measurable biophysical context in which farmers are tilling and the tillage features that have the most important environmental implications. In other words, can we narrow the number of tillage facets that we need to measure to focus on those that matter the most? For example, how critical 212 is the distinction between no-till and strip till environmentally, or is it the season of the tillage that matters more? BMP Adoption The second analysis will be divided into at least two journal articles. The nutrient management practices and the lack of consistency in their predictors will be one article. The drivers of cover crops will be another article—one that could use a step-wise regression model to look at the different categories of predictors separately before considering them all together, and one that could also interaction effects between select variables. A possible third article could focus on the probabilities of practice use and could utilize more of the response categories that were collected on the survey—including both adoption and non-adoption as possible outcomes. Future work could consider the combinations in which practices are used, and the use of one practice to predict another. This could potentially lead to the creation of practice categories that have an empirical basis and that may not be based on practice function, but on some other aspect of compatibility, such as practice timing. This data could also be used for a preliminary comparison of the predictive power of the adoption-diffusion model vs. the TPB/RAA, as I believe that there is the need for a critical look at the appropriateness of the TPB/RAA for farmer decision-making and the potential for a useful revival of the adoption-diffusion model. I also am looking forward to using more recent survey data to consider the effects of practice-specific measure of knowledge, and the relationships between knowledge, education, experience, information sources and practice use in an analysis that could be a step towards incorporating a critical realist approach to understanding practice adoption. 213 Fertilizer Use The third analysis will not be divided. Future extensions of this analysis could include the use of non-recursive models that consider the reciprocal effects between corn yield and fertilizer use. The 2017 AgCensus data (to be released in 2019) will make possible a longitudinal examination of cover crops, conservation tillage and no-till. Watershed scale models (facilitated by data integration using GIS software) could allow for the inclusion of a true feedback loop between water quality and practice use in the CNH tradition. Further developments along these lines would be greatly facilitated by more specific data at the county and watershed scale (like fertilizer spending on corn acres specifically)—which might be available from NASS by special tabulation or other arrangement. As I have discussed above, I plan to continue the work that I have started in my dissertation in a number of ways. I am particularly interested in investigating the relationships between agricultural decisions and the biophysical environment, and in developing innovative modeling strategies that integrate social and biophysical data. In doing so, I plan to utilize geographic information system (GIS) tools and agent-based modeling alongside regression-based models. 214 REFERENCES 215 REFERENCES Carolan, Michael S. 2004. “Ontological Politics: Mapping a Complex Environmental Problem.” Environmental Values 13(4):497–522. Carolan, Michael S. 2005. “Society, Biology, and Ecology: Bringing Nature Back Into Sociology’s Disciplinary Narrative Through Critical Realism.” Organization & Environment 18(4):393–421. Carolan, Michael S. 2006. “Do You See What I See? Examining the Epistemic Barriers to Sustainable Agriculture.” Rural Sociology 71(2):232–60. Fishbein, Martin and Icek Ajzen. 2015. Predicting and Changing Behavior: The Reasoned Action Approach. New York, NY: Routledge. McCright, Aaron M. 2007. “Dealing with Climate Change Contrarians.” Pp. 200–212 in Creating a Climate for Change, edited by S. Moser and L. Dilling. New York: Cambridge University Press. McCright, Aaron M. and Riley E. Dunlap. 2003. “Defeating Kyoto: The Conservative Mevement’s Impact on U.S. Climate Change Policy.” Social Problems 50(3):348–73. Rosa, Eugene A. 1998. “Metatheoretical Foundations of Post-Normal Risk.” Journal of Risk Research 1:15–44. Steel, Brent S., Denise Lach, and Vijay A. Satyal. 2006. “Ideology and Scientific Credibility: Environmental Policy in the American Pacific Northwest.” Public Understanding of Science 15(4):481–95. Stuart, D., R. C. H. Denny, M. Houser, A. P. Reimer, and S. Marquart-Pyatt. 2018. “Farmer Selection of Sources of Information for Nitrogen Management in the US Midwest: Implications for Environmental Programs.” Land Use Policy 70:289–97. Zehr, Stephen C. 2000. “Public Representations of Scientific Uncertainty about Global Climate Change.” Public Understanding of Science 9(2):85–103. 216