”Supra ”$212.5 . .z . .r fix?“ I a i... c , Ovul- .. , . Ill ufifi $1. KIA? : 2.. .JJ. JK‘hvg fig ‘1» i... u: l 31.... , 4‘1 I o: n. z a “mm... In . r y .r. 32-5. . . ‘ 33.3. x“: #337,? ”any? . 5 1335.". 32 1 .» 3.23.. In ,x. 5.53.. iii}. ”Jun... lflt .mnfuai dd“? (:8; W$¥xéwrflwxln ,3 s ‘ .. .5. . \ . , i“: ”my 1.. y... n .(«mmm 6v} . nun, as. 5&3”??? hwy? . . I .. £4.5- . 3 i 5: .i a. v‘ I '4'..- . 3...“. .9 5.13m. insane ' muZQvl. (.1- ... 3.3;: 1...: .: i .513 vi}. . 53.4... I {.15.}. .1 nwdifils . x 5) v . «543...! . It. .03 v: A; «0:! 5 i0 1 : 004 ’0‘: 5 S 75 I” LIBRARY Michigan State UNIVG TSIIY This is to certify that the dissertation entitled ECONOMIC CHOICE MODELING: THE USE OF SOCIAL PREFERENCE DATA TO INFORM WHITE-TAILED DEER MANAGEMENT IN MICHIGAN presented by Kristy Wallmo has been accepted towards fulfillment of the requirements for the Ph.D. degree in Fisheries and Wildlife ./'/WWA Major Profe'ssor’s Signature 7/3/206 3 Date MSU is an Affirmative Action/Equal Opportunity Institution PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE NW 1108 32th 6/01 c:/ClRC/DateDue.p65-p.15 ECONOMIC CHOICE MODELING: THE USE OF SOCIAL PREFERENCE DATA TO INFORM WHITE-TAILED DEER MANAGEMENT IN MICHIGAN By Kristy Wallmo A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Fisheries and Wildlife 2003 ABSTRACT ECONOMIC CHOICE MODELING: THE USE OF SOCIAL PREFERENCE DATA TO INFORM WHITE-TAILED DEER MANAGEMENT IN MICHIGAN By Kristy Wallmo White-tailed deer populations in Michigan (Odocoileus virginianus) have risen steadily over the last 30 years, with more than one million deer in the fall counts since 1981. The abundant deer population has created both benefits and costs for the Michigan public. While attitudes toward deer are generally positive, when faced with the costs of abundant deer populations, for example deer damage to agriculture or deer-vehicle collisions, preferences for deer populations may change. With the increasing attention given to public input, wildlife managers need to be informed of preferences for deer populations in light of the benefits and costs associated with deer. Choice experiment surveys are well suited for this task, as they require individuals to make constrained choices, reflecting realistic management situations where trade-offs must be made. This research uses a choice experiment mail survey to estimate choice models of preferences for deer populations and a suite of deer-related attributes. Focus groups and in-person pretests were conducted to determine which deer-related attributes are most relevant to the Michigan public and to facilitate survey development. Deer-related attributes that were used in the survey included the number of deer, the number of mature bucks, herd health, deer damage to residential property, deer damage to agriculture, deer-vehicle collisions, and deer damage to forest ecosystems. The survey was mailed to hunters (N=1,980) and nonhunters (N=2,970) in three regions of Michigan: the western upper peninsula, the northeastern lower peninsula, and the southwestern lower peninsula. Response rates ranged from 64% to 66% for hunters and 59% to 63% for nonhunters. Choice model results indicate that, in addition to deer numbers, other deer-related attributes have a significant effect on the utility of both hunter and nonhunter respondents, and both groups will consider the costs associated with deer when making choices among deer-management scenarios. Results suggest that while both groups will make trade-offs for changes in the deer population size, the types and magnitudes of the trade-offs differ among regions and between hunters and nonhunters. For example, for an increase in the deer population, hunters will accept larger increases in most, but not all, of the deer-related attributes than will nonhunters. Comparing the choice model results with other survey components demonstrates that preferences for deer and the related attributes are consistent across different measurement scales. Results of this survey can be used to inform management of the relative importance of different deer-related attributes and the types of trade-offs people are willing to make among them. ACKNOWLEDGMENTS A million thanks go to Frank Lupi, who has been an excellent advisor and has given me invaluable support and guidance during this research and writing of my dissertation. Thanks so much Frank for all your help over the last four years. Thanks also to my committee members Ben Peyton, John Hoehn, and Angela Merti g for their insight and advice, and to Peter Bull, who helped out extensively with the survey development. I would also like to thank the Department of Fisheries and Wildlife at Michigan State University and the Michigan Department of Natural Resources for funding and support of this research. I am gratefirl for the support of my parents Bill and Barbara Wallmo and the many friends who made the dissertation process easier and the years in East Lansing very enjoyable, so thanks to Laura Granack, Laura Cimo, Heather Holtzclaw, Kimberly Ludwig, Dan Lerner, Beth Dunford, Loretta Ishida, Heather Lundrigan, and the HD lab. Karen Wayland was (and is) instrumental to my life and sanity, and I feel lucky to have made such a wonderful fi'iend. Finally, many thanks to the best thing to come out of my time in Michigan, my husband Jerry Hovis. iv TABLE OF CONTENTS LIST OF TABLES ................................................................................................ vii LIST OF FIGURES ............................................................................................... ix KEY TO ABBREVIATIONS ............................................................................... x CHAPTER 1. INTRODUCTION ........................................................................ 1 White-tailed Deer in the US. ................................................................... 1 Management Concepts for Human-Wildlife Conflicts ............................. 5 Research Framework ................................................................................. 9 CHAPTER 2. ATTRIBUTE BASED DECISION MAKING ............................. 16 General Framework ................................................................................... 16 Choice Experiments....................................... ........................................... 18 Random Utility Theory ............................................................................. 21 Application to White-tailed Deer .............................................................. 24 Survey Development and Implementation ................................................ 26 CHAPTER 3. SURVEY ANALYSIS AND RESULTS ...................................... 48 Characteristics of Survey Respondents ..................................................... 48 Attitudes and Experience with Deer and Deer-related Attributes ............. 52 Respondent Segmentation Analysis .......................................................... 68 Stakeholder Satisfaction and Issue Activity Related to Alternative Deer Scenarios ................................................................................................... 77 CHAPTER 4. CHOICE MODEL ESTIMATION AND RESULTS ................... 87 Background ............................................................................................... 87 Statewide Choice Models .......................................................................... 93 Regional Results ....................................................................................... 100 Comparison of Preference Elicitation Formats ......................................... 108 CHAPTER 5. CONCLUSIONS AND MANAGEMENT IMPLICATIONS ...... 113 Preference for Deer and Deer-related Extemalities .................................. 113 Satisfaction and Issue Activity Related to Deer and Extemality Levels... 119 Limitations of the Research ...................................................................... 124 The Survey Instrument and Choice Experiment Format ........................... 126 The Big Picture for Deer Management ..................................................... 130 APPENDICES ...................................................................................................... 135 Appendix A. Survey Instruments ............................................................. 135 Appendix B. Survey Correspondance ...................................................... 172 BIBLIOGRAPHY ................................................................................................. 178 LIST OF TABLES Table 1 Focus Group Attendance ..................................................................... 32 Table 2 Model of Deer Crop Damage in Wisconsin ........................................ 38 Table 3 Pre-test Attendance .............................................................................. 42 Table 4 Survey Response Rates by Region ...................................................... 46 Table 5 Disposition of the SOS and MDNR Samples ...................................... 47 Table 6 Characteristics of Survey Respondents ............................................... 50 Table 7 Hunting Activity of Survey Respondents ............................................ 52 Table 8 Respondent Attitudes/Experience with Deer ...................................... 55 Table 9 Respondent Attitudes Toward Mature Bucks ..................................... 56 Table 10 Respondent Attitudes Toward Herd Health ........................................ 58 Table 11 Respondent Attitudes/Experience with Deer Damage to Residential Property ............................................................................................... 59 Table 12 Respondent Attitudes/Experience with Deer Damage to Agriculture ........................................................................................... 61 Table 13 Respondent Attitudes/Experience with Deer-vehicle Collisions ........ 62 Table 14 Respondent Attitudes/Experience with Deer Browsing in the Forest ................................................................................................... 63 vii Table 15 Table 16 Table 17 Table 18 Table 19 Table 20 Table 21 Table 22 Table 23 Table 24 Table 25 Table 26 Table 27 Table 28 Respondent Agreement with Belief Statements .................................. Respondent Characteristics Correlated with Deer and Buck Numbers .............................................................................................. Respondent Characteristics Correlated with Deer-related Attributes ............................................................................................. Regression Models of Respondent Satisfaction with Deer and Deer- related Attributes ................................................................................. Satisfaction and Issue Activity Associated with Deer Population Changes ............................................................................................... Types of Choice Models Estimated from Survey Data ....................... Results of Statewide Choice Models ................................................... Hypothetical Outcomes for Deer-related Attributes ............................ Marginal Rates of Substitution for Deer and Mature Bucks: Statewide ............................................................................................. Results of Regional Choice Models: Hunters ..................................... Results of Regional Choice Models: Nonhunters ............................... Hunter Marginal Rates of Substitution for Deer and Mature Bucks: Regional ............................................................................................. Nonhunter Marginal Rates of Substitution for Mature Bucks: Regional .............................................................................................. Results of Pooled versus Four-way Choice Model ............................. viii 65 74 75 81 85 90 94 96 98 102 103 106 107 110 Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 LIST OF FIGURES Hypothetical Cultural Carrying Capacity for Deer ............................. 7 Trends in Michgan’s White-tailed Deer Population ........................... 11 Annual Deer-vehicle Collisions in Michigan ..................................... 13 Hypothetical Fishing Scenarios .......................................................... 17 Hypothetical Choice Scenario ............................................................ 25 Study Areas ........................................................................................ 28 Status Quo Levels of Deer-related Attributes ..................................... 39 Choice Set for Northwest Region ....................................................... 43 Respondent Rating of Satisfaction ..................................................... 78 Example of Choice Set ....................................................................... 88 Schematic of Nested Multinomial Logit ........................................... 91 ix CCC CE MDNR QDM RUM RUT SC 303 SW USFWS WAC WDACP WDNR KEY TO ABBREVIATIONS Cultural Carrying Capacity Choice Experiment Independence from Irrelevant Alternatives Michigan Department of Natural Resources Multinomial Logit Marginal Rate of Substitution Northeast Region (Alpena, Alcona, Oscoda, Montrnorency, Presque Isle counties) Northwest Region (Marquette, Iron, Dickinson, Baraga counties) Quality Deer Management 3 Random Utility Model Random Utility Theory Stated Choice Secretary of State Southwest Region (Barry, Eaton, Calhoun counties) United States Fish and Wildlife Service Wildlife Acceptance Capacity Wildlife Damage Abatement and Claims Program Wisconsin Department of Natural Resources CHAPTER 1. INTRODUCTION White-tailed Deer in the US. White-tailed deer (Odecoileus virginianus) populations in much of the US. have increased substantially over the last century. Though there is no consensus on the size of the current population in the US, estimates have ranged from 15 to 25 million nationwide (McCabe and McCabe 1997). There is, however, consensus among many scientists that in much of their North American range deer population densities currently exceed historical levels that prevailed at the turn of the century (Alverson et a1. 1988; deCalesta 1997; Healy 1997; Woolf and Roseberry 1998). This increase has resulted in both costs and benefits to society. In the early 1900's white-tailed deer hunters faced restrictive hunting limits due to the near extirpation of deer (Woolf and Roseberry 1998). However, by the 1920's dee: herds had recovered in many states (Woolf and Roseberry 1998) and in 1996 deer were I the most popular type of big game for hunters, with 10.7 million hunters spending about 131 million days hunting deer (U SF WS 1996). While these are national figures and ’ therefore include deer species other than Odecoileus virginianus, buck harvests in 13 northeastern states which include only white-tailed deer show an average increase of 164%, with a range of 44%-525%, during the period 1983 - 1992 (Organ and Ellingwood 2000). In addition to hunting, an abundant deer population provides more opportunities for wildlife viewing and photography. In 1996 about 13 million people in the US. spent time away from home observing large mammals such as deer, bears, and coyotes and m about 17 million people spent time wildlife watching for large mammals around their 1 own home (USFWS 1996). . While increases in the white-tailed deer population have created obvious benefits in terms of more opportunities for wildlife-related activities, there are also a variety of conflicts generated from such increases. These conflicts may be viewed as the costs, or negative extemalities, hereafter referred to simply as extemalities, associated with deer populations, particularly the abundant populations present in many parts of the US. Data from the US. Department of Agriculture Wildlife Services shows that during the period 1994 - 1997 white-tailed deer were one of the top ten nuisance species in the northeastern U.S. (Organ and Ellingwood 2000), causing a variety of problems including damage to agriculture, trees, and plants, collisions with vehicles, and human health risks associated with deer. Deer browsing on commercial crops may cause economic losses to agricultural producers. Research by Conover (2002) suggests that 51% of agricultural producers in the US. stated that deer had caused damage on their farm or ranch in the previous year, and deer were most often cited as being the source of wildlife damage. Conservative estimates put deer damage to agriculture in the US. at around $100 million annually (Conover 2002). Other studies have produced regional estimates of deer damage to agriculture. In Indiana reports suggest that 90% of all wildlife-related financial losses to farmers were due to deer, and in 1993 deer were responsible for approximately $4.5 million of loss in harvestable corn in Indiana, representing about one third of all harvestable corn losses due to wildlife (McCreedy et a1. 1994). In Wisconsin, the Department of Natural Resources has implemented a Wildlife Damage Abatement and Claims Program (WDACP) to provide abatement assistance and monetary payments for wildlife damage to crops. In 1999 a WDACP report stated that deer damage represented 89% of appraised losses statewide, amounting to approximately $1.5 million. Other research suggests that deer were responsible for about $37 million of loss in corn, soybean, and wheat production in Maryland, (McNew and Curtis 1997), and that fruit growers in Western New York reported average losses of $3,000 per year (Decker and Brown 1982; McNew and Curtis 1997). Deer browsing can also cause damage to residential property by damaging ornamental plants, trees, and shrubs. In a random sample of households in the 100 largest metropolitan areas in the US. Conover (1997) states that 4% of homeowner respondents reported a problem with deer, with damage estimates of about $251 million per year and approximately $125 million spent annually to prevent the damage. Some regional estimates suggest that homeowners in southeastern and western New York reported median losses of $200 and $90, respectively, per household in 1988, while nursery producers in southeastern and western NY experienced an average loss of $21,628 and $3,813 due to deer in the same year. (Sayre et al. 1992). Abundant deer populations may also result in large numbers of deer-vehicle collisions. Conover (2002) estimates that 726,000 deer-vehicle collisions are reported annually - about one-half the number that actually occur - at an average cost of $1,644 per collision. Nationwide approximately 29,000 people are injured and 200 people lose their lives each year due to deer-vehicle collisions (Conover 2002). In some regions of the US. increases in the annual number of deer-vehicle collisions have been considerable. For example, Indiana experienced a 5 fold increase in the number of annual deer-vehicle collisions between 1981 (2,000 collisions) and 1993 (>10,000 collisions), though the number of collisions per billion miles driven in the state has remained nearly constant since 1992 (McCreedy et al. 2001). Another extemality associated with an abundant deer population is deer browsing in forests. Deer browsing on small trees and tree seedlings can affect both natural and commercially managed forests, causing compositional changes in the flora and fauna of the forest as well as economic losses to the forest industry. Conover (1997) suggests that nationwide deer may cause $750 million worth of damage to the timber industry annually, though the damage varies widely by region. Black et al. (1979) estimated that in the Pacific Northwest 5 years of deer browsing resulted in losses in the range of $90 to $190 million, whereas Marquis (1981) estimated that annual timber losses due to deer in the Allegheny hardwood forest of Pennsylvania amounted to $367 million per year. Research by T ilghman (1989) demonstrated that deer browsing affects both commercial and noncommercial tree species as well as some herbaceous ground cover plants in Pennsylvania. Other studies suggest that heavy deer browsing in forests may eliminate some woody and herbaceous species, reduce overall plant species richness, and change the composition of trees in the forest ( deCalesta and Stout 1997; Alverson et al. 1988). Deer browsing has also been shown to have negative effects on some endangered lilies and orchids (Miller et al. 1992) and can change the types of birds that forage and nest in the forest (deCalesta 1994; McShea and Rappole 2000). Other extemalities related to abundant deer populations include health-related issues concerning deer and humans. For example, deer are an important reservoir for adult ticks that cause Lyme disease in humans. Lyme disease transmission has been cited as a deer related concern by respondents in several studies (Decker and Gavin 1987; Stout et al. 1997; Cristoffel and Craven 2000). Additionally, increases in the deer population can potentially decrease the overall health of the herd and the buck to doe ratio (McShea et al. 1997), both of which have been cited as deer-related concerns (Cristoffel and Craven 2000; Bull and Peyton 1999). Management Concepts for Human-Wildlife Conflicts The purpose of the above is to demonstrate that, in general, white-tailed deer p0pulations are abundant in many regions of the US, creating both costs and benefits that affect a variety of segments of society. While many studies have shown that attitudes toward deer are positive (Lauber et a1. 2001; Cristoffel and Craven 2000; Diefenbach et al. 1997; Decker and Gavin 1987; Stout et al. 1997; Curtis and Lynch 2001), theoretical and empirical work suggest that for many species that have the potential to generate wildlife-human conflicts there is a ‘inldlife acceptance capacity” (WAC), defined as the maximum wildlife population level in an area that is acceptable to people (Decker and Purdy 1988). Minnis and Peyton (1993) expanded the WAC concept by incorporating three additional areas: the consideration of the preferences of multiple stakeholders, the consideration that wildlife populations may not only be too high but also too low, and the consideration of human responses to wildlife populations (Gigliotti et al. 2000). Minnis and Peyton (1993) refer to this expanded concept as Cultural Carrying Capacity. Cultural Carrying Capacity (CCC) is defined as the wildlife population level in a defined area that produces the most manageable amount of issue activity at a particular time (Minnis and Peyton 1993), where issue activity is considered to be activities undertaken by stakeholders or stakeholder groups that may potentially undermine the authority of a particular agency to manage wildlife. For example, activities such as letter writing, phone calls, attendance at public meetings, etc., may be undertaken by stakeholders to express their concems about wildlife management. When stakeholders believe an agency is not responsive to their concerns, they may communicate their opinions to higher authorities (Decker et al. 1985). If such communications result in an agency losing control over management, the issue is considered “disruptive” (Peyton 1984). Minnis and Peyton (1993) propose that when stakeholders find the wildlife population size to be desirable or satisfactory, they will not engage in any issue activity and the issue will be considered latent. When the population size becomes tolerable, stakeholders will begin to communicate their concerns about wildlife, and the issue emerges. When the population size becomes intolerable, stakeholders will begin to actively seek to change the situation, potentially resulting in disruptive issue activity. The reactions of stakeholders to the population size — in both attitude and action (e. g. tolerance and issue activity) are identified by a stakeholder response curve. Figure 1 represents a hypothetical scenario with two stakeholder response curves to deer management. The population size that corresponds to the region where all stakeholders lie between desirable and tolerable is considered the CCC for a wildlife population (Minnis and Peyton 1993). Public Attitude Issue Activity IntolerableI Active CCC A Tolerable Farmer Response Curve Hunter Response Desirable Cum None None Low Moderate High Very High Number of Deer Figure 1. Hypothetical Cultural Carrying Capacity for Deer Monitoring of wildlife damage reports is a fundamental tool used by many state agencies to address the concepts of WAC and CCC (Organ and Ellingwood 2000). Other tools include citizens task forces, used in New York to gain stakeholder input into the management of white-tailed deer (Carpenter et al. 2000) and the “Inquisitive Approach,” described by Decker and Chase (1997) as actively seeking input through public hearings, telephone and mail surveys, open houses, and focus groups. However, as Gigliotti et al. (2000) point out, many managers already know that deer hunters want more deer and landowners want fewer deer - indeed a variety of research has been conducted on stakeholder preferences for increasing and decreasing deer populations (Decker and Gavin 1987; Stout et a1. 1997; Diefenbach et al. 1997; Lauber et al. 1999). Further, as noted above, previous studies have shown that people generally have a positive attitude regarding deer - it is the existence of deer-related extemalities, the magnitude of which depends at least in part on the size of the deer population, for which people become intolerant. Previous research concerning attitudes toward deer and deer-related extemalities has shown that women tend to be more concerned about many deer problems than do men, and that people who have had direct experience with some deer-related extemalities are more likely to favor decreasing deer populations (Lauber et al. 2001; Decker and Gavin 1987). Other studies have documented the mportance of herd health (Cristoffel and Craven 2000) and property damage, crop damage, landscape damage, and deer- vehicle collisions (Curtis and Lynch 2001; Connelly et al. 1987; Decker and Brown 1986; Stout et a1. 1993; Sayre et al. 1992). This type of information may be useful to managers, as it can reveal something about why stakeholders might prefer increases or decreases in the deer population and their level of concern about certain extemalities. However, additional information about stakeholder preferences may facilitate managers in maintaining deer populations that are most acceptable to a variety of stakeholders. Like most types of resource management, deer management involves trade-offs. For example, a person may benefit from being able to view deer in their yard but may incur the cost of deer damage to residential property. To some extent the person will “trade-off” increases in residential damage to be able to view more deer. However, there may be a certain amount of damage for which additional increases in viewing deer do not make the person better off, and at some point the person may prefer to have less deer in order to decrease the amount of property damage. Further, while a person may accept a certain increase in residential property damage in order to view more deer, they may be less willing to accept increases in another deer-related extemality, such as deer-vehicle collisions. While most managers already know that experiences such as residential damage and deer-vehicle collisions are extemalities associated with a deer population, they may not know the relative importance of these extemalities to different stakeholder groups. Additionally, answers to questions such as “how many more deer-vehicle collisions would be acceptable to stakeholders who want a deer population increase? ", or conversely, “how much of a decrease in deer-vehicle collisions is needed to compensate stakeholders for a deer population decrease? ” may benefit wildlife managers responsible for developing management goals and strategies, as well as assisting them in public communication efforts. To date the literature has not addressed these types of questions related to the management of white-tailed deer. Research Framework White-tailed Deer in Michigan The state of Michigan provides a framework for examining stakeholder preferences for deer and deer-related extemalities. A brief history of deer management in Michigan (MDNR 1994) shows that, like many states, the deer population in Michigan was nearly decimated at the turn of the century due to market hunting and limited hunting restrictions. Deer populations at that time were estimated to be as low as 45,000 animals. Stringent hunting regulations helped the population rebound by the 1930's, and with reports of approximately 1.125 million deer talk began of a “deer problem.” Despite attempts to manage the herd, the deer problem continued until the late 1940's, when the population peaked at about 1.5 million deer. However, in combination with liberal hunting regulations and a deterioration of deer habitat, populations began to fall, declining to about 0.5 million in 1972. At that time many deer hunters voluntarily restricted their hunting effort to bucks only. In addition, the Deer Range Improvement Program was initiated to improve habitat and acquire land for deer to help attain a goal of 1 million deer in the spring of 1981 (Langenau 1994). Since 1972 deer populations have shown a marked increase (Figure 2). The Michigan Department of Natural Resources (MDNR) currently maintains a white-tailed deer population goal of 1.3 million v... , deer in the fall herd. 10 2500 — 3‘20001 a 0 Q 51500« C O a 2 31000— 0 O. o 500“ 0 HTI Ifil fitTflr—T'lil filefl l l l I I T l Twlv—T—T 00‘6’4’oPoW®obo°o°oW¢oP®o° ,9 .3 .9 ,3 ,9 .9 ,3 ,3 .9 ,3 ,3. ,9 ,3 ,9 .9 Year i Figure 2. Trend in Michigan’s White-tailed Deer Population The abundant deer population in Michigan has resulted in increases in many of the extemalities previously discussed. For example, deer—vehicle collisions have increased fiom 34,352 in 1986 to 68,233 in 1996, declining slightly to 67,669 in 1999 (Figure 3). An average insurance claim after a collision is about $1,000 (MDNR 1987). Deer crop damage is also an extemality relevant to the state of Michigan. In 1989 nearly 2,000 agricultural producers in Michigan incurred economic losses attributable to deer (Langenau 1993). Campa et a1. (1997) estimate average crop losses due to deer for alfalfa ($13/acre, 4.7%), corn ($15/acre, 6.5%), soybeans ($19/acre, 8%), and table beans 11 ($29/acre), based on farmer’s self-reports. Other deer damage estimates derived from Fritzell (1998) suggest that median percent losses of corn ranges fiom 1.8% in the southern lower peninsula of Michigan to 16.8% in the upper peninsula; losses of soybeans range from 4.3% in the south to 6.7% in the north; and losses of alfalfa range from 0.4% in the south to 21.8% in the north. Additional deer-related extemalities are more difficult to quantify, but may include health related issues, damage to residential property, and damage to commercial and natural forests. Research in northern Michigan has shown that deer browsing can affect the regeneration of Hemlock, Northern White Cedar, and Aspen (Alverson and Waller 1997; Campa et al. 1996; Verme et a1. 1986, Frelich et a1. 1985) and deer browsing may contribute to changing ecology in northern Michigan’s conifer swamps and may change the structure of plant communities in areas of high deer density (Van Deelen 1996) 12 80,000 m 70,000 » C l ,2 60.000~ '2 . 75 50.000 1 o 0 40,000 , E ‘1 '5 3°'°°°‘. E 20,000 a 3 10.000 ; n l o .1- , , . ,, . L , , a" . , ‘ 1986 1992 1993 1994 1996 1997 1998 1999 Year Figure 3. Annual Deer-vehicle Collisions in Michigan Research Objectives and Questions Deer management professionals as well as various stakeholders have questioned the basis of the current goal of 1.3 million deer in the fall herd and expressed a need to re- examine this number in the context of biological and cultural carrying capacities. The goal of this research is to use a survey approach to assess the CCC for deer in three distinct regions of Michigan by examining stakeholder preferences for deer and deer- related extemalities and determining the trade-offs stakeholders are willing to make for increases and decreases in the deer population. The four objectives below will facilitate the overall goal of the research: 13 (1) (2) (3) (4) Determine the general level of concern for deer and deer-related extemalities. Estimate a quantitative relationship of stakeholder preferences for deer and the associated extemalities. Assess stakeholder satisfaction for varying deer population sizes and the associated changes in extemalities Assess the likelihood of issue activity for varying deer population sizes and the associated extemalities The following research questions will be addressed for study sites in the western upper peninsula, northern lower peninsula, and southern lower peninsula of Michigan: (1) (2) (3) (4) How does the relative importance of deer and deer-related extemalities differ among different stakeholder groups? What kinds of changes in deer populations and associated extemalities are most likely to satisfy different stakeholder groups? What types of extemality increases will stakeholders accept for increases in the deer population, and conversely, what types of extemality decreases will compensate stakeholders for decreases in the deer population? What kinds of changes in deer populations and associated extemalities are most likely to induce different stakeholder groups to engage in issue activity? 14 The remainder of the dissertation is organized as follows: Chapter Two describes the general framework used to guide the research, including theory, methods, and a detailed description of the survey development and implementation process; Chapter Three describes the general results of the survey, including descriptive statistics and analytical results concerning stakeholder satisfaction and issue activity; Chapter Four describes the remainder of the analytical results including the quantitative relationship of stakeholder preferences for deer and the associated extemalities and the types of trade- offs stakeholders will make for increases or decreases in the deer population; Chapter Five provides a discussion of the results and the implications for white-tailed deer management in Michigan. 15 CHAPTER 2. ATTRIBUTE BASED DECISION MAKING This chapter begins with a general description of attribute based decision making and the process of developing a choice experiment. A discussion of Random Utility Theory (RUT) and choice model estimation is then presented, followed by an application of RUT to white tailed deer. The chapter concludes with a description of the choice experiment survey development and implementation process for white tailed deer in Michigan. General Framework The approach taken in this research is based on a set of methods referred to as Stated Choice (SC), which elicit preferences by asking people to choose among one or more “goods.” Generally, a survey or interview format is used to elicit preferences for what have traditionally been market goods; however, in recent years SC methods have become increasingly popular for valuing environmental goods and services (Hanley et al. 1998). The SC approach is built on the work of Lancaster (1966) and Lancastrian consumer theory, with additional behavioral foundations in judgement and decision making from economics and psychology (Adarnowicz et al. 1998a). Stated choice approaches postulate that individuals derive utility (worth or well- being) from a good based on the characteristics, or attributes, of the good (Louviere et al. 2000). For example, suppose that a fishing trip is viewed as an environmental “good.” There may be certain attributes associated with going fishing that make the trip more or less enjoyable. These attributes might include things like whether the fishing is done on a 16 lake or a river, the scenery surrounding the fishing area, the number of other people fishing in the same area, the expected success rate of catching fish, and the distance of the fishing area from ones home. A person who doesn’t mind driving in order to fish in a pristine area with few people around may prefer a fishing experience like the one described in A (Figure 4), while someone who doesn’t mind a more crowded area if it takes less time to reach the fishing spot might prefer the experience described in Scenario B. Fishing Scenario A Fishing Scenario B Fishing in this scenario involves... Fishing in this scenario involves... - Fishing on a relatively pristine - Fishing on a lake with other fishers river with few other people present as well as some boat traffic around 0 A high chance of catching one or 0 A low chance of catching a legal more legal size fish size fish ° Driving less than an hour to reach 0 Driving several hours to reach the lake the river Figure 4. Hypothetical Fishing Scenarios These hypothetical attributes of a fishing trip may or may not be attributes that are relevant to a particular population. In practice, when using a SC approach determining the relevant attributes of a good and describing them requires substantial research. However, the point of the above is to illustrate the concept proposed in Lancastrian consumer theory - that utility (for a fishing trip) can be decomposed into separable utilities for the attributes of the good (type of water body, scenery, crowd/congestion, success rate, and distance to fishing spot). The fact that a person would prefer A to B in 17 Fig. 4, or conversely, B to A, reveals something about the relative importance placed on the attributes of the good. It is important to note that SC methods themselves do not constitute a theory of behavior, but rather, they are a means to generate data about an individual’s behavior (Adamowicz et al. 1998a). Choice Experiments Choice experiments are a specific type of SC method that elicit preference data. A choice experiment (CE) elicits data by asking people to choose between one or more constructed scenarios. These constructed scenarios describe a particular good by decomposing the good into relevant attributes, as in the fishing trip scenarios above. Alternative scenarios are constructed by varying the levels of the attributes according to an experimental design plan. For example, consider the attribute “success rate” in the fishing trip scenarios. Hypothetically, if “success rate” has three levels - low, moderate, and high - an experimental design plan would be used to determine the level “success rate” would take in each alternative scenario. Similarly, the design plan would specify the levels of the other attributes in each scenario, such that each scenario is different. To elicit preference data individuals are then asked, typically in either a written survey or in- person interview format, to choose which of the alternative scenarios they prefer. This type of preference data allows Random Utility Models (RUM) to be estimated, producing a quantitative model of the utility derived from attributes of a good. There are several stages involved in developing a CE. The first is to identify the salient attributes of the good in question. Adamowicz et al. (1998a) note that this stage is 18 the most important stage of the study, as it characterizes a decision problem in terms that are both relevant and understandable to the decision maker. Ultimately, sparse attention to this stage can lead to irrelevant, meaningless, or biased models. Louviere and Tirmnermans (1990) suggest using multiple qualitative techniques such as in-depth interviews, focus groups, and direct questioning, in addition to planning and policy guidelines, to identify a set of salient, nonredundant attributes that characterize the good. The next stage involves specifying the levels for each attribute. The levels may be specified using a variety of criteria, including research objectives, current or future planning horizons, physical constraints, prior research or experience, and the believability of levels by potential respondents (Louviere and Timmerrnans 1990; Adamowicz et al. 1998a). Generally the levels are set by the researcher, though exploratory research can be used to determine level ranges or endpoints. The next stage involves the use of statistical design theory to combine the levels of the attributes into different scenarios. Generally each combination of attribute levels (scenario) is referred to as a treatment in the design literature (Louviere and Timmennans 1990). The use of a hill factorial design would imply that all possible combinations of levels and attributes have been designated as a treatment for the experiment. For example, suppose that a good is characterized by four attributes, where two of the attributes have three levels and two have four levels. The full factorial would be 32 x 42, or 144 treatments. As this number of treatments would be impractical, subsets of the full factorial are generally used, with “main effects” designs - orthogonal subsets of the full factorial - being a popular design used in many CE’s (Adamowicz et al. 1998a). One 19 drawback of the main effects plan is the strictly additive (no interaction terms) model that can be estimated from this design, though Louviere et a1. (2000) state that for linear models main effects typically account for 70% to 90% of the explained variance. Statistical designs can then be used to combine alternative scenarios into a choice set, from which individuals are asked to choose their preferred scenario. The next stage involves developing a questionnaire/interview script in which an individual is ultimately faced with at least one, but usually multiple, choice sets. Theoretically, an individual evaluates the alternatives in a choice set based on the utility derived fi'om the attribute levels of one alternative as compared to the other alternatives in the choice set, implicitly trading off levels of one attribute against levels of another (Adamowicz et al. 1998a). While the overall goal of this stage is to collect information on choices, the questionnaire or interview can also be used to collect other types of data such as demographic, attitudinal, or behavioral data. Additionally, questionnaires and interviews can be used to set the stage for respondent’s decision making, conveying information about the good, the selected attributes, status quo information, policy information, or any other material the respondent would need in order to make an informed choice. Pre-testing and pilot testing of the questionnaire/interview script is recommended by most researchers involved in CB studies (Adamowicz et al. 1998a). The final stages of a CB involve data collection from the desired sample population and model estimation. As with any experiment, if statistical inferences are to be made from the sample population, simple random or stratified random sampling of the target population is used. Random Utility Theory (RUT) forms the basis for model 20 estimation. The choice model structures are based on RUT and are appropriate for describing discrete choices in a utility maximizing fiamework (Boxall et al. 1996). Maximum likelihood techniques are frequently used to estimate binary or multinomial logit models, though other specifications such as the multinomial probit and the nested multinomial logit have been applied to CE data (Adamowicz et al. 1998a). Random Utility Theory Choice models are based on Random Utility Theory (RUT), a utility maximizing framework that allows individuals to value attributes of a good (Adamowicz et. al 1998b). Random utility theory specifies that utility (U) for a good consists of a systematic, known component (V) and a random component (B). Based on RUT, the utility that individual i receives from a given alternative, a, can be expressed as Uia=via+Eia (1) where Ui, is the unobservable utility that i associates with a, V,, is the quantifiable, known portion of utility, and Bi, is the random, unobservable effects associated with a for individual i. Alternative a can be decomposed into specific attributes, and the systematic component of utility Via can be expressed as Via = fix. (2) where X, is a vector of attributes and the associated levels for alternative a and [3 is a vector of attribute coefficients. Although theoretically the attributes could vary among individuals, for the CE used in this research the attributes are the same for all individuals, thus the subscript i is dropped from the right hand side of equation (2). Substituting the 21 expression for V,,, the utility function can be expressed as U1. = 13X. + Bi. (3) The presence of the random component allows probabilistic statements to be made about preferences and/or behavior, ultimately allowing for model estimation. Under the assumption that individuals are utility maximizers, RUT specifies that the probability that an individual i will choose alternative a from a set of C alternatives is equal to the probability that the utility derived from a is greater than the utility derived from any other alternative in the choice set C. This can be expressed as Probi (alC) = Prob(Ui, > Uij) V j e C = Prob(Vi, + E,, > Vij + Eij) Vj e C = 1’r0b(l5Xa + E. > 13X,- + E11)» Vi E C (4) Equation (4) implies that the probability that an individual will choose alternative a is equal to the probability that the systematic component of utility plus the associated error for a is greater than the systematic and associated error components of all other alternatives in the choice set C (Adamowicz et al. 1998b). To operationalize (4), an error distribution must be assumed. Type I extreme value distributions, such as Gumbel or Weibull distributions, are used extensively in discrete choice models (Louviere et al. 2000). Ifa type I extreme value distribution is assumed for the random component of (4), a multinomial logit model (MNL) can be expressed as Probi (alC) = expo. Vi,)/ Zj e C expo. Vij) = expo BXJ/ 21' e c expo BX.) (5) 22 The parameter A is a scale parameter that is imbedded in RUT based models, but not identifiable separately from 0. Generally A is normalized to one, and the model is estimated as if A B are the attribute coefficients (Hensher et al. 1999). An additional property of MNL is the Independence from Irrelevant Alternatives (11A). The HA property means that the ratio of the probability of choosing one alternative over the probability of choosing another is unaffected by the presence or absence of any additional alternatives in the choice set. In many situations, the HA property is not wholly desirable, and can be relaxed by nesting the model (see Chapter 4). Equation (5) can be estimated using maximum likelihood techniques. Ifchoice observations are ordered so that the first n1 individuals chose alternative a, the next n2 individuals chose alternative b, the next n3 individuals chose alternative c, and so on for all j elements of the choice set C, the likelihood function can be written as n: Hr+n2 nr+n2+n3 l L = H Pia... H Pib... 1] Pic... Pij i=1 i=nr+l i=n1+n2+r i=I—nj+l Defining a dummy variable f-- 11’ where g = 1 when alternative j is chosen and fil- = 0 otherwise, the function can be simplified to the log-likelihood function lnL= Z nylnpg (6) The term Pij in (6) can be replaced with (5) so that the only unknown terms are A [3. Using maximum likelihood techniques, parameter estimates for elements of D can be 23 estimated for each of the attributes that have been selected to characterize the good. Application to White-tailed Deer Random Utility Theory can be used to determine preferences for white-tailed deer populations and some of the extemalities associated with white-tailed deer that have been discussed previously. For example, assume that individuals derive some level of utility (well being) from deer, and that this utility can be “decomposed” into a set of deer-related attributes, which may include some of the deer-related extemalities. Suppose that the utility for deer is a function of the population size, the number of deer-vehicle collisions, and the amount of deer crop damage. Assuming that the utility function is linear, equation (3) can be expressed as U = V + E = BX + E = 01*Deer population size + BfDeer-vehicle collisions + [33*Deer crop damage + E. where [3k are the parameter estimates that indicate the relative importance of each attribute. Preference data can be collected by showing individuals alternative scenarios that contain varying levels of population size, collisions, and crop damage and asking them to choose which scenario they like best, as shown in the hypothetical example in Figure 5. 24 Which Deer Scenario Do You Prefer? Scenario A Scenario B Scenario Scenario A includes... Scenario B includes... Scenario C includes... 0 High deer densities - Moderate deer - Low deer densities densities ° A moderate ° A moderate ° A moderate increase in annual increase in annual decrease in annual deer-vehicle deer-vehicle deer-vehicle collisions relative collisions relative collisions relative to the number of to the number of to the number of collisions last year collisions last year collisions last year 0 About the same ' A small decrease in - A large decrease in amount of deer- the amount of deer- the amount of deer- crop damage that crop damage that crop damage that occurred in the last occurred in the last occurred in the last year year year Figure 5. Hypothetical Choice Scenario In this example the attributes population size, collisions, and crop damage represent X in equation (5) and Scenarios A, B, and C represent thej alternatives. By asking individuals in a sample population to make choices among a sufficient nrunber of alternative scenarios, obtained through a statistical design plan, preference data are collected and equation (6) can be estimated. The output includes coefficients for each of the deer- related attributes that indicate the relative importance each attribute has on utility. The output can also be used to examine the trade-offs individuals will make among the attributes. For example, for a specified change in the size of the deer population, the model output can indicate the amount of change in annual collisions, or any other attribute, that individuals will “trade-off” (accept increases or decreases) and still be as 25 well off (have the same utility) as they were before any changes occurred. These trade- offs are referred to as the marginal rate of substitution (MRS). MRS is expressed by the negative of the ratio of two attribute coefficients, 6. g. - [31/02. The concept of MRS if further explained in Chapter 4. In the above example three attributes served to characterize the good “deer.” As stated previously, the identification of the salient attributes of a good is of critical importance to a CB study, and in practice more (or less) attributes may be needed to characterize preferences for deer. The following sections describe the process of identifying and describing deer-related attributes for a CB and the subsequent stages of survey development and implementation. Survey Development and Implementation Background and Description of Study Sites A choice experiment survey was developed to determine preferences for deer populations and deer-related extemalities, referred to as deer-related attributes in a CB setting. The survey was developed for two primary stakeholder groups - white-tailed deer hunters and a more general group of the public. While other stakeholder groups do exist, e. g. landowners or farmers, time and resources constrained the number of stakeholder groups that were able to be included, in particular because choice experiment surveys require a large number of survey versions due to the experimental design. However, information was collected in the survey pertaining to respondent land use and ownership so that this, as well as other socio-demographic characteristics, can help explain 26 preference heterogeneity. Survey regions were selected primarily to fit in with a larger MDNR project that consisted of three components: habitat modeling, deer population modeling, and stakeholder preferences. Habitat and deer population components dictated that the survey regions should be distinct in deer densities and available deer habitat. Additionally, representation from both the upper and lower peninsula of Michigan was desired. Three regions of Michigan were surveyed: a northwestern region (Baraga, Dickinson, Iron, and Marquette counties), a northeastern region (Alpena, Alcona, Montrnorency, Oscoda and Presque Isle counties), and a southwestern region (Barry, Calhoun and Eaton counties) (Figure 6). Each region is briefly characterized below, with attention to characteristics that may relate to deer and deer management. 27 Northwest Region: Baraga, Dickinson, Iron, Marquette counties Northeast Regim: Alcona, Alpena, Montrnorency, Oscoda, Presque Isle counties SouthwestRegion: Barry, Calhoun, Eatnn counties Figure 6. Study Areas The population of the northwest region (Region I) is approximately 110,369, with a population density of about 24 people per square mile. Per capita personal incomes range from $18,500 - $21,175, depending on the county. About 1.7% of the land area is devoted to agriculture, with hay and alfalfa being the predominant crops grown. Of the three regions, the northwest region has the highest percentage of forest land, with more than 80% of the land in each county classified as forest land. The forest industry owns less than 10% of forest land in Iron county, 10% - 20% in Marquette and Iron counties, and 40% of the forest land in Baraga county. State and national forest ownership is at 28 least 20% and can be higher than 40%, depending on the county. The MDNR estimates deer densities to be fairly low in this area relative to the northeast and southwest regions. Additionally, during the winter months, deer tend to migrate from the northern counties to the southern counties in this region, resulting in higher deer concentrations in the southern part of the region. This type of winter movement is not as pronounced in the other two study regions. The population density of the northeast region (Region II) is about the same as the northwest, with about 25 people per square mile (total pop. 74,834). Per capita personal income ranged from $16,541 - $22,000, depending on the county. The northeast region has more cropland than the northwest, with 8.2% of the land in agricultural use. Grain or seed corn is the predominant crop, though the region also grows wheat, soybeans, table beans, and alfalfa. In Presque Isle, Alpena, and Alcona counties between 60% and 80% of the land is classified as forest land, while over 80% of the land in Montrnorency and Oscoda counties is forest land. Less than 10% of the land in each county is owned by the forest industry, though state and national ownership ranges from less than 10% to over 40%, depending on the county. The MDNR has estimated moderate deer densities in Presque Isle and Alpena counties and higher densities in Montrnorency, Alcona, and Oscoda counties. Relative to the northwest region deer densities are generally higher throughout the northeast. The southwest region (Region III) is the most densely populated of the three regions, with 161 people per square mile and a total population of 295,527. Per capita personal incomes are also higher in the southwest than in either the northeast or 29 northwest, ranging between $23,111 - $24,690, depending on the county. F orty-three percent of the land in the southwest region is devoted to agriculture, growing predominantly grain and seed corn. As a reference, the southwest region grows almost 15 times the amount of com grown in the northeast region. In each county in the southwest region 20% - 40% of the land is classified as forest land. Less than 10% of forest land is used by the commercial forest industry in the southwest, and state and national ownership is less than 20% in all counties. Deer densities in this region are higher than the northeast and northwest, and the MDNR considers this region to have some of the highest deer densities in the state. The choice experiment survey used here is one component of a larger deer management project, hereafter referred to as Eco-Deer, which includes a habitat component, biological component, and the stakeholder preference component. A choice experiment survey format was chosen for the stakeholder preference component because this format requires respondents to make trade-offs in their decision-making, which often typefies real life decisions concerning natural resources. The choice experiment format also allows preferences to be estimated in a quantitative model, facilitating the Eco-Deer project goal of integrating results of the habitat, biological and stakeholder components to develop a holistic deer management tool. The following describes the choice experiment survey development and implementation process. 30 Attribute Selection In a choice experiment framework, attributes of the deer herd are things that are related to the existence of a deer population. For example, crop damage may occur in some areas due to the deer population. Likewise, a deer population means that recreational opportunities for viewing and hunting exist. Having a deer population in an area may even affect local econorrries during peak hunting seasons. Presumably, there is a large set of deer-related attributes; however, due to limited space, appropriate experimental design, and task complexity faced by the respondent, it was not possible to use the entire set of deer-related attributes in the choice experiment survey. Thus it was necessary to select a subset of attributes for use in the survey. The initial step in selecting the subset was conducting informal interviews and discussions with deer management professionals and researchers on the Eco-Deer project to determine which attributes are salient to deer management, thus narrowing the list of all possible deer-related attributes. These interviews and discussions occurred during the period of May to August 2000. After the initial interviews, three focus groups were conducted in October 2000 and February 2001 to facilitate survey development. The primary goals of the focus groups were (1) to narrow the subset of deer-related attributes to those that are most relevant to hunters and public and (2) to develop a choice experiment framework that would enable participants to make well—informed, meaningful trade-offs among the attributes. The October focus group was held in Lansing, Michigan. Participants were randomly selected from the Greater Lansing Area phone book. The two February focus groups were held in Gaylord, MI. Participants were selected from computer generated 31 telephone listings for Gaylord and the surrounding areas. Initial contacts were made by telephone. Efforts to recruit a sample of the general public (for October and February groups) and a sample of deer hunters (for February group only) required that potential participants be screened for various socio-demographic characteristics. From the initial telephone contacts 13, 14, and 12 people agreed to participate in the October general public, February general public, and February hunter focus groups, respectively. Participants were sent a confirmation letter and map to the focus group location approximately 5 days before the discussion. The evening before the discussion participants were called to remind them of their commitment. All participants were paid $30. Attendance at the focus groups is reported in Table 1. Table 1. Focus Group Attendance Focus Group Attendance October General Public (Lansing) 6 (4 women, 2 men) February General Public (Gaylord) 7 (6 women, 1 man) February Hunter (Gaylord) 11 (all men) In the focus group discussions a set of deer-related attributes emerged that were relevant to the public and hunters, and thus were selected for the survey. Attributes for the public survey included the number of deer, herd health, residential property damage, agricultural damage, deer-vehicle collisions, and the effect of deer on the forest ecosystem. All attributes are specific to each of the three regions. Attributes for the hunter survey were identical to the attributes for the public with the inclusion of one additional attribute, the number of mature bucks. Focus group discussions revealed that 32 participants were able to consider the types of trade-offs that could be made among the attributes - for example, participants were able to think about what they may have to trade-off in some of the attributes in order to have an increase or decrease in the deer population. Attribute Description The next step in developing the survey was to determine how to describe each attribute in a manner that facilitated quantitative measurement and would allow levels of the attribute to be specified. Focus groups and informal discussions among eco-deer researchers were used to develop these descriptions. Focus groups revealed that while the meaning of some of the attributes, such as deer-vehicle collisions, was very straightforward, other attributes, such as deer browsing and the forest ecosystem, needed more detailed descriptions. Feedback from focus group participants, informal discussions with eco-deer researchers and deer management professionals, and literature reviews were used to the develop the attribute descriptions that were provided in the survey. Abbreviated attribute descriptions are provided below, along with the associated units of measure used to vary the attribute levels. Complete survey instruments for hunters and public are contained in Appendix A. Abbreviated Attribute Descriptions The Number of Deer refers to the current number of deer in the region (see graphic in survey instrument). Unit of measure is the percent change in the population relative to the status quo. 33 The Number of Mature Bucks refers to the number of bucks in the region that are at least two and a half years old with at least four antler points on one side. Unit of measure is the percent change in the population relative to the status quo. Herd Health refers to the physical well being of the herd. The health of the herd may be considered excellent, even though a few individual deer may be in poor health. When the number of deer in poor health increases, the health of the deer herd decreases. When a deer is in poor health it may have at least one of these characteristics: 0 smaller body size than expected - low reproductive success 0 disease ° lower chance of surviving long winters Unit of measure is the percent of deer in the region that have at least one characteristic of poor health. Residential Property Damage is caused by deer feeding on plants, trees, and shrubs that people plant in their yards. Unit of measure is the percent of residential properties in the region experiencing some damage due to deer. Deer Damage to Agriculture is caused by deer feeding on agricultural crops. Unit of measure is the deer damage per acre of cropland in region, expressed as $ per acre. Deer-Vehicle Collisions refers to the number of collisions between deer and vehicles during a given time period. Unit of measure is the number of reported annual deer-vehicle collisions. Deer and the Forest: In a forest, deer generally feed, or browse, on plants, shrubs, and tree seedlings. Over time, deer browsing may change the types of plants, trees, and animals that live in the forest. The extent of the changes depends on how much browsing occurs in the forest. In forest areas that experience deer browsing for 5 to 10 years: - some types of wildflowers may be eliminated while some grasses and ferns increase. 0 Some trees, like white cedar and maples, may be eliminated while trees like spruce and fir may increase. 0 The elimination of certain kinds of trees may cause losses in the commercial forest industry. 0 Some birds, like warblers, may be eliminated while cardinals and bluejays may increase. - In general, the habitat will support fewer kinds of plants and wildlife. 34 Besides the changes described above, many scientists believe that changes in the forest may also cause the forest to function differently in the long-term, with uncertain consequences. For example, the forest may be less able to adjust to events like fires and floods. Unit of measure is the percent of forest area in the region experiencing heavy deer browsing. In order to help “ground” respondents, information about the status quo of each attribute was also provided in the descriptions. For example, in the choice experiment levels of the attributes are specified to create alternative situations. Knowledge of the current level, or status quo, of the attribute would presumably help respondents in making an informed choice concerning their preferences for the attributes and attribute levels in their region. Because limited published data on the status quo for each attribute exists, particularly at the regional level, a variety of unpublished sources were consulted to develop estimates for each region. Status quo estimates for herd health were developed by consulting regional white- tailed deer biologists and veterinarians in the MDNR and asking them to provide estimates of the percent of deer with at least one characteristic of poor health. Eco-deer researchers from the population and habitat components were then asked to review these estimates and provide an estimate of their own. For the herd health attribute estimates from each source were consistent in the northwest and southwest regions. In the northeast region there was some disagreement among sources. Using input from Eco- deer researchers, the status quo estimate provided in the survey was slightly lower than the estimates provided by the MDNR. To develop status quo estimates for deer damage to residential property private landscaping and tree nursery companies in each region were contacted and asked to 35 provide an estimate of the percent of residential properties in the region that had experienced any deer damage. In each region between 4 and 7 companies were contacted. Efforts were made to contact a company in each of the counties that comprised a region; however in the northeast this was not possible, as no landscaping or tree nursery companies were listed in the phone book or on the intemet for two of the counties. Once companies were contacted discussions with managers and technicians lasted between half an hour to an hour each. With the exception of one company in the northwest region, all estimates within a region varied by less than 20%. Averages were calculated and used as the status quo estimate. One potential bias of the estimates results from most companies acknowledging that they were quite familiar with deer damage occurring in the surrounding area, but much less familiar with damage occurring throughout the entire county and the multiple-county region. To address this issue the attribute description emphasized that the estimate was for the entire region, and damage could be higher or lower in certain areas, depending on deer density in an area. To develop status quo estimates for deer browsing in the forest regional white- tailed deer biologists and management unit supervisors from the forestry division of the MDNR were consulted and asked to provide estimates of the percent of forest area in their region experiencing heavy deer browsing. Similar to the estimates provided for deer damage to residential property, the majority of management unit supervisors stressed that they were more familiar with certain areas within the region but less familiar with deer browsing throughout the entire region. Eco-deer researchers from the population and habitat components were also asked to provide an estimate. Variance was higher among 36 estimates of deer browsing than for any of the other status quo estimates. Generally, input from the Eco-deer researchers resulted in lowering the estimates provided from other sources. In the survey the attribute description emphasized that the estimate was for the entire region, and damage could be higher or lower in certain areas, depending on the deer density in an area. Unlike most of the attributes, some estimates for deer crop damage in Michigan are reported in the literature, though the variance of the estimates is rather large. In addition, the Wisconsin Department of Natural Resources (WDNR) maintains a database containing information on deer crop damage on a county level basis. Using the Wisconsin database and Wisconsin deer densities, regression models were developed at the Wisconsin county level for predicting deer crop damage. A number of different independent variables, all calculated per county, were included in different models, including acres of cropland, number of farms, average size of farms, percent of mixed hardwoods, percent of aspen, acres of corn, acres of soybean, acres of alfalfa, and overwinter deer density categories. For these models the dependent variable was the total amount of deer damage per county. Data for these variables was obtained from the WDNR database on deer crop damage and the Wisconsin Department of Agriculture. The only significant predictor of deer damage was the overwinter deer density category. The dependent variable was then recalculated to represent the deer damage per acre of cropland in the county. Independent variables that were significant in this model included the percent of land in the county devoted to agriculture and the overwinter deer density category. The deer crop damage model implies that deer crop damage increases 37 as the overwinter deer density increases and as the percent of land devoted to agriculture in a county decreases (Table 2). Table 2. Model of Deer Crop Damage in Wisconsin Independent Variable Coefficient (std. error, p-value) Overwinter deer density 0.28 (0.09, 0.00) Percent of land devoted to agriculture -0.01 (0.01, 0.00) Model Significance F=10.37, p=0.00, R-squared=0.22 Overwinter deer densities for the counties in Michigan comprising the survey regions and the percent of land in these counties devoted to agriculture were then plugged into the model to produce estimates of deer crop damage in the northwest, northeast, and southwest survey regions. These estimates were compared to other deer crop damage estimates reported in the literature (Campa et. a1 1997; Fritzell 1998) and weighted averages were computed for the final crop damage estimate. The Michigan State Police maintains county level data on the annual number of reported deer-vehicle collisions. The most recent available data (1999) were used for the status quo estimates of deer vehicle collisions. Estimates were not obtained for the number of deer or the number of mature bucks (hunters only) in the region due to the complexity of deer density estimates, thus the number of deer and number of mature bucks was referred to simply as the current number in the region. Figure 7 shows the status quo estimates for each attribute by region. 38 Northwest Northeast Southwest aar- Nurnber of Deer In Current Number Regron , : , _ . Percent of Deer in j _ , ‘ Region with at Least One ,‘ ft 25% , 2 ' - Characteristic of Poor " - ‘ . a : . f, ' Health -‘ Percent of Properties in Region Experiencing Some Deer Damage Deer Damage per Acre of l Cropland in Region =:-_. (Slacre) Annual Number of Deer- Vchicle Collisions in Region Percent of Forest Area in Region Experiencing Heavy Deer Browsing _ _ _ _ .. _ .. Figure 7. Status Quo Levels of Deer-related Attributes After reading the attribute descriptions and information about the status quo respondents were asked several questions concerning attitude towards and experience with each attribute. These questions were intended to break up large amounts of text and encourage respondents to read the attribute descriptions carefully. Additionally, they were intended to collect data that not only describe attitudes and experience but also may prove insightful in interpreting choice model results. Because it was expected that some respondents would disregard or disagree with the baseline estimates provided in the survey, a set of Likert scale agree/disagree statements was developed to help determine the credibility respondents gave to the estimates. Responses to these statements may 39 help in ad-hoc explanation of the choice experiment results. Experimental Design Choice experiment surveys require respondents to compare a number of different situations, referred to as a choice set, and choose which situation they prefer. Different situations are created by varying the “levels” of each of the attributes. Depending on the research needs, the “status quo” may be provided as one situation in the choice set, in which case the levels of each attribute would simply be the current status of the attribute. Experimental design plans are used to vary the attribute levels and form alternative situations, and designs generally require that the levels of each attribute vary independently. However, independent variation Mong all the deer-related attributes would have created some counterfactual situations, e. g. a situation with a large increase in deer numbers but a large decrease in deer-vehicle collisions, relative to the status quo. Feedback from focus groups revealed that when given choices that contained these types of counterfactual situations most respondents refused to make a choice, citing that the situations were not logical or possible to exist in real life. For this reason 3 types of choice sets were constructed in the survey: (I) attribute levels increase relative to the status-quo (2) attribute levels decrease relative to the status-quo (3) attribute levels increase or decrease but changes are marginal (small) relative to the status-quo. The public survey contained a fourth choice set which included the status-quo and one increasing, decreasing, and marginal alternatives. Because of the additional attribute ‘number of mature bucks’ in the hunter survey, which took an additional page for 40 description, the hunter survey did not contain the fourth choice set. In summary, respondents receiving the public survey version were asked to make four separate choices: a choice among the status quo and alternative situations in which attribute levels increase; a choice among the status quo and alternative situations in which attribute levels decrease; a choice among the status quo and altemative situations in which attributes either increase or decrease, but any changes are marginal; and a choice that contained all types of alternative situations. Respondents receiving the hunter survey version were asked to make only the first three choices. In each type of choice set each attribute could take one of two levels. For the increasing choice set each level represented a specified increase fi'om the status quo. For the decreasing choice set each level represented a decrease from the status quo, and for the marginal situation each level represented a small change - either an increase or a decrease - fiom the status quo. The magnitude of the change for each level was set primarily by consensus among eco-deer researchers with some input from individuals who participated in pre-testing. To create alternative situations in each type of choice set, levels of each attribute were varied according to a 6 x 2 (public) and 7 x 2 (hunters) main effects experimental design plan. Thus independent variation was maintained among the attributes within each type of choice set. The design plan assumes linear preferences and no interactions between any attributes. The design plan was generated using Minitab software, and resulted in 16 different survey versions per region (8 for hunters and 8 for public), for a total of 48 versions. 41 Pre-testing Regional specific survey instruments were developed for hunters and public and pre-tested in July and August 2001. To select pre-test participants names were randomly selected from computer generated telephone listings in each of the three survey regions. Participants were contacted first by telephone and asked several screening questions to determine selected socio-demographic characteristics and whether or not they hunt white- tailed deer in their region. From these initial telephone contacts 29, 28, and 25 people agreed to participate in the pre-tests in the northwest, northeast, and southwest region, with approximately one third of the participants in each region being hunters. The same protocol used to remind focus group participants of their commitment was followed for pre-test participants. All participants were paid $35. Attendance at pre-tests is reported in Table 3. Table 3. Pre-test Attendance Pre-test Location Attendance Marquette (northwest region) 21 (9 women, 12 men) Alpena (northeast region) 22 (6 women, 16 men) Battle Creek (southwest region) 19 (11 women, 8 men) Pre-tests revealed that when comparing alternative situations to the status quo some respondents did not compare among all attributes, but tended to make choices based solely on changes in deer numbers. In efforts to encourage respondents to compare among all the attributes, three alternative situations were provided in a choice set - a “status-quo” (the current situation of the attributes in a specific region) choice, and two 42 alternatives each with the same change in deer numbers but different changes among the other attributes. An example of a choice set for the Northwest region is shown in Figure 8. Current Situation Scenario A Scenario B Number of Deer In Current Number Regron ‘ : ., j » Percent of Deer in Region with at Least One Characteristic of Poor Health Percent of Properties in Region Experiencing Some Deer Damage Deer Damage per Acre of Cropland in Region (Slacre) Annual Number of Deer- Vehicle Collisions in Region Percent of Forest Area in Region Experiencing Heavy Deer Browsing Which do you prefer for your region? Current Cl Scenario A D Scenario B C] (Check one) Figure 8. Choice Set for Northwest Region Satisfaction and Issue Activity Models Questions were developed to enable satisfaction and issue activity models to be estimated. Data for the satisfaction models were collected by asking respondents to rate how satisfied they would be if the first alternative to the status quo, e.g. Scenario A in 43 Fig. 8, was the situation in their region. Satisfaction was rated using a 5 point Likert scale ranging from “Extremely satisfied” to “Extremely Dissatisfied.” The question was asked after each choice set (increasing, decreasing, and marginal). Because an experimental design was used to create the scenarios for the choice experiment, there were a total of 144 different “Scenario A’s” (48 survey versions times 3 types of choice sets). After respondents rated their satisfaction with “Scenario A” they were asked if they would engage in specific types of issue activity if “Scenario A” was the situation in their region. Specific types of issue activity included “contacting someone with authority to get the situation changed” or “taking steps myself to change the situation.” Two options were also available for respondents who wOuld not engage in issue activity if “Scenario A” were the situation in their region. Like the satisfaction ratings, 144 different situations were evaluated among all respondents concerning engagement in issue activity. Sample Population In an effort to sample two stakeholder groups, deer hunters and a more general group referred to as “public,” the survey sample was drawn from two separate sources, the Michigan Secretary of State (SOS) and a database of white-tailed deer hunters maintained by the MDNR. The SOS provided a random sample of 5,000 names (including addresses, age, and sex) for each region, and the MDNR provided a random sample of 2,000 hunter names per region. Both sources were asked to provide names of 44 individuals over 20 years of age. Sources were cross-checked to eliminate any duplicate names, though it was expected that a percentage of names on the SOS lists would also hunt deer. Random samples of 990 and 660 names were then drawn from SOS and MDNR lists for each of the three regions, for a total of 4950 names. The male to female ratio was 1:1 for the public sample (SOS) and 9:1 for the hunter sample (MDNR). The age distribution for both hunters and public was a slightly skewed normal distribution with the mean age being 47 for hunters and 41 for public. Survey Mailing Using guidelines outlined in Dillrnan (2000), respondents were contacted five times, unless a reply was received, in efforts to increase response rates. Examples of all contact letters are contained in Appendix B. The first contact consisted of a pre-notice letter sent in late August 2000. In each region the letter was sent to the 990 and 660 randomly selected names from the SOS and MDNR lists, for a total of 4,950 letters. The letter informed respondents that as a resident of the northwest/northeast/southwest region they had been selected to participate in a survey, which they would be receiving shortly. The first survey mailing occurred on Sept. 11, 2001 and included a cover letter, a return postage-paid survey booklet, and three first class stamps as a token incentive to complete the survey. There were 286 nondeliverable prenotice letters and 11 refusals, thus a total of 4,653 surveys were sent in the first survey mailing. A reminder postcard was sent on Sept. 25, 2001. The reminder postcard was sent to all individuals who had received a survey on the first mailing, thanking those 45 individuals who had already sent in their completed survey and reminding those who had not to please do so, as their input was important to deer management in their region. A second survey mailing occurred on Nov. 15, 2001 and included a new cover letter and a survey booklet. After removing nondeliverables and refusals from the first mailing, a total of 2,607 surveys were sent in the second mailing. A final survey mailing occurred on Dec. 6, 2001, and included a new cover letter and survey booklet. Nondeliverables and refusals from the second mailing were removed and a total of 1,895 surveys were sent in the third mailing. Table 4 shows the regional response rates and Table 5 shows the disposition of the MDNR and SOS samples and response rate. All surveys returned by Jan. 5, 2002 were included in the dataset for analysis. Table 4. Survey Response Rates by Ream Sample Group MDNR (Hunter) SOS (Public) Region NW NE SW NW NE SW Response Rate 65% 66% 64% 63% 62% 59% 46 Table 5. Disposition of the SOS and MDN R Samples Number . Non— Response Cum. t . ' Con “t Mailed Comp letrons Refusals deliverablesz Rate Response Prenotice MDNR 1,980 2 36 sos 2,970 NA 9 250 NA NA Total 4,950 l l 286 First Mailing MDNR 1,942 897 5 14 47% NA SOS 2,711 1,036 17 77 39% Total 4,653 1,933 22 91 42% Second $31: 1,026 244 3 9 24% 59% SOS 1,581 427 4 25 27% 56% Total 2,607 671 7 34 26% 57% Third Mailing NIDNR 770 126 0 3 l 6% 66% SOS 1,125 138 2 8 12% 62% Total 1,895 264 2 l l 14% 63% Full Surve Tm] y 14,105 2,868 42 422 63% 'Refusal is defined as a blank survey returned by a respondent asking to be taken off the mailing list or a returned unopened survey with “Refused” or “Return to Sender” written on the envelope. 2Nondeliverable is defined as a survey returned due to an incorrect address or a respondent who is no longer at the address. 47 CHAPTER 3. SURVEY ANALYSIS AND RESULTS Partial results of the choice experiment survey are presented in this chapter. The chapter begins with the background characteristics of respondents, then discusses respondents attitudes and experience with deer and deer-related attributes and differences between hunters and nonhunters. Attitude differences among respondents with different demographic characteristics are also examined. The chapter concludes with an assessment of issue activity and stakeholder satisfaction with different deer/extemality situations. Characteristics of Survey Respondents Data were obtained for several demographic variables. As can be seen in Table 6, in the public sample slightly more men returned the survey than did women. From the 2000 Census data for Michigan, the population consists of about 49% men and 51% women, thus the survey returns based on sex are slightly different than the general population; however, this may be due to the nature of the survey and the fact that more men hunt than do women. The mean age of respondents from the MDNR sample is 48 (median = 47) and 51 (median = 49) for respondents from the SOS sample. For convenience these samples will be referred to as hunters and public, unless otherwise noted. Median annual income category is $35,000—$49,000 for both groups. Both hunters and public have lived an average of 46 years in Michigan. The survey sample Compares well to the 2000 Census data for Michigan for income variables; however, the Inedian age of respondents is a little over 10 years older than the median age in Michigan 48 (36), according to the Census. The 2000 Census data shows that the median income in Michigan is $44, 667, which is within the median annual income category for survey respondents. Education level of survey respondents is similar to that of the general population. The 2000 Census data shows that about 30% of the population in Michigan has a high school degree, about 23% has some college, 7% has an associate’s degree, 14% has a bachelor’s degree, and 8% has a graduate degree. It should be noted that the characteristics of the sample population for this survey are not expected to match those of the 2000 Census, as the current sample is drawn from primarily rural regions in Michigan and includes only people over 21 years of age. On average, hunters have spent 35 years in their region, while public have spent 33 years. Approximately 6%, 4%, and 5% of the public respondents derived income from farming, forestry, and tourism, respectively. For hunters, 9%, 8%, and 4% of respondents derived income from farming, forestry, and tourism. The majority of both hunters and public owned property in their region, with the median property size for hunters of 18 acres (mean acreage is 67) and the median size for public of 5 acres (mean acreage is 81). Respondents stated that their property was used for a variety of purposes. For respondents in the public sample, uses included primary residence (68%), hunting (21%), wildlife viewing (21%), recreational residence (13%), farming (10%), and forest products (7%). Hunters stated primary residence (63%), hunting (42%), wildlife viewing (35%), recreational residence (20%), farming (15%) and forest products (13%) as Property uses. More detailed respondent characteristics, stratified by region, are provided in Table 6. To adjust for item non-response, percentages are based on the number of 49 respondents answering the question. Table 6. Characteristics of Survey Respondents Stakeholder Group Hunter (MDNR) Public(SOS) Region NW NE SW NW NE SW Sex* % Male 89 91 89 55 57 56 % Female 11 9 11 45 43 44 N= 415 448 410 516 600 484 Age* % 18-34 20 21 21 20 20 23 % 35-44 28 24 21 28 24 20 % 45-54 20 19 22 20 19 22 % 55-64 16 19 18 17 l9 l8 % 65-74 10 12 12 9 13 12 % Over 75 6 6 6 6 5 5 N= 415 448 410 516 600 484 Education % Less than high school 5 10 6 5 8 5 % High school graduate/GED 37 33 31 32 36 24 % Vocational/Trade 8 8 9 8 8 5 % Associates degree 18 21 24 20 20 25 % Some college 9 13 13 8 10 13 % College graduate 15 10 1 1 l6 9 18 % Graduate/Professional 8 5 6 11 9 10 N= 400 403 388 483 533 453 Item non-response (n) 15 38 22 33 67 31 Income % Less than $14,999 6 9 5 12 12 8 % $15,000 - $24,999 14 19 7 15 16 12 % $25,000 - $34,999 19 17 12 l6 19 13 % $35,000 - $49,999 23 23 20 26 23 17 % $50,000 - $74,000 22 20 32 17 17 28 % Over $75,00 16 12 24 14 13 22 N= 359 349 354 426 448 404 Item non-response (n) 56 92 56 90 152 80 Years in Michigan % 0 - 10 4 l 2 5 2 4 % 11 - 25 15 7 7 ll 6 8 % 26 - 40 23 25 29 26 21 29 % 41 - 60 39 39 45 36 40 42 % Over 60 19 28 17 22 31 17 N: 402 430 395 484 552 462 Item non-response (n) 13 11 15 32 48 22 50 Table 6 cont’d. Stakeholder Group Hunter (MDNR) Public(SOS) Region NW NE SW NW NE SW Property ownership in region % Yes 80 85 83 80 87 82 % No 20 15 17 20 13 18 N= 404 431 394 479 536 455 Item non-response (n) 11 10 16 37 64 29 Property Size (acres) % Less than 2 22 17 13 32 27 45 % 2 - 10 26 29 27 24 30 32 % 11- 50 27 28 34 25 20 15 % 51 - 100 13 10 10 10 9 4 % Over 100 12 16 16 9 l4 4 319 319 323 3w 416 * Se: and age were not asked in the survey but were provided in the SOS and MDNR sample frames, thus there is no item non-response for these characteristics. Data concerning hunting activity were also collected in the survey. While over 90% of respondents in the hunter sample stated that they hunt deer in the region, a relatively large percentage (40%) of the public sample also stated that they hunt in the region. This potential bias is addressed in the next section ‘Attitudes and Experience with Deer and Deer-related Attributes.’ The mean number of years hunting in a region was 26 and 25 for hunters in the MDNR sample and SOS sample, respectively. On average hunters in the MDNR sample hunted for 16, 17, and 17 days in the northwest, northeast, and southwest regions, respectively, in the year 2000. Hunters from the SOS sample hunted an average of 14, 14, and 13 days in the northwest, northeast, and southwest regions in 2000. More detailed respondent characteristics, stratified by region, are provided in Table 7. Percentages have been adjusted for item non-response. 51 Table 7. Hunting Activity of Survey Respondents Stakeholder Group Hunter (MDNR) Public (SOS) Region NW NE SW NW NE SW Hunt in region % Yes 92 95 91 49 50 30 % No 8 5 9 51 50 70 N: 401 434 396 469 535 454 Item non-response (n) 14 7 14 47 65 30 Years hunted in region % 0 - 10 17 18 27 21 16 34 % 11 — 25 36 28 35 32 34 34 % 26 — 40 27 29 30 31 33 24 % 41 — 60 19 23 8 13 17 8 % Over 60 l 2 <1 3 <1 <1 N= 376 414 363 241 274 137 Days hunted last year % 0 — 5 l8 19 20 29 31 34 % 6 — 15 49 48 46 49 42 34 % 16—30 22 23 22 15 17 24 % Over 30 11 10 12 7 10 8 N= 3&2 3mm Attitudes and Experience with Deer and Deer-related Attributes As stated previously, to familiarize respondents with the choice experiment attributes they were provided with attribute descriptions and then asked to complete several attitude and experience questions. While the primary purpose of these questions was to familiarize respondents with the attributes, the questions do provide data concerning perceptions of and experience with extemalities associated with deer. A general summary of responses is described below, with more detailed regional responses provided in tables. As described in the ‘Survey Development’ section, respondents were also asked to rate their level of agreement with several statements concerning the credibility of the attribute descriptions and baseline estimates. An analysis of these 52 responses shows that the majority of hunters and public accepted the baseline estimates and attribute descriptions provided in the survey. In the state of Michigan, a sample of “the general public” would generally contain a percentage of hunters, ranging between 3% and 20%, depending on the region (Peter Bull, pers. comm.) However, the percentage of respondents from the public (SOS) sample who stated that they hunt was considerably larger than the percentages given above. For this reason, and due to an expressed interest in comparing different segments of the general population, hunter responses from the SOS sample were analyzed with responses from the MDNR hunter sample. This segmentation created two distinct groups of hunters and nonhunters. The remaining analyses in this chapter have been conducted on responses from these two groups, as opposed to responses from the public (SOS) and hunters (MDNR). Deer Combining all regions 38% of hunters stated that they frequently saw deer in their own yard or neighborhood. The percentage was higher in the northwest (46%), lower in the northeast (28%) and about the same as all regions combined for the southwest. Twenty-eight percent of hunters stated that they frequently took a drive or a walk for the specific purpose of viewing deer, with only slight differences (< 2%) among the three regions. On average, relative to the year 2000, hunters in the northwest and southwest regions stated they would like about 25% more deer in future years, while hunters in the northeast region wanting about a 50% increase. Forty-five percent of hunters stated they 53 would be very concerned if the number of deer in their region decreased in future years by 20%. In the northeast region this percent was much higher (57%) while the northwest and southwest regions were lower, at approximately 38%. Similar to the hunter sample, combining all regions 33% of nonhunter respondents stated that they frequently see deer in their own yard or neighborhood, but only 10% of this sample stated that they frequently took a drive or a walk specifically to see deer. The percentage of nonhunters who see deer frequently was highest in the northwest (41%), with 32% in the northeast and 35% in the southwest. In all regions the percentage of nonhunters who took drives or walks specifically to see deer was between 8% and 11%. On average, relative to the year 2000, nonhunters stated that they would like about the same number of deer in future years,with minimal differences among regions. Thirty-eight percent of nonhunters stated that they would be very concerned if the number of deer in their region decreased in future years by 20%, with only marginal differences among regions. Table 8 provides detailed regional responses of hunters and nonhunters. 54 Table 8. Respondent Attitudes/Experience With Deer Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW Frequency of seeing deer in own yard or neighborhood last year % Never 10.4 13.5 13.6 15.5 17.9 25.1 % Rarely 14.5 26.4 11.7 16.7 20.7 18.7 % Sometimes 29.3 32.3 34.2 27.1 28.9 26.9 % Frequently 45.8 27.8 40.5 40.7 32.5 29.2 N= 598 659 486 258 280 342 Item non-response (n) 7 23 12 7 7 10 Frequency of taking drive/walk specifically to view deer % Never 12.6 12.7 11.5 32.4 31.6 48.5 % Rarely 19.7 19.6 17.8 25.7 27.4 20.7 % Sometimes 38.8 39.4 41.8 30.8 30.1 22.9 % Frequently 29.0 28.3 28.7 11.1 10.9 7.9 N= 580 639 478 253 266 328 Item non-response (n) 25 43 20 12 21 24 Number of deer desired in region in future years % none 0.2 0.0 0.0 0.8 1.4 2.3 % 75% less 1.7 0.7 1.2 3.8 2.9 3.5 % 50% less 9.6 6.6 5.3 17.3 12.3 15.4 % 25% less 9.6 2.8 8.4 11.5 8.0 10.2 % Same as last year 27.7 19.7 39.4 28.8 24.3 33.4 % 25% more 14.3 13.9 14.9 6.5 5.8 3.5 % 50% more 20.2 29.3 17.3 5.4 12.7 5.5 % 75% more 3.5 5.4 2.9 1.5 0.4 0.3 % Twice as many or more 7.3 16.6 6.9 1.9 5.1 0.9 % Unsure 4.0 4.6 2.7 15.8 21.0 16.9 % Don’t care 1.9 0.4 1.0 6.5 6.2 8.1 N: 593 670 490 260 276 344 Item non-response (n) 12 12 8 5 ll 8 Concern if the number of deer in the region decreased by 20% in future years % Very concerned 37.9 59.2 37.3 13.4 16.6 10.1 % Concerned 26.1 21.6 27.8 25.6 30.7 19.7 % Somewhat concerned 14.5 9.3 18.1 21.4 20.5 26.3 % Not concerned at all 21.56 10.0 16.8 39.7 32.2 43.6 N= 601 681 493 262 283 346 W (n) 4 1 5 3 4 55 The MDNR hunter sample was also asked about the number of mature bucks in their region — defined as the number of bucks two and a half years or older with at least four antler points on one side. As compared to last year (2000), hunters stated that they would like to have about 50% more mature bucks in their region in future years, with only slight differences among regions (<1%). Sixty-three percent of hunters stated that they would be very concerned if the number of mature bucks in their region decreased in future years by 20%. The percentage was higher in the northeast (69%) and lower in the southwest (57%). Table 9. Respondent Attitudes Toward Mature Bucks Stakeholder Group (MDIN‘Illtzlmple) N onhunter Region NW NE SW NW NE SW Number of deer desired in ‘ region in future years % none 0.3 0.5 0.3 % 75% less 0.3 0.2 0.3 % 50% less 1.4 0.7 1.7 % 25% less 0.8 0.2 1.7 % Same as last year 8.8 7.3 17.1 % 25% more 13.5 13.6 17.7 % 50% more 37.9 33.3 32.6 NA NA NA % 75% more 4.7 6.6 2.8 % Twice as many or more 28.8 30.8 21.9 % Unsure 1.9 4.4 3.1 % Don’t care 1.6 2.4 0.8 N= 403 433 397 Item non-response (n) 12 8 l3 Concern if the number of mature bucks in the region decreased by 20% in future years % Very concerned 64.8 70.8 58.5 % Concerned 24.6 18.6 26.6 % Somewhat concerned 8.2 6.3 10.9 NA NA NA % Not concerned at all 2.5 4.3 3.9 N= 406 435 398 W (n) 9 L__12 56 Herd Health In the attribute description of herd health four characteristics of a deer in poor health — smaller body size than expected, low reproductive success, disease, and low chance of surviving long winters — were provided along with an estimate of the percent of deer in the region that have at least one of these characteristics. Fifty-eight percent of hunters indicated that they would be very concerned if the percent of deer in their region with at least one characteristic of poor health increased (relative to the status quo) in future years. There were slight differences (<4%) among regions. Respondents were also asked to rate the health of the deer herd in their region based on their own opinion and experience. Twenty-nine percent of hunters rated the health of the herd in their region as excellent. In the northwest and northeast regions only 23% of hunters thought that herd health was excellent, while in the southwest the percentage was much higher at 44% . Thirty-eight percent of nonhunters said they would be very concerned if the percent of deer with at least one characteristic of poor health increased in future years. This percent was highest in the northeast region (51%), with slight differences between the northwest (34%) and the southwest (31%). In contrast to the hunter sample, only 11% of nonhunters rated the health of the herd in their region as excellent. The percent was highest in the southwest (17%), and lower in the northwest (10%) and northeast (6%). The majority of nonhunters in all regions rated the health of the herd as “good.” Table 10 shows detailed, regional responses. 57 Table 10. Respondent Attitudes Toward Herd Health Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW Concern if percent of deer in region with at least one characteristic of poor health increased by 10% in future years % Very concerned 56.3 56.0 60.7 34.1 51.8 31.7 % Concerned 33.7 32.0 31.5 40.9 34.0 40.1 % Somewhat concerned 7.7 8.8 6.9 19.3 9.9 19.3 % Not concerned at all 2.4 3.1 1.0 5.3 4.3 8.9 N= 597 671 496 264 282 347 Item non-response (n) 8 11 2 1 5 5 Rating of herd health in region % Excellent 23.0 23.2 44.6 10.0 5.7 17.3 % Good 38.8 38.2 22.4 45.2 27.2 30.3 % Fair 31.8 26.9 27.8 19.5 23.0 11.5 % Poor 3.2 4.2 1.6 2.7 6.4 <1.0 % Unsure 3.2 7.4 3.6 22.6 37.8 40.3 N: 595 672 496 261 283 347 Item non-response (n) 10 10 2 4 4 5 Deer Damage to Residential Property Combining all regions 38% of hunters had experienced residential property damage from deer. In the northwest region the percentage was higher (50%), while in the northeast and southwest approximately 30% of hunters experienced residential damage from deer. In the northwest deer damage caused 43% of respondents to change the types of plants in their yards. This percentage was lower in the northeast (30%) and southwest (20%). In the northwest about 16% of hunters felt that it was very important to substantially decrease the current level of property damage, while in the northeast and southwest this percentage was approximately 8%. 58 Thirty-five percent of nonhunter respondents had experienced residential property damage from deer. Similar to the hunter sample, the percentage was higher in the northwest (45%) than in the northeast or southwest (30%). In the northwest and northeast the damage caused 50% of nonhunters to change the plants in their yards, while 30% of nonhunters in the southwest changed the types of plants in their yard. In the northwest and northeast less than 20% of nonhunters felt that it was very important to decrease the current level of property damage, while in the southwest this percentage was about 28%. Table 11 shows detailed regional responses. Table 11. Respondent Attitudes/Experience with Deer Damage to Residential Property Damage Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW Experienced residential property from deer % Yes 50.6 31.0 32.0 45.4 32.7 28.7 % No 49.4 69.0 68.0 54.6 67.3 71.3 N= 593 661 490 262 278 342 Item non-response (n) 12 21 8 3 9 10 Damage resulted in changing types of plants in yard % Yes 43.9 30.8 20.7 51.3 47.3 29.3 % No 56.0 68.3 79.3 47.1 51.6 68.7 % Unsure <1.0 <1.0 0 0.8 1.1 2.0 N= 303 224 169 1 19 93 99 Importance of decreasing current level of damage by 10% % Very important 16.2 7.8 8.7 16.8 11.8 13.0 % Important 25.4 21.7 26.2 31.3 28.0 27.1 % Somewhat important 35.0 32.3 34.1 34.7 35.8 35.2 % Not important at all 23.4 38.2 30.9 17.2 24.4 24.8 N= 594 668 492 262 279 347 Item non-response (n) 11 14 6 3 8 5 59 Deer Damage to Agricultural Crops Fifty—five percent of hunters had either experienced themselves or knew someone in the region who had experienced deer damage to agriculture. On a regional level this percent was much higher in the southwest (70%) than in the northwest and northeast (50%); however in all regions the damage was most frequently categorized as relatively unimportant economic loss. Less than 20% of hunters in each region stated that they would be very concerned with a large increase in the amount of deer damage to agriculture. In contrast to the hunter respondents, between 36% and 40% of the nonhunter respondents experienced or knew someone in the region who experienced deer damage to agriculture. In all regions the damage was most frequently categorized as a moderately important economic loss. Approximately 25% of nonhunters in each region stated that they would be very concerned with a large increase in the amount of deer damage to agriculture. Table 12 shows detailed regional responses. 60 Table 12. Respondent Attitudes/Experience with Deer Damage to Agriculture ‘ Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW Experienced or know someone in region who experienced deer damage to agriculture % Yes 47.0 51.4 71.2 39.8 37.4 36.0 % No 53.0 48.6 28.2 60.2 62.6 64.0 N= 592 666 486 261 273 342 Item non-response (n) 13 16 12 4 14 10 Economic loss from damage: % Major loss 6.4 4 4.3 9.6 5.9 6.5 % Moderately important loss 36.5 27.1 38.5 41.3 30.4 37.4 % Relatively unimportant loss 37.2 53.0 43.0 29.8 30.4 30.9 % Unsure 19.9 15.6 14.2 19.2 31.4 23.6 N= 282 247 351 104 102 123 Concern with large increase in deer damage to agriculture % Very concerned 21.5 12.9 18.7 25.1 23.0 24.4 % Concerned 32.4 27.1 35.2 36.5 31.7 34.9 % Somewhat concerned 33.4 36.2 32.7 28.9 33.5 30.2 % Not concerned at all 12.7 23.8 13.4 9.5 11.9 10.9 N= 599 668 492 263 278 348 Item non-response (n) 6 l4 6 2 9 4 Deer-vehicle Collisions Seventy-one percent of hunters were aware that the majority of deer-vehicle collisions take place on local or rural roads rather than highways and freeways. Eighty- one percent of hunters stated that they or someone in their immediate family had been involved in a deer-vehicle collision, with only marginal differences among regions. In the northwest and southwest regions approximately 30% of hunters said it would be very important to decrease the annual number of collisions, while only 20% of hunters in the northeast stated that this would be very important. 61 Sixty-seven percent of the nonhunters were aware that most deer-vehicle collisions occurred on local or rural roads, and approximately 75% of nonhunters stated that they or someone in their immediate family had been involved in a deer-vehicle collision. Importance placed on decreasing the number of collisions was highest in the northwest, and lower in the northeast and southwest. Table 13 shows detailed regional responses. Table 13. Respondent Attitudes/Experience with Deer-vehicle Collisions Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW Yourself or someone in immediate family involved in deer-vehicle collision % Yes 83.9 78.6 82.6 77.4 77.2 74.1 % No 15.6 22.1 » 17.2 21.8 21.4 24.1 N= 596 673 489 261 281 344 Item non-response (n) 9 9 9 4 6 8 Importance of decreasing annual deer-vehicle collisions by 10% in future years % Very important 32.4 20.0 28.2 43.8 38.1 46.7 % Important 29.7 27.1 33.7 35.8 34.5 30.7 % Somewhat important 29.0 30.0 26.5 15.4 22.1 17.7 % Not important at all 8.9 22.9 11.6 5.0 5.3 4.9 N= 596 669 490 260 281 345 Item non-response (n) 9 13 8 5 6 7 Deer Browsing and Forest Ecosystems Thirty percent of hunters stated that they had seen effects of deer browsing in forested areas. The percentage was higher in the northwest (42%) than in the northeast (26%) and southwest (22%). Hunters in all regions were more concerned with the effects 62 deer browsing has on other types of wildlife and the forest ecosystem and less concerned about losses to the forest industry caused by deer browsing. Twenty-six percent of nonhunters stated that they had seen effects of deer browsing. Similar to hunters, a higher percentage of nonhunters in the northwest (40%) had seen effects of deer browsing than in the northeast (22%) or southwest (20%). Nonhunters in all regions were most concerned with the effects deer browsing on wildlife and the forest ecosystem and least concerned about losses to the forest industry caused by deer browsing. Table 14 shows detailed regional responses. Table 14. Respondent Attitudes/Experience with Deer Browsing in the Forest Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW Have seen examples of deer browsing in the forest % Yes 42.5 26.1 22.5 37.9 22.3 19.5 % No 52.8 68.6 71.4 48.3 62.8 64.4 % Unsure 2.3 3.5 4.1 13.8 14.8 16.0 N= 598 678 493 261 282 343 Item non-response (n) 7 4 5 4 5 9 Concern if percent of forest area experiencing heavy deer browsing increases by 10% in future years % Very concerned 21.8 11.7 13.4 23.1 15.9 22.6 % Concerned 32.9 28.1 32.0 36.4 33.2 34.0 % Somewhat concerned 33.2 30.8 36.3 28.0 33.6 30.9 % Not concerned at all 12.1 29.4 18.3 12.5 17.3 12.6 N= 602 676 493 264 283 350 Item non-response (n) 3 6 5 1 4 2 63 Table 14 cont’d Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW Concerned about browsing because of changes to wildlife % Strongly agree 10.8 5. 0 7.1 13. 9 10.4 14.2 % Agree 33.3 20. 4 36. 9 44.8 37.4 40.8 % Neither agree/disagree 34.3 34.7 32.7 30. 5 34.5 35.8 % Disagree 16. 5 26.7 15. 8 8.5 14.4 7.2 % Strongly disagree 5. 0 13.2 7. 5 2. 3 3. 2 2. 0 N= 600 666 493 259 27 8 346 Item non-response (n) 5 16 5 6 9 6 Concerned about browsing because of long-term effects to forest ecosystem % Strongly agree 14.8 5. 7 9. 8 19.4 10.7 19.6 % Agree 36.9 21.4 39.2 42.2 39.9 40.1 % Neither agree/disagree 27.8 30.5 24.8 27.1 31.7 31.7 % Disagree 16.1 29.3 18. 9 9.3 13.93 6.6 °o Strongly disagree 4. 3 13.0 7. 3 1.9 3. 9 2. 0 N= 601 668 492 258 281 347 Item non-response (n) 4 14 6 7 6 5 Concerned about browsing because of losses to commercial forestry % Strongly agree 9. 7 2. 6 3. 9 8. 5 5. 8 8. 4 % Agree 26.0 12.5 18.4 34.5 231 25.5 % Neither agree/disagree 34.4 30.5 36.2 32.6 43.7 41.4 % Disagree 20. 0 30.9 24.3 19.0 17. 7 18.0 % Strongly disagree 9. 8 23.6 17.2 5. 4 9. 7 6. 7 N= 599 666 489 258 277 345 Item non-response (n) 6 16 9 7 10 7 Belief Statements Near the end of the survey, respondents were asked to express their level of agreement with a set of “belief statements.” In part, these belief statements were designed to aid in determining whether respondents found the extemality information presented in the survey credible, with the expectation that a low percentage of 64 respondents would express disagreement with statements C, E, G, and J, and a high percentage of respondents would express disagreement with statement F in Table 15. Results in Table 15 support the expectation, but there are respondents who do not accept the baseline estimates. Other belief statements were included in the survey because of their potential as explanatory variables in various types of models, as well as a means for gathering additional data that may be usefirl to deer management. Table 15. Respondent Agreement with Belief Statements Stakeholder Group Hunter Public Region NW NE SW NW NE SW A. When deer-vehicle collisions increase in an area, car insurance rates usually increase % Strongly Agree/Agree 56 59 63 60 55 70 % Neither Agree nor Disagree 29 26 24 27 27 20 % Strongly Disagree/Disagree 15 15 l3 l3 18 10 N= 401 423 392 475 540 456 Item non-response (n) 14 18 18 41 60 28 B. People can avoid most deer- vehicle collisions % Strongly Agree/Agree 47 56 49 35 43 28 % Neither Agree nor Disagree 19 15 l9 14 17 20 % Strongly Disagree/Disagree 34 29 32 51 40 52 N= 403 424 392 475 542 455 Item non-response (n) 12 1 7 1 8 41 58 29 C. Deer browsing can prevent some types of trees from reproducing in the forest % Strongly Agree/Agree 68 53 59 69 54 69 % Neither Agree nor Disagree 20 27 28 21 27 24 % Strongly Disagree/Disagree 12 20 13 10 19 7 N= 399 421 385 476 538 455 Item non-response (n) 16 20 25 40 62 29 65 Table 15 cont’d. Stakeholder Group Hunter Public Region NW NE SW NW NE SW D. The effects of deer browsing in the forest are significant in the nw/ne/sw region of Michigan % Strongly Agree/Agree 32 19 22 37 32 29 % Neither Agree nor Disagree 38 35 40 42 40 53 % Strongly Disagree/Disagree 30 46 38 21 28 19 N= 402 423 392 473 539 453 Item non-response (n) 13 1 8 18 43 61 3 l E. When comparing scenarios I accepted that 30% of the forest areas in the nw/ne/sw region of Michigan experience heavy deer browsing % Strongly Agree/Agree 42 23 41 45 33 45 % Neither Agree nor Disagree 32 33 30 30 35 36 % Strongly Disagree/Disagree 26 44 29 25 32 19 N= 401 419 386 470 530 453 Item non-response (n) 14 22 24 46 70 31 F. A deer with at least one characteristic of poor health has a disease % Strongly Agree/Agree 21 19 20 21 24 28 % Neither Agree nor Disagree 35 29 38 38 32 37 % Strongly Disagree/Disagree 43 52 42 41 44 35 N= 403 420 391 473 535 451 Item non-response (n) 12 21 19 43 65 33 G. When comparing scenarios I accepted that 35%/30%/5% of deer in the nw/ne/sw region have at least one characteristic of poor health % Strongly Agree/Agree 44 26 55 48 40 64 % Neither Agree nor Disagree 33 36 30 34 30 28 % Strongly Disagree/Disagree 23 38 15 18 30 8 N= 401 420 389 471 534 451 Item non-response (n) 14 21 21 45 66 33 Table 15 cont’d. Stakeholder Group Hunter Public Region NW NE SW NW NE SW H. There is a significant amount of deer damage to agricultural crops in the nw/ne/sw region of Michigan % Strongly Agree/Agree 42 25 49 45 35 49 % Neither Agree nor Disagree 34 37 31 33 36 36 % Strongly Disagree/Disagree 24 38 20 22 29 15 = 403 425 392 474 537 454 Item non-response (n) 12 16 18 42 63 30 I. There is a significant amount of deer damage to residential properties in the nw/ne/sw region of Michigan % Strongly Agree/Agree 40 14 21 41 22 31 % Neither Agree nor Disagree 28 30 34 28 34 40 % Strongly Disagree/Disagree 42 56 45 31 44 29 N= 403 425 391 475 537 455 Item non-response (n) 12 16 19 41 63 29 J. When comparing scenarios I accepted that 30%/20%/20% of residential properties in the nw/ne/sw region of Michigan experience some deer damage % Strongly Agree/Agree 60 41 49 55 49 57 % Neither Agree nor Disagree 20 29 29 27 28 30 % Strongly Disagree/Disagree 20 30 12 28 23 13 N= 402 423 391 471 534 452 Item non-response (n) 13 18 19 45 66 32 K. The number of bucks can increase in a deer herd even if the total number of deer stays the same % Strongly Agree/Agree 54 47 55 36 34 34 % Neither Agree nor Disagree 20 27 22 39 43 49 % Strongly Disagree/Disagree 26 26 21 25 23 17 = 401 424 391 472 531 452 Item non-response (n) 14 17 19 44 69 32 67 Respondent Segmentation Analysis Segmentation Based on Respondent Characteristics Data on selected respondent characteristics were collected in the survey and cross-tabulated or correlated with attitudes toward deer and deer-related attributes in order to help explain preference heterogeneity. Selected characteristics are listed below: 0 hunter or nonhunter ° number of years living in the northwest/northeast/southwest region 0 type of residential area, e.g. rural, semi-rural, town/city - number of years hunting deer in northwest/northeast/southwest region - income category - education category - age Cross tabulation analyses were conducted to examine differences between hunters and nonhunters. All other variables contained at least 4 ordered categories, with the exception of “type of residential area” which contained 3. Assuming ordinal data, Pearson correlation coefficients were computed. Significance level was set at p s 0.05 for both chi-square and correlation statistics. Significant relationships are described below. Test statistics and significance levels are presented in Tables 16 and 17. All significant correlations have a Pearson’s correlation coefficient ranging fiom 0.07 to 0.12. In the remainder of this chapter, a correlation of this magnitude will be referred to as weakly significiant. Measurement scales for questions asking respondents about their ‘concern for an extemality increase’ or the ‘importance placed on decreasing an 68 extemality’ ranged from 1= “very concerned” or “very important” to 4 = “not concerned at all” or “not important at all,” thus the coefficient signs in Tables 16 and 17 may appear the reverse of what would be expected. Number of Deer Desired for the Region In all regions hunters and nonhunters differ significantly in the number of deer they would like to have, with hunters generally preferring more deer than the current number and nonhunters preferring about the same as the current number. In the northwest region income was weakly correlated with the number of deer desired for the region, suggesting that respondents with higher incomes tend to prefer less deer for the region. In the northeast region the number of years hunting was weakly correlated with the number of deer desired for the region, suggesting that hunters with more years of hunting tend to prefer less deer. In the southwest region the type of residential area (city/town, scattered neighborhoods outside the town, rural area) was weakly correlated with preferences for deer numbers, with respondents living in rural areas tending to prefer less deer. Concern for Deer Numbers In all regions hunters were also more concerned with potential decreases in the number of deer than were nonhunters, although 38% of the public stated they would be very concerned if the number of deer decreased by 20% (relative to the current number) in future years. In all regions education level was weakly correlated with concern for a 69 decrease in deer numbers, suggesting that respondents with higher education levels tend to be less concerned with a 20% decrease in the number of deer in the northwest region than do respondents with less education. In the northeast respondents with more years hunting in the region tend to be more concerned with a 20% decrease in deer numbers. Number of Mature Bucks Desired for the Region Analyses concerning mature bucks were conducted on hunters from the MDNR sample only, as hunters from the SOS sample did not receive a survey version that contained questions about bucks. The only significant correlations that exist concerning the number of mature bucks desired for the region occur in the northwest region, where hunters with more years of hunting tend to desire-fewer mature bucks than do hunters with less years of hunting. Concern for the Number of Mature Bucks In the southwest region hunters with more years of hunting tend to be more concerned with a 20% decrease in the number of mature bucks. There were no significant correlations in the northwest or northeast. Concern for Herd Health In all regions hunter and nonhunter ratings were significantly different, with hunters generally rating the health of the herd higher than nonhunters. In the northeast respondents living in more rural areas tended to rate health higher, whereas respondents 70 with higher levels of education and older respondents tended to rate the health of the herd lower. hr the southwest respondents with more years living in the region tended to rate the health of the herd higher, as did respondents with more years of hunting and respondents living in more rural areas. In the southwest respondents with higher levels of education tended to give lower ratings to herd health. Concern for Deer Damage to Residential Property Significant difference exists between hunters and nonhunters in the northeast region in the importance placed on decreasing the amount of deer damage to residential property, where hunters felt it was less important to decrease the amount of damage than did nonhunters. However, for both groups of respondents less than 15% felt that it was very important to decrease the amount of deer damage to residential property. In all regions the number of years hunting was weakly correlated with importance of reducing residential damage, with hunters with more years hunting tending to place more importance on decreasing the amount of deer damage to residential property. In the northeast and southwest regions respondents with higher incomes tended to place less importance on decreasing the amount of residential property damage, and in the northeast older respondents tended to place more importance on decreasing the amount of deer damage to residential property. 71 Concern for Deer Damage to Agriculture Hunters and nonhunters differed significantly in the northeast region in their concern for any increases in deer damage to agriculture, with hunters tending to be less concerned than nonhunters with an increase. In the northwest and northeast regions respondents with more years living in the region tended to be more concerned with any increases in deer damage to agriculture. Concern for Deer- Vehicle Collisions In all regions there were significant differences between hunters and nonhunters in the importance they placed on reducing the annual number of deer-vehicle collisions, with hunters generally placing less importance on reductions than did nonhunters. In the northeast region respondents with higher incomes tended to place less importance on reducing the annual number of deer-vehicle collisions. Concern for Deer Browsing in the Forest Significant differences exist between hunters and nonhunters in the northeast and southwest in their concern for increases in the amount of heavy deer browsing in forests, with nonhunters tending to be more concerned with an increase than hunters. In the northwest region respondents with more education tended to be more concerned with an increase in the amount of deer browsing in the forest. In all regions there were Significant differences between hunters and nonhunters in their concern about the effects deer browsing has on wildlife and concern about the long- 72 term effects of browsing on the forest ecosystem. In the northeast and southwest significant differences exist between hunters and nonhunters in their concern about the effects deer browsing has on commercial forestry. In all of the above cases hunters tended to be less concemed than nonhunters. In the northwest region respondents with more education tended to be more concerned about the effects deer browsing has on other wildlife and the long-term effects on the forest ecosystem, while respondents with more years living in the region tended to be more concerned about the effects of deer browsing on commercial forestry. In the northeast respondents living in more rural areas tended to be less concerned about the long term effects of deer browsing on the forest ecosystem and the effects on commercial forestry. Also in the northeast respondents with more education tended to be more concerned about the long-term effects deer browsing has on the forest ecosystem, and respondents with more years living in the region tend to be more concerned about the effects of deer browsing on commercial forestry. In the southwest region respondents with higher incomes tend to be less concerned about the effects of deer browsing on commercial forestry. 73 Table 16. Respondent Characteristics Correlated with Deer and Buck Numbers Attitude Northwest Northeast Southwest Who tends to prefer more deer? Who tends to be more concerned with deer number Hunters x2 = 110.7 (0.00) Resp. with less income 7" =-0.10 (0.00) Hunters )8 = 66.7 (0.00) Hunters x2 = 200.4 (0.00) Resp. with fewer years of hunting x2 =-0.10 (0.01) Hunters x2 = 170.8 (0.00) Resp. with more Hunters )8 = 245.5 (0.00) Resp. living in less rural areas x2 =-0.8 (0.01) Hunters x2 = 130.8 (0.00) decreases? , years of huntrng x2 =-0.09 (0.02) Resp. with less Resp. with less Resp. with less education education education )8 = 0.09 (0.01) x2 = 0.11 (0.00) x2 = 0.12 (0.00) Who tends to Resp. with fewer prefer more Years Of huntmg mature bucks? XZ ='0-14 (0°01) Who tends to be Resp. with more years of hunting x2 =-0.11 (0.04) more concerned with buck number decreases? 74 Table 17. Respondent Characteristics Correlated with Deer-related Attributes Attitude Northwest Northeast Southwest Who tends to be Hunters more concerned with x2 = 52.1 (0.00) increases in the percent of deer in poor health? Who tends to place more importance on decreasing the amount of deer damage to residential property? Who tends to be more concerned with increases in deer damage to agriculture? Who tends to place more importance on decreasing annual deer-vehicle collisions? Who tends to be more concerned with increases in the amount of deer browsing in the forest? Resp. with more years hunting x2 =-0.08 (0.05) Resp. with more years living in region x2 =-0.09 (0.01) Nonhunters x2 = 37.2 (0.00) Resp. with more education x2 =-0.07 (0.03) Nonhunters x" = 22.3 (0.01) Resp. with less income )6 = 0.07 (0.05) Resp. with more years hunting )8 =-0.07 (0.05) Older resp. x2 =-0.08 (0.01) Nonhunters x2 = 35.9 (0.00) Resp. with more years living in region x2 =-0.08 (0.01) Nonhunters x2 = 74.0 (0.00) Nonhunters x2 = 21.0 (0.01) Hunters x2 = 98.5 (0.00) Resp. with less income 3" = 0.11 (0.00) Resp. with more years hunting )8 =-0.11 (0.01) Nonhunters x2 = 51.4 (0.00) Nonhunters x’ = 24.0 (0.02) 75 Segmentation Based on Experience with Extemalities A separate analysis was conducted to determine whether respondents who had experience with the negative extemalities associated with the deer population tended to prefer less deer than those respondents who did not have extemality experience. Preferences for deer were obtained from the question asking respondents to state the number of deer they desired for their region in future years, using a scale relative to the current number of deer in the region (see Appendix A). In all regions cross tabulation analyses suggest that respondents who have experience with deer damage to residential property, deer damage to agriculture, and deer browsing tend to prefer less deer for their region than those without these extemality experiences (all )8 statistics were significant at p < 0.01). There were no significant differences between respondents with and without deer-vehicle collision experience in the number of deer preferred for the region; however, this is most likely due to the high percentage of respondents in each region with deer- vehicle collision experience (78% or more in all regions). Correlation analyses were conducted to examine whether respondents’ rating of herd health is significantly correlated to the number of deer desired for the region. The analysis suggests that respondents in the southwest who gave lower health ratings tend to prefer less deer for the region r = 0.15, p < 0.01). There were no significant correlations between health rating and preference for deer populations in the northwest and northeast regions. 76 Stakeholder Satisfaction and Issue Activity Related to Alternative Deer Scenarios Satisfaction Model Estimation Recall that respondents were asked to rate their satisfaction level of the alternative deer scenarios presented in the survey using a five point Likert scale (see Figure 9), in addition to choosing which scenario they would prefer for their region. Because an experimental design was used to vary the levels of the deer related attributes, there were a total of 144 different “Scenario A’s” (48 survey versions times 3 types of choice sets, e. g. increasing, decreasing, or marginal - see Survey Development), and respondents indicated their satisfaction level for these different scenarios. Item non-response for the satisfaction ratings ranged between 4% and 9% for hunters and 9% and 10% for nonhunters, depending on the region and on the type of alternative situation. Generally item non- response was highest when respondents were asked to rate their satisfaction with a marginal situation, e. g. a situation where changes in the attribute levels are marginal relative to the status quo, though this was also the last situation respondent were asked to rate and respondent fatigue may have occurred at this point. The dataset for the satisfaction models was formed by combining ratings for increasing, decreasing, and marginal situations, thus allowing a wider range of variation among the attributes, which served as the independent variables for satisfaction models. 77 Current Situation Scenario A Scenario B Number ofDeerrn Regron 'C- . N ”i Percent of Deer in Region with at Least One Characteristic of Poor Health Percent of Properties in Region Experiencing Some Deer Damage Deer Damage per Acre of Cropland in Region ($/acre) Annual Number of Deer- Vehicle Collisions in Region Percent of Forest Area in Region Experiencing Heavy Deer Browsing Which do you prefer for your region? (Check one) Current CI Scenario A D Scenario B D How satisfied would you be if Scenario A were the situation in your region? (Check only one) C] Extremely CI Satisfied 0 Neither Satisfied D Dissatisfied U Extremely Satisfied Nor dissatisfied Dissatisfied Figure 9. Respondent Rating of Satisfaction Satisfaction models were estimated using LIMDEP 7.0. The models assume that respondent satisfaction level with a deer situation depends on the levels of the deer- related attributes. Respondent characteristics were also included to determine their effect on satisfaction level. A systematic procedure, outlined below, was used for estimation. 1. General model: OLS model estimated to determine the effects of the deer-related attributes on Satisfaction. 78 2. A second OLS model was estimated which included deer-related attributes and the respondent characteristics education, age, years living in the region, type of residence area (rural, semi-rural, city/town), and years hunting in the region (hunters only) . 3. A third OLS model was estimated which included deer-related attributes and the significant respondent characteristics from (2) above. Respondent characteristics were retained if they were significant and improved the R-square value. Ifno respondent characteristics were significant, the General model was retained as the Satisfaction model. All satisfaction models were significant at p< 0.05. Although the hunter and nonhunter models all contained significant attributes, the R—squared value on all models was low - ranging from 0.02 to 0.12 - thus their usefirlness in explaining the variance in satisfaction ratings is quite limited. In an effort to improve model performance, ordered probit models were estimated from the satisfaction dataset; however, these models also performed poorly. One reason for the poor model performances may be the limited time allocated to the satisfaction questions during the qualitative research phase. During this phase most of the qualitative research was devoted to the choice experiment, as choice tasks are generally complex and ideally require the respondent to understand what the attributes of a good are as well as the trade-offs they are making among the attributes by choosing one scenario over another. Time and resource constraints dictated that the primary focus of the survey development phase was on determining and defining the choice experiment attributes and construction and presentation of the alternative 79 scenarios. Clearly additional qualitative work related to stakeholder satisfaction questions would have been useful. Results of the satisfaction models estimated using OLS regression are presented in Table 18. Regional Results of Stakeholder Satisfaction Northwest Regj on The only significant attribute for hunters in the northwest region was the number of mature bucks in the region. This suggests that as the number of mature bucks in the region increases hunter satisfaction increases. Significant attributes of nonhunter satisfaction included herd health and education, suggesting that nonhunter satisfaction decreases as the percent of deer with at least one characteristic of poor health increases and as education increases. WM Significant attributes of hunter satisfaction in the northeast region included the number of deer, the number of mature bucks, and deer vehicle collisions. Results suggest that hunter satisfaction increases when the number of deer and number of mature bucks increase, and satisfaction decreases when the number of annual deer vehicle collisions increases. 80 Table 18. Regression Models of Stakeholder Satisfaction with Deer and Deer-related Attributes Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW R—square 0.02 0.08 0.05 0.12 0.08 0.12 Deer Number coefficient (std. error) 0.4697 2.2080 1.6326 0.9244 1.3883 1.6147 (0.5786) (0.5946) (0.5894) (0.6429) (0.6790) (0.5201) 2.3434 2.0407 -1 .5254 BuckNumb" (1.2266) (1.2111) (1.3225) 0.0048 -0.0076 -0.0351 -0.0288 -0.1622 -0.0293 He'd Health (0.0132) (0.0133) (0.0284) (0.0151) (0.0146) (0.0257) De" Damage“ 0.0103 0.0092 0.0175 0.0102 -0.0090 -0.0142 Residential 0.0132 0.0123 0.0142 0.0149 0.0141 0.0129 Property ( ) ( ) ( ) ( ) ( ) ( ) Deer Damage to -0.0140 -0. 1705 0.1635 -0.0833 -0.2630 0.0407 Agriculture (0.0722) (0.1264) (0.1362) (0.0798) (0.1399) (0.1229) Deer-vehicle 0.0001 -0.0005 -0.0003 -0.0002 -0.0002 -0.0006 Collisions (0.0001) (0.0002) (0.0001) (0.0002) (0.0002) (0.0001) Deer Browsing in -0.0203 -0.0021 -0.0083 -0.0107 —0.0245 —0.0070 the Forest (0.0134) (0.0132) (0.0135) (0.0151) (0.0147) (0.0124) . -0.0659 Education (0.0189) A 0.0053 .0.0042 ge . . (0.0022) - ,_ (0.0021) Significant attributes of nonhunter satisfaction included the number of deer in the region, deer damage to agriculture, and age. These results suggest that nonhunter 81 satisfaction increases when the number of deer increases, and decreases when the amount of deer damage to agriculture increases. Southwest Region Significant attributes of hunter satisfaction in the southwest region included the number of deer in the region, deer vehicle collisions, and age. Results suggest that satisfaction increases when the number of deer increases. Similar to the northeast nonhunters, older nonhunters in the southwest region tend to be less satisfied than younger hunters. The sign on deer vehicle collisions is counter-intuitive, as it suggests that increases in the annual number of deer vehicle collisions increases satisfaction. Significant attributes of nonhunter satisfaction included the number of deer and the number of deer-vehicle collisions. Results suggest that nonhunter satisfaction increases with increases in the number of deer, and decreases with increases in deer vehicle collisions. Issue Activity Model Estimation In addition to the satisfaction rating, respondents were asked if they would engage in certain types of “issue activity” if Scenario A (see Fig. 9) were the situation in their region. Types of issue activities respondents were asked about included (1) contact someone with authority to get the situation changed, or (2) take steps myself to change the situation. Respondents could also choose (3) do nothing because the situation would not be that bad, or (4) do nothing because it would not change the situation. As explained above, the experimental design used to determine the attribute levels for the 82 choice experiment resulted in 8*3 within each region for both hunters and non-hunters difl'erent “Scenario A ’s ” for each type of choice set, thus respondents were questioned about engagement in issue activity for 48 different deer situations. Item non-response was less than 6% for hunters and ranged between 7% and 9% for nonhunters. The dataset for the issue activity models was formed by combining ratings for increasing, decreasing, and marginal situations. Issue activity models were estimated using maximum likelihood techniques for discrete choice data. The data on issue activity was coded as a 1 if respondents stated they would engage in issue activity (either 1 or 2 above) and a zero if respondents stated they would not engage in issue activity (3 or 4 above), given the particular deer situation. The models were estimated using LIMDEP 7.0. The models assume that respondent engagement in issue activity depends on the levels of the deer-related attributes in the scenario. Respondent characteristics variables were also included to determine their effect on satisfaction level; however, none of these variables were retained in the selected Issue Activity models as they generally confounded any significant effects of the deer- related attributes. The only models containing any significant attributes were the northeast and southwest hunters and the southwest nonhunters. In the northeast the attribute deer numbers was significant (p<0.01), suggesting that as deer numbers decreased the likelihood of engaging in issue activity increased. In the southwest the attributes deer numbers and deer damage to agriculture were both significant (p<0.02 for both attributes). These results suggest that the likelihood of engaging in issue activity increases as deer numbers decrease and as deer damage to agriculture increases. 83 The nonhunter issue activity model in the southwest contained two significant attributes: deer damage to agriculture and deer vehicle collisions. While the interpretation for deer vehicle collisions is logical, e. g. the likelihood of engaging in issue activity increases when deer vehicle collisions increase, the coefficient on deer damage to agriculture has a negative sign, implying that the likelihood of issue activity increases as the amount of deer damage to agriculture decreases. Again the lack of qualitative research in developing these questions may be a factor in the poor model results and counter-intuitive sign. Given the above results, it may be more beneficial to simply examine percentages rather than modeling stakeholder satisfaction and issue activity. For example, calculating the percentages of responses at each point of the agreement scale when the situation is one other than the status quo may be somewhat insightful in examining stakeholder satisfaction. Similarly, examining the percentage of respondents who would engage in issue activity if the situation were one other that the status quo may be more useful than modeling issue activity, particularly given the poor results of the models. This information is summarized in Table 19. 84 Table 19. Satisfaction and Issue Activity Associated with Deer Population Changes Stakeholder Group Hunter Nonhunter Region NW NE SW NW NE SW Satisfaction level and engagement in issue activity when deer numbers and extemalities increase % Extremely Satisfied 7 9 13 5 3 5 % Satisfied 33 33 29 23 22 19 % Neither Satisfied nor Dissatisfied 25 25 23 23 27 29 % Dissatisfied 26 26 28 36 35 32 % Extremely Dissatisfied 9 7 7 13 12 15 % Engage in issue activity when deer numbers and extemalities 35 30 34 32 29 33 increase Satisfaction level and engagement in issue activity when deer numbers and extemalities decrease % Extremely Satisfied 7 4 8 12 13 12 % Satisfied 30 25 29 47 43 50 % Neither Satisfied nor Dissatisfied 20 21 17 26 22 19 % Dissatisfied 30 27 27 12 19 16 % Extremely Dissatisfied 13 23 19 3 3 3 % Engage in issue activity when deer numbers and extemalities 38 39 38 15 19 14 decrease Results in Table 19 suggest that hunters are generally more satisfied than nonhunters when deer numbers and extemalities increase, and about the same percentage of hunters and nonhunters in all regions would engage in issue activity in this type of situation. When deer numbers and extemalities decrease nonhunters appear to be more satisfied than hunters, and for all regions the percentage of hunters engaging in issue activity is higher than the percentage of nonhunters. It is interesting to note that when 85 extemalities and deer numbers increase, there are no significant differences between hunters and nonhunters in engagement in issue activity. However, when deer numbers and extemalities decrease, hunters and nonhunters in each region differ significantly, with a larger percentage of hunters engaging in issue activity in this type of “decreasing” situation (Northwest )8 = 44.00, p < 0.01; Northeast x2 = 35.14, p < 0.01; Southwest x2 = 56.60, p < 0.01). These results generally support previous findings presented in this chapter that suggest hunters want more deer than nonhunters and may be less concerned about deer-related extemalities. In addition the results suggest that hunters may be more apt to engage in some level of issue activity when they are not satisfied with deer situations in their region. 86 CHAPTER 4. CHOICE MODEL ESTIMATION AND RESULTS This chapter begins with a brief summary of the choice experiment framework presented in detail in Chapter 2. Results of the regionally aggregated models are then presented, followed by models for the regional level. The chapter concludes with a comparison of models estimated via different preference elicitation formats. Background CE’S and the resulting choice models assume that people make choices based on the utility provided by different options. In the white-tailed deer CE, Options (referred to as scenarios) consisted of different levels of deer-related attributes. The attributes included the number of deer, herd health, deer damage to residential property, deer damage to agriculture, deer-vehicle collisions, and deer browsing in the forest. Hunter surveys (from the MDNR sample only) contained an additional attribute - the number of mature bucks. Respondents were shown a choice set (Figure 10) and asked to choose which scenario they would prefer for their region. Attribute levels for each alternative (except the status quo) are determined by an experimental design plan, but must be able to vary independently. This was a significant problem for the white-tailed deer CE, as independent variation created several counterfactual scenarios, e. g. deer numbers increase but deer vehicle collisions decrease, relative to the status quo. To overcome this problem three types of choice sets were presented to respondents: a choice set where attributes levels in each of two alternatives 87 Current Situation Scenario A Scenario B Number of deer in region Number of mature bucks in region Percent of deer in region with at least one characteristic of poor health Percent of residential properties in region experiencing some deer damage Deer damage per acre of cropland in region ($/acre) Annual number of deer- vehicle collisions in region Percent of forest area in region experiencing heavy deer browsing Which do you prefer for your region? Current D Scenario A El Scenario B C] (Check one) Figure 10. Example of Choice Set increase relative to the status quo; a choice set where attribute levels in two alternatives decrease relative to the status quo; and a choice set where attribute levels in two alternatives increase or decrease relative to the status quo, but changes are small, referred to as the marginal choice set. All of these models can be referred to as 3-way models, as each choice set offers respondents three options - the status quo and two alternatives. 88 Respondents from the SOS sample were also given a choice set that contained the status quo, an increasing scenario, a decreasing scenario, and a marginal scenario, referred to as a 4-way choice set. Appendix A contains a complete survey instrument for the MDNR and SOS samples, illustrating each type of choice set. Choice models were estimated by combining data fi'om the increasing, decreasing, and marginal choice sets and estimating a separate model from the 4-way choice (SOS version only). In total, 12 different models were estimated from the data. Table 20 outlines the types of models estimated. Multinonrial logit (MNL) models, described in detail in Chapter 2, were estimated from the data. To begin the modeling process, the deer-related attributes and a dummy variable representing the status quo were entered as independent variables. When the data permitted, nested MNL models were estimated. Nested MNL models allow more flexible error structures by specifying an hierarchical choice setting and estimating additional parameters for each choice set partition. Figure 11 shows the hierarchical setting used for white-tailed deer. Using a nested MNL and partitioning the choice sets into respondents who chose a scenario other than the current situation, labeled “Change,” and respondents who stayed with the current situation, labeled “No Change,” allows the variance of the random components to vary across subsets of the partitions. This relaxes the independence assumption (for alternatives sharing a partition) as well as the identical distribution assumption between alternatives in different partitions (Louviere et al. 2000) 89 Table 20. Types of Choice Models Estimated from Survey Data Model No. of Data Used for Obs. Estimation Hunters - MDNR Sample Northwest Hunter 1,139 1 Northeast Hunter 1,202 1 Southwest Hunter 1,105 1 Statewide Hunter 3,446 2 Nonhunters - SOS Sample Northwest Nonhunter 683 3 Northeast Nonhunter 762 3 Southwest Nonhunter 892 3 Statewide Nonhunter 2,337 4 Northwest Nonhunter 4-way 221 5 Northeast Nonhunter 4-way 250 5 Southwest Nonhtmter 4-way 300 5 Statewide Nonhunter 4-way 771 6 1 = Combined increasing, decreasing, marginal choice sets = Combined increasing, decreasing, marginal choice sets from all regions = Combined increasing, decreasing, marginal choice sets fiom nonhunter respondents 4 = Combined increasing, decreasing, marginal choice sets from nonhunter respondents from all regions 5 = 4-way choice set fi'om nonhunter respondents 6 = 4-way choice set from nonhunter respondents from all regions 90 Deer Scenario Choice Change No Change Scenario A Scenario B Status Quo Figure 11. Schematic of Nested Multinomial Logit Nesting results in the estimation of an inclusive value parameter for the choice set partition (change or no change). For all nested models the inclusive value parameters for change and no change were constrained to be equal. The interpretation of parameter estimates is that inclusive value parameters not significantly different than one suggest that the nesting specification could be collapsed into a non-nested model (Louviere et al. 2000) During the modeling process additional variables were interacted with the dummy variable and entered into the model in an effort to improve model results. These variables 91 included respondent demographic characteristics, respondent experience with deer-related extemalities, and likert scores from several of the belief statements. Because these variables were interacted with the dummy variable for the status quo, their interpretation concerns only the likelihood of changing fiom the status quo. For example, a negative sign on a belief statement coefficient would indicate that the higher a respondent’s score is on the belief scale, the more likely they are to change from the status quo. While a handful of these variables were significant in the models, they generally (1) confounded the effects of the deer-related attributes, and (2) precluded the use of a nested model (nested models always improved the overall model results). Therefore, these variables were dropped, and the final models included only the deer-related attributes and a status quo dummy. All models were estimated using LIMDEP 7 .0. Results are presented below, beginning with statewide models, followed by regional models and a comparison of 3- way and 4-way models for the nonhunters. Goodness of fit, as measured by the likelihood ratio index, for regional and statewide models ranged fiom 0.02 to 0.29. For all models that retained a nested structure, the nested version increased the value of the likelihood ratio index. It is important to note that this index does not have the same interpretation as an R-squared value fiom linear regression models. The likelihood ratio index, sometimes referred to as McFadden’s R-squared, is the percent increase in the log likelihood function above the value taken when all parameters are zero (or the value of no model). However, the index can only be used to compare models estimated from the same data and the same set of alternatives, and cannot be used to compare models 92 estimated from different data sets. Although generally a model with a higher likelihood ratio index is said to fit the data better than a model with a lower index, the index has no intuitively interpretable meaning (Train 2003). Statewide Choice Models Statewide models are non-nested MNL models. The statewide hunter model contained 6 significant attributes: the number of deer, the number of mature bucks, herd health, deer-vehicle collisions, deer browsing, and the status quo dummy variable. Similar results were found for nonhunters, where significant attributes included the number of deer, herd health, deer-vehicle collisions, deer browsing, and the status quo dummy variable. For both hunters and nonhunters, deer damage to residential property and deer damage to agriculture were non—significant attributes. For both hunters and nonhunters, significant attributes had the expected sign, indicating that deer-related extemalities have a negative effect on utility and deer numbers (and buck numbers for hunters) have a positive effect. Table 21 presents the results of the statewide hunter and nonhunter models. 93 Table 21. Results of Statewide Choice Models Hunters Nonhunters Number of deer 2.6818 2.1377 (std. error) (0.5667) (0.8037) Number of mature 7.5800 bucks (1.2365) —0.0741 -0.0496 new health (0.0108) (0.0150) Deer damage to 0.0076 -0.0145 residential property (0.01 10) (0.0139) Deer damage to -0.0537 -0.05 83 agriculture (0.0559) (0.1087) Deer-vehicle -0.0009 -0.0019 collisions (0.0001) (0.0002) Deer browsing in -0.0097 -0.0310 forest (0.0041) (0.0156) 1.0895 0.8340 Status qu" dummy (0.0475) (0.0521) Log-L -3361.9 -1889.7 These results suggest that, at a regionally aggregated level, hunters and nonhunters experience gains in utility when deer populations and buck populations (hunters only) increase, all else being equal. Because the units are the same, a direct comparison between deer number and buck number coefficients can be made for hunters. This comparison reveals that buck numbers has a much stronger effect on utility than do deer numbers. Both hunters and nonhunters suffer disutility when three of the five deer- related extemalities increase: the percent of deer in poor health, the number of annual deer-vehicle collisions, and the percent of forest area that is heavily browsed by deer. Additionally, the significance of the dummy variable suggests that moving away from the 94 status quo would bring disutility to both hunters and nonhunters. Utility Comparisons Coefficients from CE models do not have the same interpretation as OLS coefficients in that they do not directly reveal the change in the dependent variable given a 1 unit change in an independent variable. However, they can be used to compare utilities of a variety of potential deer scenarios. From Chapter 2, the probability that an individual prefers (and thus chooses) one scenario over all other scenarios, assuming the scale parameter is 1, can be expressed as Pr alC = exp(flXa) /ZC: exp(E(,-) Having estimates of [3, potential X’s can now be plugged into the model to determine choice probabilities for alternative scenarios. To illustrate, suppose the scenarios in Table 22 below are the outcomes of two potential management strategies for the aggregate region of the Northwest, Northeast, and Southwest regions. 95 Table 22. Hypothetical Outcomes for Deer-related Attributes Scenario A Scenario B Number of deer 10% increase 15% increase Number of mature bucks 5% increase 1% increase Percent of deer with a characteristic of poor 20% 25% health Percent of properties 10% 10% experrencmg deer damage Deer damage per acre of 5$ 5$ cropland Annual deer-vehicle 3,500 3,000 collrsrons Percent of forest area experiencing heavy deer 30% 20% browsing ‘ Using the models to predict choice probabilities for Scenario’s A and B shows that, for hunters Prob (AIC) = 0.97 Prob (BIC) = 0.03 and for nonhunters Prob (AIC) = 0.25 Prob (BIC) = 0.75 Given the hypothetical outcomes above, if a manager is considering only the human dimension of deer management, Scenario A provides greater utility for hunters and thus is the preferred scenario for the aggregate region, while Scenario B is the preferred scenario for the region for nonhunters. However, when no increases in the 96 number of mature bucks occurs in either Scenario A or B, the hunter choice probabilities become Prob (AIC) = 0.42 Prob (BIC) = 0.58 The difference between the two sets of choice probabilities for hunters illustrates the importance hunters place on mature bucks. When there are no increases to the number of mature bucks, the management decision, if based on the wishes of the majority of each stakeholder group, becomes the same for both hunters and nonhunters, given the hypothetical scenarios A and B. It should be noted at this point that both hypothetical scenarios A and B involve increases to both deer and mature buck populations. Chapter 5 introduces two additional types of situations, situations where deer populations increase but mature bucks decrease, and situations where mature bucks increase but deer populations decrease. Marginal Rates of Substitution CE results can also be used to calculate the marginal rates of substitution among the deer-related attributes. Marginal rates of substitution (MRS) represent the ratio of the change in utility with respect to one attribute and the change in utility with respect to a second attribute. Stated another way, MRS is the rate at which individuals will substitute (changes in) one attribute for (changes in) a second attribute such that overall utility remains constant. From the CE results, this ratio is expressed by - 131/132, for two attribute coefficients. MRS are relevant to deer management as they can reveal the amount of 97 extemality changes stakeholders will trade-off for a given percent change in the number of deer or number of mature bucks. For example, hunters may be willing to accept, or tolerate, a decrease in the number of deer or mature bucks in order to see a decrease in the number of deer-vehicle collisions. Similarly, hunters may accept an increase in deer- vehicle collisions in order to have more deer or mature bucks. MRS defines the increase (decrease) in deer that is necessary to keep utility constant when deer-vehicle collisions, or other attributes, increase (decrease). MRS were calculated for all attributes to examine the trade-offs hunters and nonhunters would accept for a 1% increase in the deer and mature buck population, relative to the status quo (Table 23). Table 23. Marginal Rates of Substitution for. Deer and Mature Bucks: Statewide Acceptable trade-offs Acceptable trade-offs for a 1% increase in for a 1% increase in deer mature bucks Hunters Hunters Nonhunters Percent °.f deer With a 1% more“ 0.4% more“ 0.4% more“ charactenstrc of poor health Percent of residential properties experiencing deer 10.0% more 3.5% more 1.5% more damage Deer damage per acre of $1.41 more $0.50 more $0.37 more cropland Annual deer-vehrcle 84 more“ 30 more“ 11 more" collrsrons Percent of forest area experiencing heavy deer 8% more* 3% more“ 0.7% more“ browsing * Indicates that the attribute is significant in the statewide choice model. 98 Results in Table 23 show that hunters will accept greater extemality increases for increases in mature bucks than increases in deer, and will also accept greater increases in deer-vehicle collisions and deer browsing than will nonhunters for increases in the number of deer. Interestingly, hunters and nonhunters will accept about the same increase in poor herd health for an increase in deer. In calculating MRS, it is possible to determine the trade-offs stakeholders would accept for large increases in deer or mature bucks by sealing the ratio, e. g. for a 100% increase in mature bucks hunters would accept approximately 8,400 more deer-vehicle collisions per year. It is most likely that hunters would not accept this many more deer- vehicle collisions, nor is it likely that mature bucks could increase by 100%. This is a problem with CE models that assume linear preferences and predictions outside the range of the data, as does the model for deer. However, more complex functional forms for preference models require larger sample sizes and more resources. Given the constraints of this CE, linear models seemed the most appropriate functional form. Further, research has demonstrated that within a certain range linear preferences predict well (Louviere et al. 2000), and thus, up to a subjective cut-off point, knowledge of MRS would be useful for deer management professionals. In summary, results of the aggregated choice models Show that, for both hunters and nonhunters, the extemalities ‘deer damage to residential property’ and ‘deer damage to agriculture’ do not have a significant effect on utility, while the extemailities poor herd health, deer-vehicle collisions, and deer browsing provide disutility to both groups. Hunters will generally trade-off greater increases in extemalities for increases in deer than 99 nonhunters, and will accept larger extemality increases for more mature bucks than for more deer. Among all attributes, mature bucks seems to be a key attribute for hunters. Models were simulated using hypothetical management outcomes that increase and decrease deer and buck numbers, and the extemalities, by various amounts. Generally, when outcomes included changes to buck numbers, choice probabilities for hunters and nonhunters were different. However, when there were no changes to buck numbers but all other changes were the same, both groups generally had similar probabilities. Chapter 5 expands on model simulation and presents an array of choice probabilities under different management outcomes for the statewide models as well as regional models. The results above are based on pooling data from all three regions. In the following section model results are presented separately for the Northwest, Northeast, and Southwest, illustrating some differences when the data is disaggregated by region. Regional Results Northwest Region Hunter and nonhunter models for the northwest region were estimated using the hierarchical scheme discussed above. Inclusive value parameters for hunters are significantly different from one at p < 0.12. For nonhunters, inclusive values are significantly different from one at p < 0.01 . Although the nesting structure did not result in highly significant inclusive value parameters for hunters, the structure was retained due to larger log-likelihood values. 100 For hunters in the northwest mature bucks have a positive effect on utility and the extemalities poor herd health, deer-vehicle collisions, and deer browsing have a negative effect. The dummy variable for the status quo is significant. The number of deer, residential damage, and agricultural damage are not significant for hunters. For nonhunters, all extemality attributes with the exception of deer browsing are significant. The number of deer is not a significant attribute for nonhunters. The status quo dummy is significant for both hunters and nonhunters. Tables 24 and 25 Show model results for hunters and nonhunters. Northeast Region Hunter and non-hunter models for the nOrtheast region are non-nested models. For hunters, deer and mature bucks are both significant attributes, with percentage increases in mature bucks having almost twice the effect on utility as do deer. Significant extemalities include poor herd health, deer damage to agriculture, and deer-vehicle collisions. The dummy variable for status quo is also significant. For nonhunters deer numbers is significant, as are the extemalities deer damage to agriculture, deer-vehicle collisions, deer browsing, and the status quo dummy variable. (Tables 24 and 25). Southwest Region Hunter models for the southwest region were nested, with inclusive value parameters significantly different from one at p < 0.01. The nonhunter nested model failed to converge, thus the non-nested model was retained. For hunters, deer numbers 101 and buck numbers were both significant, and, similar to the northeast hunters, mature buck numbers have nearly twice the effect on utility as do deer numbers. The extemalities poor herd health, deer-vehicle collisions, deer browsing, and the status quo dummy are also significant. For nonhunters the extemalities poor health, deer damage to agriculture, deer-vehicle collisions, and the status quo dummy are significant (Tables 24 and 25). In the nonhunter model the sign on deer damage to agriculture is positive, suggesting that this extemality has a positive effect on utility. Table 24. Results of Regional Choice Models: Hunters Northwest Northeast Southwest Number of deer 1.2135 4.7903 8.1879 (std. error) (2.3200) (1.0729) (3.0490) 12.8086 8.1711 14.5112 Num‘x” °f mam“ bud“ (4.9232) (2.2068) (6.2176) -0.0891 0.0864 -0.3046 He’d health (0.0252) (0.0199) (0.0720) Deer damage to residential -0.0063 0.0025 -0.0181 property (0.0341) (0.0195) (0.0299) . 0.0621 0.6199 0.3845 De“ “wage t° agnwlm’e (0.1669) (0.1840) (0.3455) . . . 0.0007 0.0012 0.0013 Deer-vehicle collrsrons (0.0003) (0.0003) (0.0004) Deer browsin in forest 0.0110 0.0182 0.0916 ‘3 (0.0039) (0.0203) (0.0383) 1.2155 1.0577 1.8138 Status ‘1‘” “WW (0.2714) (0.0842) (0.4778) Inclusive Value Parameters 0.3429 0.4224 Change;No change (0.2167) (0.1650) Log-L -1 144.3 -1165.3 0023.9 102 Table 25. Results of Rggional Choice Models: Nonhunters Northwest Northeast Southwest Number of deer 1.7862 2.7995 1.5774 (std. error) (5.6832) (1.5355) (1.3380) Herd health -0.0931 0.0022 -0.1710 (0.0457) (0.0254) (0.0593) Deer damage to residential -0. 1013 0.0145 00107 Property (0.0424) (0.0070) (0.0229) Deer damage to agriculture -1.0740 -0.8768 0.5356 (0.4525) (0.2465) (0.2557) Deer-vehicle collisions -0.0047 -0.0024 -0.0021 (0.0009) (0.0004) (0.0002) Deer browsing in forest -0.1880 -0.0507 -0.0276 (0.0517) (0.0268) (0.0257) Status quo dummy 0.9118 0.8263 0.8143 (0.3521) (0.0268) (0.0879) Inclusive Value Parameters 0.2791 Change;No change (0.0918) Log-L -53 1.9 -639.2 —689.2 As in the statewide model, significant attributes for hunters in all regions include the number of mature bucks, herd health, and deer-vehicle collisions. However, the regional models show some differences from the aggregate model. For example, deer are Significant in the northeast and southwest but not in the northwest. This finding is consistent with the results of other questions in the survey that demonstrate that while 21% of hunters in the northwest would prefer less deer than last year for their region, only 10% of hunters in the northeast and 15 % of southwest hunters prefer less deer than last year. Additionally, anecdotal evidence fi'om pretest respondents in the northwest indicated that, although abundant, deer in this region are relatively small, which supports 103 both the insignificant coefficient on deer and the strong effect mature bucks have on lltlllt}l. Other regional differences include the Significance of deer browsing in the northwest and southwest and the significance of deer damage to agriculture in the northeast. The latter is interesting as deer damage to agriculture is not significant in the statewide model, and only 12.9% of hunters in the northeast stated they would be very concerned with increases in deer damage to agriculture, compared to 21.5% and 18.7% of hunters in the northwest and southwest regions. Additionally, a greater proportion of hunters in the northeast than in the northwest or southwest said they would be “not concerned at all” with an increase (see Chapter 3). However, these questions were posed without constraints, e. g. there were no trade-offs involved. In the choice scenarios respondents were indirectly making trade-offs among changes in deer damage to agriculture and changes in other extemalities (and deer and buck numbers), thus introducing some additional variables into the decision-making process. Nonhunter models vary by region, and generally differ fi'om the statewide model to a greater extent than the hunter models do. Deer-vehicle collisions and deer damage to agriculture are significant for all regional nonhunter models. Deer damage to residential properties, which is not significant in any other model, is significant for northwest nonhunters. Subj ectively, this result is consistent with qualitative research conducted during the survey development phase, as discussions with private landscaping and tree nursery companies in the northwest indicated that deer damage was a problem for the region. 104 Deer numbers are significant in the northeast but not in the northwest or southwest, a finding consistent with other parts of the survey. For example, in both the northwest and southwest regions slightly over 30% of respondents stated they would prefer less deer in their region than they had last year, while only 25% of respondents in the northeast felt this way. Additionally, a greater percentage of nonhunters in the northeast (24%) preferred more deer for their region than did nonhunters in the northwest (15%) or southwest (10%). An underlying reason for these attitudes may be, at least in part, the presence of bovine tuberculosis in the northeast. Due to bovine TB, liberal deer harvest policies have been in place over the last several years in the northeast, in an effort to eradicate the disease. During the qualitative research period people frequently discussed their concern over decreasing deer numbers and the liberal harvest policies in the region. While the perception of deer population declines was evident in the northeast, deer population decreases were generally not a concern in the northwest and southwest. This anecdotal evidence is borne out by other survey questions indicating that 47% of nonhunters in the northeast would be concerned or very concerned if the deer population decreased in their region, while only about 39% and 30% of nonhunters in the northwest and southwest felt this way. Marginal Rates of Substitution For comparison with the statewide model, MRS were calculated for regional hunter and nonhunter models, although the number of deer was not Significant in each 105 case (Tables 26 and 27). Theoretically MRS can be calculated to determine trade-offs among extemalities, e.g. acceptable trade-Offs between deer browsing and deer-vehicle collisions. Since the focus of this chapter is on trade-offs between changes in the deer and buck populations and changes in extemalities, the full set of MRS among all attributes is not displayed. Table 26. Hunter Marginal Rates of Substitution for Deer and Mature Bucks: R 'onal Acceptable trade-offs for a 1% increase in deer Northwest Northeast Southwest Hunters Hunters Hunters Percent of deer With a 0.1% more* 0.5% more* 0.2% more* characterrstrc of poor health Perm?“ oi’eS'dem‘al pmpemes 1.9% more 19.0% less 4.5% more experrencmg deer damage Deer damage per acre of cropland $0.19 more $0.07 more* $0.21 more Annual number 0f deer-vehicle 17 more* 40 more* 63 more* collrsrons Percent Of forest area 1% more* 2.6% less 0.9% more* experiencing heavy deer browsing Acceptable trade-offs for a 1% increase mature bucks Percent of deer with a 0 * 0 * 0 * characteristic ofpoor health 1.4 /o more 0.9 A. more 0.5 A) more Fem?“ °ire$demlal pmpemes 20.3% 32.6% less 8.0% more experrencmg deer damage Deer damage per acre of cropland $2.06 $0.13 more* $0.38 more Ami“?! number Of deer-vehicle 183 more* 68 more* 111 more* collrsrons Percent Of forest area 12% more* 4.4% less 1.6% more* experiencing heavy deer browsing * Indicates that the attribute is significant in the regional choice model. 106 Table 27. Nonhunter Marginal Rates of Substitution for Deer: Regional Acceptable trade-offs for a 1% increase deer Northwest Northeast Southwest Nonhunters Nonhunters Nonhunters Percent of deer wrth a 0.2% more* 12.7% more .1% more* characterrstrc of poor health Percent of residential properties experiencing deer 0.2% more* 1.9% more 1.5% more damage Deer damage per acre 0f $0.01 more* $0.03 more* $0.03 less* cropland Annual number of deer-vehicle 4 more* 11 more* 8 more* collrsrons Percent of forest area experiencing heavy deer 0.1% more 0.5% more* 0.6% more browsing * Indicates that the attribute is significant in the regional choice model. Comparing MRS for hunters shows that hunters in the northwest will trade-off fewer deer in poor health and fewer deer-vehicle collisions for an increase in the number of deer than will hunters in the northeast or southwest. However, for increases in the number of mature bucks northwest hunters will accept more of all significant extemalities than will northeast or southwest hunters, illustrating the preference northwest hunters have for mature bucks relative to deer and deer-related extemalities. For hunters in the northwest and southwest the acceptable extemality trade-offs are considerably greater for mature bucks than for deer, though in the northeast this relationship is not as strong. As discussed previously, bovine TB is present in the northeast region, and liberal harvest quotas may elevate the overall importance of increasing the deer population relative to 107 increasing the number of mature bucks. Comparing the northeast hunters to northeast nonhunters shows that the acceptable extemality trade-offs for nonhunters are lower than those for hunters. This preference difference is consistent with results presented in Chapter 3, where descriptive results show that nonhunters are generally more concerned with increases in deer-related extemalities than are hunters. In summary, the regional hunter models are more consistent with the aggregate statewide hunter model than are regional nonhunter models. For hunters, mature bucks, herd health, and deer-vehicle collisions are significant in all regional models, while the significance of deer numbers, deer damage to agriculture, and deer browsing vary by region. Hunters have stronger preferences for mature bucks than for deer in all regions, though this relationship is strongest in the northwest and southwest. For nonhunter regional models deer damage to agriculture and deer-vehicle collisions are significant in all regions, while the significance of other attributes varies by region. For both hunter and nonhunters in all regions the status quo dummy variable is significant, indicating that people tend to prefer the current situation in their region to potential changes. Comparison of Preference Elicitation Formats In addition to the three 3-way choice sets (increasing, decreasing, and marginal), the SOS versions of the survey contained an additional choice set which included a 4-way choice among the current situation, a situation where all attributes increase, a situation where all attributes decrease, and a situation with small attribute changes in either direction (see Chapter 2 for a detailed description). Using this design, two separate 108 choice models can be estimated: a model pooling the increasing, decreasing, and marginal scenarios and a second model using data from the 4-way choice. The pooled model is the statewide nonhunter model presented above in Table 21. The 4-way model was estimated separately using data on the 4-way choice, and comparisons of the two models are presented below. One important distinction between the two types of models concerns the number of alternative scenarios in each choice set. As previously discussed, the pooled model is based on pooling choices from increasing, decreasing, and marginal sets, each of which contained 3 scenarios in the choice set. Thus respondents had to evaluate three alternatives before making a choice. In the 4-way choice model, respondents evaluated four scenarios before making a choice, which may be a slightly more complex task. On the other'hand, the 4-way choice was presented near the end of the survey, and at this point respondents may have become familiar with the choice experiment format and may have even devised heuristic tools to help them answer choice questions, thus the addition of a fourth scenario would not add significantly to the task complexity. Several authors have demonstrated a learning effect in repeated measures experiments (Morrison 2000; Bradley and Daly 1994) though generally more than three choice tasks were required to reveal evidence of a learning effect. A comparison of the pooled and 4-way model was conducted to test the hypothesis that different preference elicitation formats over the same range of attributes and attribute levels will result in similar preference structures. An informal hypothesis test was conducted by using the coefficients of the pooled model as starting values for the 4-way model, constraining the model to zero iterations, and conducting a log-likelihood 109 ratio test on the likelihood functions of the pooled and 4-way models. Results of this informal test Show that the models are significantly different (x2 = 25.23, p < 0.01). Results of pooled and 4—way statewide models are presented in Table 28. MRS are not compared between the two models as deer was not significant in the 4—way model. Table 28. Results of Pooled versus Four-way Choice Model Pooled 4-way Number of deer 2.1377 0.9590 (std. error) (0. 8037) (1.1540) -0.0496 41.0579 He’d health (0.0150) (0.0220) Deer damage to -0.0145 -0.0465 residential property (0.0139) (0.0244) Deer damage to —0.0583 -0.1927 agriculture (0.1087) (0.1624) Deer-vehicle -0.0019 -0.001 1 collisions (0.0002) (0.0002) Deer browsing in -0.03 10 -0.0406 forest (0.0156) (0.0247) 0.8340 0.6803 Status qu° dm‘m‘y (0.0521) (0.1180) Log-L -1889.7 -690.3 pseudo R2 0.26 0.35 Results demonstrate that there are differences in the significant attributes fi'om each model. For example, residential damage is significant in the 4-way model but non- significant in the pooled model, while the opposite pattern holds for deer browsing. Further, the number of deer is not significant in the 4-way model. Deer-vehicle collisions, herd health, and the status quo are significant in both models. Full results are not presented for the regional comparisons. However, regional 4- 110 way models generally showed a fewer number of significant attributes, and those that were significant had larger p-values than the pooled model, as expected given the larger sample size. The same informal hypothesis tests were conducted on pooled and 4-way regional model and similar results to the statewide tests were obtained at the regional level. Both statewide and regional comparisons between pooled and 4-way models suggest that different elicitation formats result in different preference structures. One reason for this may be differences in choice complexity of the pooled and 4-way choice. In the 4-way choice respondents evaluated the current situation and three alternative scenarios, each of which was moving in a different direction relative to the status quo. This may have been a more difficult task than the choices from the pooled model, where respondents evaluated fewer alternatives per choice set (the current situation and two alternatives), both of which were moving in the same direction relative to the status quo. On the other hand, it is possible that, for some respondents, the 4-way choice was easier, as they may have been looking at the way the attributes move in making their decision. For example, respondents who were looking for decreases may have had an easier time choosing in the 4-way choice because only one scenario was decreasing. In the 3-way model, even if respondents knew they wanted a decrease in the attributes, they still had to compare two decreasing alternatives before making a choice. Further, the alternatives in the 3-way choice set contained identical changes in deer numbers, in an effort to encourage those respondents who tend to make choices based solely on the number of deer to consider all the attributes before making a choice. Survey questions do not 111 provide insight as to which type of choice was more complex for whom; however, if choice task complexity was different between the pooled and 4-way models this may affect the respondent’s decision process and thus the model results. Status quo choices were examined in each of the increasing, decreasing, marginal, and 4-way choice sets in the survey. In each of the increasing, decreasing, and marginal choice sets the status quo made up a larger percentage of choices than in the 4-way choice. For example, in the increasing, decreasing, and marginal choice set, 84%, 32%, and 44% of choices were for the status quo, respectively. Note that these choice sets are constrained, e. g. a respondent may want fewer deer, but in the increasing choice set they are not offered that choice, thus they default to the status quo. When all choice sets were pooled, the status quo represented 53% of the choices made. In contrast, in the 4-way choice set only 23% of the choices were the status quo. In the 4-way choice set respondents did not have constraints on their choice, e. g. if they wanted fewer deer they could choose that option from the choice set rather than defaulting to the status quo. Thus, in addition to choice task complexity, it is possible that the different constraints respondents faced at each choice set may also affect the results of the pooled and 4-way model. 112 CHAPTER 5. CONCLUSIONS AND MANAGEMENT IMPLICATIONS This chapter begins with a discussion of preferences for deer and deer-related extemalities, based on the results of the choice experiment and other survey components. This discussion is followed by describing the general limitations of the research and a retrospective examination of issues that are specific to the use of a choice experiment survey format. The chapter concludes with a discussion of the big picture for deer management. Preferences for Deer and Deer-related Extemalities The purpose of this research was to examine and quantify stakeholder preferences, with the intention of informing management abOut preferences, trade-offs, satisfaction, and issue activity related to changes in the deer population and related extemalities. Results of both the CE and other survey components demonstrate that people in Michigan care about the deer population and many of the extemalities associated with the population, and are willing to make trade-offs for increases or decreases in the number of deer and the level of particular extemalities. To a large extent the research was able to quantify these trade-offs and provide management with guidance and recommendations, though the preference information is not perfectly consistent (see Limitations in this chapter). In general, choice model results show that the number of deer and the number of mature bucks both have a positive effect on utility, although there are some regional differences. Further, all significant extemalities, with the exception of deer damage to 113 agriculture in the southwest nonhunter model, have a negative effect on utility. The extemalities herd health and deer vehicle collisions are generally significant across regions and stakeholder groups, and appear to be the extemalities respondents are most concerned about. In Chapter 1, four research questions were introduced. These questions are now examined using a synthesis of the results presented in Chapters 3 and 4. Question (I): How does the relative importance of deer and deer-related extemalities difier among diflerent stakeholder groups? At the statewide level, the relative importance of deer and deer-related extemalities are similar for hunters and non-hunters. Both groups have positive utility for deer, and negative utility for extemalities. Additionally, neither deer damage to residential property nor deer damage to agriculture was a significant attribute for either group, suggesting that respondents are least concerned with these extemalities relative to the other extemalities described in the survey. The hunter choice model shows that, at the statewide level, mature bucks provide about three times more utility for hunters than do deer. Choice model results show that hunters in all regions place relatively more importance on having more mature bucks in their region than on having more deer. Another consistency across hunters in each region is the importance placed on the extemalities herd health and deer-vehicle collisions, both of which have negative and significant coefficients in all regions. None of the regional hunter models show significant coefficients on the residential property damage coefficient, suggesting that, for 114 hunters, it is a relatively unimportant extemality associated with the deer population. Similar to the hunter models, all nonhunter models have significant and negative coefficients on deer-vehicle collisions, suggesting that nonhunters also place relatively high importance on this extemality. Also similar to the hunter models, the attribute herd health appeared to be relatively important to nonhunters, though only in the northwest and southwest regions. Deer damage to residential property also appeared to be relatively unimportant to nonhunters as a group. In contrast to hunters, the deer ntunbers coefficient was only marginally significant in one region (northeast), suggesting that nonhunters place less importance on the number of deer in the region than do hunters. At the regional level there are some differences among hunters and among nonhunters. For example, in the northwest and southwest, deer browsing in the forest appears to be an extemality that hunters feel is relatively important, whereas hunters in the northeast place relatively more importance on deer damage to agriculture. Among nonhunters, deer damage to residential property was significant only in the northwest model, and deer browsing in the forest was marginally significant in the northeast. Deer damage to agriculture was significant and brought disutility in the northwest and northeast, though in the southwest the coefficient sign implied that deer damage to agriculture has a significant positive effect on utility. Searching for data coding errors and re-examining the experimental design for errors did not shed any insight as to why deer damage to agriculture would have an unexpected sign. As discussed in Chapters 2 and 3, the survey also contained questions that asked respondents to express their level of concern over changes in the attribute levels. 115 Although the format of these questions is different than the choice experiment in that respondents do not evaluate a bundle of attributes and make trade-offs among them, many responses to the concern questions generally support the choice model results. For example, when asked to express concern about a decrease in the number of deer, more than 60% of hunters in each region stated they would be concerned or very concerned, while 45% or less of nonhunters in each region expressed this level of concern. This finding is generally consistent with the significance of the deer number attribute from the choice models. Over 70% of each stakeholder group in each region stated they would be concerned or very concerned if herd health decreased, and approximately 60% or more of respondents in each stakeholder group, with the exception of the northeast hunters, would be concerned or very concerned with an increase in deer-vehicle collisions. These results generally support the significance of herd health and deer-vehicle collisions in five of the six choice models. The majority of hunters and nonhunters in all regions stated that it was ‘somewhat important’ or ‘not important at all’ to decrease the amount of deer damage to residential property, though the split between somewhat important/not important at all and important/very important was closest for the northwest nonhunters. Similar results are demonstrated from the choice models, where deer damage to residential property was a non-significant attribute in all regional models except the northwest nonhunters. Other components of the survey support the conclusion that preferences for deer and deer-related extemalities differ among hunter and nonhunter stakeholder groups. For example, cross tabulation analysis with Chi-square test of significance suggest that, in all 116 regions, hunters tend to prefer more deer and tend to be more concerned with deer decreases than nonhunters. In general, cross tabulations also suggest that hunters tend to be more concerned than nonhunters with herd health, and nonhunters tend to be more concerned with deer damage to agriculture and residential properties, deer-vehicle collisions, and deer browsing in the forest. The sample population was segmented by other socioeconomic strata to determine whether respondent characteristics other than hunter or nonhunter are correlated with preferences for deer and deer-related extemalities. Few conclusions can be drawn from this analysis, as most of the significant correlations were weak (Pearson correlation coefficients < 0.12). Several of the “stronger” correlations suggest that, in all regions, respondents with more years of hunting tend to place more importance on decreasing deer damage to residential property, and respondents with less education tend to be more concerned about decreases in the number of deer. However, none of the correlation coefficients is greater than 0.12, thus the usefulness of segmenting respondents by strata other than hunter/nonhunter or region may be limited. The respondent population was also segmented by their experience with deer- related extemalities. Respondent experience with the extemalities varied by group and by regions, and cross tabulation analyses suggest that respondents with experience with deer damage to residential property, agriculture, and deer browsing in the forest tend to prefer less deer for their region than respondents without this experience. Over 74% of respondents in each group and each region had either themselves or someone in their immediate family experienced a deer-vehicle collision, which may be a reason for the 117 significance of the deer-vehicle collision coefficient in all of the choice models. In summary, there are differences and similarities among regions, between stakeholders, and among various segments of the respondent population in the relative importance placed on deer and deer-related attributes. This situation precludes a succinct list of definitive conclusions concerning each specific attribute; however, the list below, drawing from all components of the survey, summarizes the general conclusions pertaining to research question (1). Deer and deer-related extemalities are important to respondents, though their relative importance varies among regions and between stakeholder groups. For hunters, deer and mature bucks generally have a significant and positive effect on utility, while. for nonhunters deer generally have a non- significant positive effect. Hunters prefer more deer for their region than do nonhunters Hunters place more importance on the number of mature bucks than on deer numbers When statistically significant, deer-related extemalities generally have a negative effect on utility. Deer-vehicle collisions and herd health appear to be important to the majority of respondents. Residential property damage appears to be least important to the majority of respondents . Respondents with experience with deer-related extemalities tend to prefer less deer for their region than respondents without this experience. 118 Satisfaction and Issue Activity Related to Deer and Extemality Levels The conclusions in the section above are consistent with previous research documenting both positive attitudes towards deer in general and the importance of deer- related extemalities. However, in addition to examining the relative importance of deer and deer-related attributes, this research addresses additional questions concerning satisfaction and issue activity when the level of deer and deer-related attributes change, which may be useful in making management decisions. Research Question (2): What kinds of changes in deer populations and associated extemalities are most likely to satisfy diflerent stakeholder groups? Satisfaction models that regressed attribute levels in the choice experiment on a likert satisfaction rating were estimated at theistatewide and regional level. The models generally have low R-square values (< 0.13) and few significant coefficients, perhaps due, in part, to the limited amount of qualitative research devoted to the satisfaction questions. It is also possible that there is not sufficient variance in the independent variables to treat them as linear variables in the regression. Generally, the models suggest that satisfaction increases when the number of deer or mature bucks increase, and satisfaction decreases when extemalities increase, though there is a counter-intuitive Sign on deer-vehicle collisions in the southwest hunter model. The usefirlness of the satisfaction models in explaining variance is relatively limited. However, if the concept of utility, a central concept in addressing research question (3), can act as an indicator of satisfaction, then marginal rates of substitution and utility scenario predictions can be used to address respondent satisfaction when attribute 119 levels change. Research Question (3): What types of extemality increases will stakeholders accept for increases in the deer population, and conversely, what types of extemality decreases will compensate stakeholders for decreases in the deer population? The marginal rate of substitution (MRS) is the rate at which individuals will substitute changes in one attribute for changes in a second attribute such that their overall level of utility remains constant. As applied to preferences for and satisfaction with deer and deer-related extemalities, MRS can illustrate the increases in extemality levels that respondents will accept for increases in the deer population, and similarly, the extemality decreases needed to compensate respondents for deer population decreases. When comparing hunter and nonhunter MRS at the statewide level, results suggest that for an increase in the deer population, hunters will accept (utility will remain constant) 2.5 to 3 times the increases in deer-vehicle collisions and the percent of heavily browsed forest area than nonhunters; however, the two groups will accept about the same increase in the percent of deer in poor health. Both hunters and nonhunters would accept larger increases in the percent of heavily browsed forest area than in poor health and deer- vehicle collisions. At the statewide level, increases in mature bucks provide more utility than do increases in deer, as hunters will accept 2.5 to 3 times the extemality increases for increases in mature bucks versus increases in deer. There are some differences at the regional level, however. For example, in general hunters in the northeast and southwest will accept larger extemality increases for increases in deer than will hunters in the 120 northwest. For increases in mature bucks, the reverse is true. Hunters in the northwest will accept larger increases - more than twice as much for some extemalities - for an increase in mature bucks than hunters in the northeast and southwest. As discussed in Chapter 4, choice models can also be used to predict choice probabilities for different deer scenarios. Higher choice probabilities indicate higher utility levels associated with a given scenario, and may also indicate higher satisfaction levels. Utility scenario predictions were made for the four deer scenarios below, to examine the extent to which choices would vary between stakeholders. The deer scenarios are relative to the status quo of the region, and represent hypothetical changes for analytical purposes. Other scenarios can be constructed based on different predictions of management outcomes. Using the choice models to predict choice probabilities for each scenario shows that, when changes to bucks occur, for hunters and nonhunters in each region the highest probabilities are associated with Scenario D and the lowest with Scenario B, with the exception of nonhunters in the northeast, where Scenarios A and D were equally high, and Scenarios B and C were equally low. One reason for these results may be that Scenario D has the largest increase in buck numbers, which is important for hunters, but also the largest decreases in extemality levels, which may be important to the nonhunters. When the hypothetical management outcomes do not include changes to buck numbers, hunters in the northeast and northwest tend to prefer Scenarios B, while hunters in the southwest tend to prefer Scenarios A and B with about the same probabilities. 121 A E Q 2 Number of deer 5% more 8% more 5% less 8% less Number of mature bucks 3% less 5% less 3% more 5% more Percent of deer in poor health 3% more 8% more 3% less 8% less :zlggletniirgzgegfiagopemes 3% more 10% more 3% less 10% less Deer damage per acre of cropland 1% more 3% more 1% less 3% less Annual deer-vehicle collisions 2% more 4% more 2% less 4% less Percent of forest area experiencing o o o 0 heavy deer browsing 3 /o more 6 /o more 3 /0 less 6 /0 less Hunter Choice Probability Northwest 0.2066 0.1792 0.2852 0.3289 Northeast 0.2133 0.1896 0.2810 0.3161 Southwest 0.2022 0.1729 0.2881 0.3369 Hunter Choice Probability when no “Haggai? °°°‘" 0.2664 0.2689 0.2334 0.2313 Northeast 0.2718 0.2812 0.2273 0.2197 Southwest 0.2626 0.2613 0.2374 0.2386 Nonhunter Choice Probability Northwest 0.241 1 0.2227 0.2574 0.2787 Northeast 0.2538 0.2464 0.2462 0.2536 Southwest 0.2313 0.2066 0.2652 0.2969 This analysis demonstrates that, for the magnitude and direction of the changes proposed, hunters and nonhunters prefer the same management scenarios when changes to bucks occur. However, without changes to buck numbers hunters tend to pick scenarios with more deer and extemality increases while nonhunters tend to choose scenarios with less deer and extemality decreases. 122 In summary, several conclusions can be drawn in relation to research questions (2) and (3): - For a deer population increase, hunters will generally accept larger increases in extemality levels than will nonhunters. Herd health is an exception, and hunters and nonhunters will generally accept about the same decrease in herd health for an increase in the deer population. - Hunters prefer to have more mature bucks than more deer, and will accept larger extemality increases for increases in mature bucks. This pattern exists in each region, though it is strongest in the northwest. - Generally, increases in the annual number of deer-vehicle collisions and increases in herd health are the least acceptable type of extemality increase, followed by deer browsing in the forest. - Based on hypothetical management outcomes which increase and decrease attribute levels, hunters tend to prefer scenarios which increase mature bucks even when the deer numbers decrease. Nonhunters tend to prefer scenarios which decrease extemality levels, even when deer numbers decrease. Question (4): What kinds of changes in deer populations and associated extemalities are most likely to induce diflerent stakeholder groups to engage in issue activity? Similar to the satisfaction models, the models of issue activity based on the levels of deer and deer-related attributes performed poorly, perhaps due in part to lack of qualitative research devoted to the questions concerning issue activity. As suggested in Chapter 3, it may be more informative to examine the percentages of respondents who state they would engage in some form of issue activity when situations change fiom the status quo. For example, when deer and extemalities increase, about 30% of hunters and nonhunters in a region would engage in some form of issue activity, with no significant differences between the two groups. However, hunters and nonhunters differ 123 significantly when deer and extemalities decrease, with hunters being more likely to engage in issue activity. Approximately 40% of hunters in each region would engage in issue activity when deer, mature bucks, and extemalities decrease (decreases were between 0 -5% for mature bucks and 10 -20% for deer). Limitations of the Research In collecting preference information, survey respondents were provided with the status quo levels of deer-related attributes and then asked to make a choice about different deer scenarios for their region. It is assumed that respondents perceive the status quo information as credible, and make an informed choice using this information. It is worth noting that, on average, less than 25% of respondents did not disagree with information provided about the status quo. Further, if the respondents from the northeast are removed the figure drops to less than 20%. Some respondents in this region may be more skeptical of any information related to deer due to a somewhat unstable relationship with deer management professionals in recent years. Though in general the level of skepticism seems limited, any skepticism at all can affect decision-making. For example, it is possible that a respondent may be concerned about an attribute but, because the information presented is perceived as incorrect, they disregard the attribute altogether when making choices. This could render attribute parameters insignificant. For this survey, the best available information concerning deer-related attributes was presented to respondents; however, respondents were cautioned that, in some cases, the information was an estimated average for a three or four county region. More precise estimates, if 124 they existed, might improve credibility in the survey and subsequently increase the validity of survey results. In the future, any information managers and educators can provide concerning deer and deer-related attributes will improve efforts to examine stakeholder preferences. Generally, when survey results are applied for management purposes, response bias should be considered. Time and resources did not allow for any formal examination of response bias; however, for this survey, the effects of response bias may be mitigated by two factors. First, the response rates were relatively high (62% for the general public and 66% for hunters), particularly given that the first mailing occurred on September 11, 2001. Second, because approximately 40% of public respondents stated that they had hunted deer, two distinct groups, hunters andnonhunters, were created to reduce bias by hunters on the public responses. It is possible that responses are biased by other characteristics. One obvious characteristic would be that only respondents with an overt interest in deer or experience with deer-related attributes returned the survey. Generally this does not appear to be the case. For example, only about 29% of hunters and 10% of nonhunters stated that they frequently took a drive or walk to view deer, and less than 50% of respondents had experience with deer damage to residential property, deer damage to agriculture, or deer browsing in the forest. The survey has demonstrated preference inconsistencies across different elicitation formats, e. g. the 4-way versus the pooled nonhunter choice models. These differences, described in detail in Chapter 4, may be due to the difference in choice complexity or the difference in the number of observations used to estimate each model. 125 The pooled model was estimated from 2,337 observations, while the 4-way model had only 771 observations. The difference in observations is a function of survey design (and not item non-response), and it may, in part, be responsible for the preference inconsistencies. Deer management should be cautious in generalizing the precise choice model results outside of the study regions. The statewide model is based only on the aggregate of observations from the three regions. Each of these regions is fairly unique, and relatively rural - particularly the northeast and northwest regions. Further, preferences differed, at times considerably, among regions. Noting this, the use of exact parameter estimates for analyses outside of the study regions may be inappropriate; however, using results as general guidance outside of the study area may be suitable in some situations. The Survey Instrument and Choice Experiment Format In developing the choice experiment survey instrument, a fairly extensive qualitative research phase was undertaken, consisting of three focus groups and sixty two in-person pretests. The qualitative research was vital in identifying problems in the survey instrument, and the strength of this phase ultimately led to a credible survey instrument. This section describes the insight gained during the qualitative research and presents some of the issues related to the survey instrument that emerged after the CE data was analyzed. Some CE’s ask respondents to make choices among attributes with which they are relatively familiar. In other CE’s respondents’ knowledge of the attributes in question 126 may be limited, or the way in which the attributes are quantified may be unfamiliar. The latter was the case in the white-tailed deer CE. Qualitative research suggested that, while people generally had a conceptual understanding of most of the attributes, choosing a metric to express different levels of the attributes presented a challenge. Feedback during the survey development phase proved crucial in identifying and describing a metric to quantify attribute level changes. In addition to the above, the qualitative research phase identified problems that occurred when all attributes were allowed to vary independently, e. g. the creation of counterfactual situations, and the problem of respondents choosing a scenario based only on the number of deer. Feedback from focus groups and pretest interviews induced, in part, the experimental design plan, which divided the choice sets into “increasing,” “decreasing,” and “marginal” sets and added a third alternative to each set. Thus in keeping with other researchers who stress the importance of a qualitative phase, the attention given to qualitative research in the white-tailed deer CE was critical in developing a credible survey instrument. Given the scope of the research project and the time and resource constraints, the qualitative research phase was relatively extensive, and was used to obtain feedback on issues that were critical to the choice experiment. However, there are several survey items that may have benefitted from more attention during qualitative research, including endpoint levels for attributes, and satisfaction and issue activity questions. For example, endpoint levels for changes to the deer population were set at 20% and 30% more or less for increasing and decreasing scenarios. It is possible that these changes are not dramatic 127 enough to induce people to consider the attribute when making choices, and thus deer numbers was not a significant attribute in several of the regional models. The same may be true for other nonsignificant attributes. Questions concerning the magnitude of attribute level changes were not formally incorporated into the qualitative research, and more feedback from focus group and pretest participants would have been useful. Additionally, the questions concerning satisfaction and issue activity, which followed choice questions but are not directly related to the CE, needed further development. This is evidenced in part by the poorly performing satisfaction and issue activity models. Admittedly neither the satisfaction nor issue activity scales were given much attention in the qualitative research. An additional issue for discussion in focus groups or pretest interviews concerns a mechanism for attribute level changes. In the survey respondents were given limited information on how the changes to the status quo might occur in the alternative scenarios, with a caveat that... “alternative scenarios may not always seem logical, but they are projections of possible deer scenarios...” Explanations or mechanisms for how the changes may arise could be tested to see if they enhance the credibility of the alternative scenarios. On the other hand, information of this type may result in respondents valuing a mechanism for change, and not necessarily the attributes themselves. Given additional time and resources, qualitative research may have provided insight into this issue. The survey format was the same for all respondents in that each choice scenario contained the status quo and two or three alternatives, depending on whether a respondent received the SOS or MDNR survey version. In either case, the current situation, or status 128 quo, was always presented in the first (left-most) column, followed by the two or three alternatives. Ordering problems may arise from this positioning. For example, respondents may simply choose the first scenario available, which in this case would always be the current situation. The white-tailed deer CE data do not support this, however, as the current situation was chosen most frequently (by the SOS respondents) when the alternatives were increasing, but not when alternatives were decreasing. When the alternatives involved marginal changes, the current situation and an alternative were chosen with similar fi'equencies. However, what does appear to occur due to this ordering is that, for those respondents who chose other than the status quo, the majority of them always chose the first alternative presented. This could suggest that respondents who already know they do not want the status quo simply choose the first available alternative. Focus groups and/or pretest interviews could be conducted to detennine whether the ordering of scenarios appears to have an effect on choices, or alternatively, the survey sample could be split to test for an “ordering” effect. However, the latter requires a larger number of survey versions to be developed and adds considerably to printing and coordination costs. In summary, the survey instrument benefitted from an extensive qualitative research phase, given time and financial constraints, and the importance of the qualitative phase should not be underestimated. 129 The Big Picture for Deer Management At the most general level, this research has demonstrated two very important points for deer management. First, management Should consider the preferences of more than just deer hunters when developing policies, as deer provide positive utility to both hunters and nonhunters. Second, while hunters do care about deer, they also care about the extemalities associated with deer, and neither hunters nor nonhunters want increases in the deer population at any cost (e. g. more deer-vehicle collisions, more deer browsing in the forest, etc...). Two of the costs that are least acceptable to both hunters and nonhunters are deer-vehicle collisions and poor herd health, while residential property damage appears to be the most acceptable cost associated with deer. As expected, hunters generally want more deer than nonhunters, but they are also willing to accept higher (but not unlimited) costs, with one exception. Results show that nonhunters are just as concerned with herd health as are hunters, and both groups are willing to make similar trade-offs between herd health and increases in deer populations — for a 1% increase in deer numbers both groups could incur about 0.4% more deer in poor health without a change in utility. Ifmanagement seeks to reduce deer populations, this finding emphasizes the importance of demonstrating the relationship between herd health and population size to the public. Another finding that should be of interest to management is the importance hunters place on mature bucks. Choice model results show that hunters will incur 2.5 to 3 times the cost, in terms of increased extemalities, for increases in mature bucks than for increases in deer. Further, hunters are most likely to choose deer scenarios with the 130 largest increase to mature bucks for their region, even if the number of deer decrease. It is interesting to compare these results with the results of MDNR surveys concerning Quality Deer Management (QDM) (F rawley 2003). As stated by the MDNR, the goals of QDM include maintaining a balanced sex ratio in the deer herd, keeping the herd in balance with the habitat, and increasing the number of older-aged bucks. QDM survey results suggest that, in 5 out of 6 deer management units surveyed, there is insufficient support to recommend QDM for implementation, or for continued antler point restrictions in harvest. One reason for the disparity between the choice model and QDM survey results may be the lack of information provided in the choice experiment survey addressing how a region will get more mature bucks. While the CE survey does not describe this mechanism, the QDM surveys discuss specific mechanisms for attaining goals, most of which pertain to harvest restrictions. The different survey formats and results may suggest that, while hunters would like more mature bucks, they do not want to incur additional harvest restrictions to get them. However, if an increased number of mature bucks is a management goal, it may be possible to use the strong preferences for herd health to gain support for QDM. For example, the QDM surveys do not specifically discuss the overall health of the herd in relation to QDM. Emphasizing any relationships between improved herd health and QDM may generate increased support for the management concept. The survey instrument produced consistent results across most types of measurement. For example, when respondents were asked how many deer they would like in their region, hunters wanted about 25% more than last year and nonhunters wanted 131 about the same as last year. This finding is consistent with the choice model results indicating that deer provide positive utility to both stakeholder groups. Further, attitude questions indicate that respondents are less concerned about deer damage to agriculture and residential property than they are about deer browsing in the forest, deer-vehicle collisions, and herd health, a finding borne out by the choice models as well. When models were estimated at the regional level, the significance of the attributes varied by region. For example, choice models suggest that deer browsing in the forest is not significant to hunters in the northeast but significant to hunters in the other two regions. This finding is consistent with the fact that more hunters in the southwest and northwest agreed with the statement that ‘the effects of deer browsing in the forest are significant’ than did hunters in the northeast. Also, more nonhunters in the northwest agreed with the statement that ‘there is a significant amount of deer damage to residential properties’ than did nonhunters in the other regions, a finding supported by the significance of that attribute in the nonhunter northwest regional choice model. The consistency across different types of preference measurement underscores the credibility of the survey results and the management implications derived from the results. In addition to knowledge of preferences for deer and deer-related attributes, deer managers may need to know what types of changes are likely to produce issue activity, and subsequently, the cultural carrying capacity for deer (see Chapter 1). Although the issue activity models from Chapter 4 performed poorly, using other components of the survey some conclusions about issue activity and cultural carrying capacity can be drawn: 132 ° Generally, nonhunter issue activity is limited to situations when deer and extemalities increase, while hunters tend to engage in issue activity when deer numbers decrease and when extemalities increase. However, the largest amount of total issue activity (combining hunters and nonhunters in all regions) arises when deer, bucks, and extemalities increase. 0 Among all extemalities, decreases in the percent of deer in poor health and the number of deer-vehicle collisions would provide the strongest “compensation” for a decrease in the deer population, while a decrease in deer damage to residential property would likely provide the weakest “compensation.” - Based on the hypothetical outcomes used to simulate the choice models, a win-win deer scenario for hunters and public is a scenario that decreases the number of deer and the level of deer-related extemalities but increases the number of mature bucks. Given the magnitude of the mature bucks coefficient it seems reasonable that additional simulations, based on a realistic range of attribute outcomes, will produce similar results. In conclusion, using all components of the survey, several management recommendations are presented below: - Consider the preferences of both hunters and nonhunters when setting deer management goals. Results suggest that, all else equal (e. g. extemalities remain at the status quo level), more deer will increase the well-being of both groups. - Recognize that hunters as well as nonhunters do not want more deer at any cost, and both groups can make trade-offs between deer population size and extemalities. - Preference differences among regions reinforce the existing design of small deer management units, and illustrate the need for the human as well as biological aspects of deer management to be incorporated at small scales. - Outreach and education efforts would be well spent demonstrating the relationship between herd health and deer population size. It may also be useful to direct some effort at minimizing deer-vehicle collisions. 133 ° Hunters appear to be more likely than nonhunters to engage in issue activity. However, the least amount of issue activity is undertaken when mature bucks increase and deer-related extemalities decrease, even if the overall size of the deer population decreases. Because deer and deer-related extemalities matter to the public, future research concerning cultural carrying capacity as defined in Chapter 1 may want to explore mechanisms to better link deer and extemalites to the concept of issue activity. Further research could be undertaken concerning the scale used to measure issue activity, the changes required to induce an individual to engage in some form of issue activity, and other unexplored variables that may affect the likelihood of issue activity. In addition, the relationship between respondent characteristics, experience/attitudes, and preferences for deer populations could be examined. Stated choice techniques are becoming increasing popular as a method to incorporate a human dimension into natural resource management. The format places individuals in situations where trade-offs must be made, a common situation that many resource managers face. This research shows that choice experiment surveys can generate a significant amount of information, not only from the choice models themselves but from the additional survey components provided to help people make an informed decision. 134 APPENDIX A SURVEY INSTRUMENTS 135 Hunter Version f LIVING WITH DEER IN THE NORTHWESTERN UPPER PENINSULA OF MICHIGAN A Survey Conducted by Michigan State University § Marquette Marquette, Baraga, Iron, and Dickinson counties. Region 1 136 Michigan State University is conducting surveys about deer management in many areas of Michigan. We are asking you about deer management in the region of Marquette, Baraga, Iron, and Dickinson counties. These four counties are referred to as the Northwest (NW) region. Deer management in the NW region affects YOU because these decisions affect issues such as the number of deer in the region available for viewing and hunting, deer-vehicle collisions, or deer damage to forestry and agriculture. Because these issues affect you, it is important that your opinions and concerns are heard by state deer management professionals. Taking a few minutes to fill out this survey will help ensure that your input about deer management in the NW region is heard. Please complete this survey if: 0 you live in Marquette, Baraga, Iron, or Dickinson county OR hunt deer in any of these counties If you do not live or hunt deer in Marquette, Baraga. Iron, or Dickinson county please check the box below and return the survey. [:1 I am not a current resident and do not hunt deer in Marquette. Baraga. Iron, or Dickinson county. 137 ‘ Section I. Aspects of the Deer Herd . 1. During the last year. approximately how often did you . . . (circle only one) ....See deer in your own yard or in Never Rarely Sometimes Frequently your neighborhood ....Take a drive or a walk for the Never Rarely Sometimes Frequently specific purpose of viewing deer NUMBER OF DEER 2. In the diagram below. the letter E represents the number of deer last year in the NW region. Please circle the letter that best represents the number of deer that you would like to have in future years in the NW region. (Circle one letter below. If you are not sure or don’t care. check the appropriate box below). Nonem" 50%less‘ : Same as ‘. 50% more Eil’wlce as? ; last year i manyor " ' “i ‘ . more 1‘ A e I c o E ‘_ IA Fl-w 1 -5 Less More 0 I am unsure C] I don't care 3. How ooncemed would you be if the number of deer in the NW region decreased in future years by 20%? (Check only one) 0 Very ooncemed D Concerned 0 Somewhat ooncemed 0 Not ooncemed at all 138 NUMBER OF NATURE BUCKS 4. In the diagram below, the letter E represents the number of mature bucks in the NW region that are two and a half years or older and have at least four antler points on one side. Please circle the letter that best represents the number of mature bucks that you would like to have in future years in the NW region. (Circle one letter below. If you are not sure or don't care. check the appropriate box below). .‘ '- None - ;":..5'0% lessen? Same as 50% more '; iiTwlce‘asié I“ 3 ~ that year , many or , 7: ' ' ‘ more 4.:- A . a c ; o E 5 F e H i; . Less More Glam unsure 0 I don’t care 5. How ooncemed would you be if the number of mature bucks in the NW region decreased in future years by 20%? (Check only one) 0 Very ooncemed Cl Concerned 0 Somewhat ooncemed Cl Not concerned at all 139 7. (NFORNATION ABOUT HERD HEALTH Herd health refers to the physical well-being of the deer herd. e The health of a deer herd may be considered excellent. even though a few individual deer may be in poor health. 0 When the number of deer in poor health increases, the health of the deer herd decreases. When a deer is in POOR HEALTH it may have AT LEAST ONE of these characteristics: 0 Smaller body size than expected 0 Low reproductive success 0 Disease 0 Lower chance of surviving long winters The percent of deer in the herd that have AT LEAsT ONE of the characteristics of poor health varies. depending on the region of the state. There are no EXACT figures. but professionals ESTINATE that: 0 In Michigan the percent of deer with AT LEAsT ONE of the characteristics of poor health ranges from 5% to 50%. e In the NW region approximately 25% Of the deer have AT LEAsT ONE of the characteristics of poor health. Howconcernedwouldyou beifthe percentofdeerinpoorhealthintheNW region increased from the current level of 25% to 35% in future years? (Check only one) 0 Very ooncemed D Concerned 0 Somewhat ooncemed 0 Not concerned at all Based on your own opinion and experience with deer. how would you rate the health of the deer herd in the NW region? (Check only one) 0 Excellent 0 Fair 0 Good 0 Poor 0 I am unsure 140 (NEORNATION ABOUT DEER DANAOE To ResiosNTIAL PROPERTY Residential property damage is caused by deer feeding on plants. trees, and shrubs that people plant in their yards. 0 Professionals esTiNATE that approximately 30% of the residential properties in the NW region experience some deer damage. e This is an ssTNIATEo AVERAGE for the W- Damage may be higherin areas where there are moredeerand Iowerin areas with less deer. 8. Have you experienced residential property damage from deer? DYOS » 8a. lers,hasthedeerdamageceusedyoutO 0N0 changethetypesof plants in your yard? DYes DNO‘ D I am unsure 9. How important would it be to you if the percent of residential properties in the NW region experiencing some deer damage decreased from the current level of 30% to 10% in future years? (Check only one) 0 Very important 0 Important 0 Somewhat important 0 Not important at all 141 «FORMATION ABOUT DEER DAMAGE To AGRICULTURAL CROPS Deer damage to agriculture is caused by deer feeding on agricultural crops. e There are approximately 47,000 acres of cropland in the NW region. e Professionals esTINATE that deer cause approximately $6 of damage per acre of cropland in the NW region. 0 This is an Es'mIATEO AVERAGE for WW Damage may be higher for certain crops and lower for others. Damage may also be higher in areas where there are more deer and lower in areas with less deer. 10. Haveyouorhassomeoneyou knowinthe NWregionexperienced deerdamageto agricultural crops? 0 Yes I...» 10a. If Yes. has the damage resulted in: D No (Check only one) 0 Major economic loss 0 Moderately important economic loss 0 Relatively unimportant economic loss 0 i am unsure 1 1. How concerned would you be if the average amount of deer damage to agricultural crops in theNWregionincreasedfromthecurrentlevelofSBperacretOSQperacreinfutureyears? (Check only one) a Very ooncemed D Concerned 0 Somewhat concerned 0 Not ooncemed at all 142 IMPORNATION ABOUT DEER-VEHICLE COLusioNs Deer-vehide collisions refers to the number of reported collisions between deer and automobiles during a given time period. e In 1999 in the STATE OF MiCHreAN there was a total of 67.689 reported deer-vehicle collisions in the state Of Michigan. e In the NW region there were 3.562 reported deer-vehicle collisions. e About65% ofthecollislonsintheNWregionoccurredoncountyorlocal roads. 9 About 35% of the collisions in the NW region occuned on Interstates. US Routes. or State Routes. 12. Before reading about deer-vehicle collisions. did you know that the majority of reported deer- vehlcle collisions in the NW region occurred on county or local roads? DYes 0N0 D I am unsure 13. Have you or anyone in your immediate family been involved in a deer-vehicle collision? 0 Yes ONO DI am unsure 14. How important would it be to you if the annual number of reported deer-vehicle collisions in the NW region decreased from the 1999 level of 3,562 to 2.000 collisions in future years? (Check only one) 0 Very Important 0 Important 0 Somewhat important 0 Not important at all 143 “FORMATION ABOUT DEER AND THE FOREsT In a forest. deer generally feed, or browse, on the plants. shrubs, and tree seedlings. Over time. deer browsing may change the types of plants. trees. and animals that live in the forest. The extent of the changes depends on how much browsing occurs in the forest. In forest areas that experience heavy deer browsing for 5 - 10 years: 0 Sometypesofwiidflowersmaybeeliminated.whiiesornegmssesandfemsmay increase. e Some trees. like white cedar and maples. may be eliminated. while trees like spruce and fir may increase. a The elimination of certain kinds of trees may cause losses in the commercial forest industry. e Some birds. like warblers may be eliminated. while cardinals and bluejays may increase. 0 In general. the habitat will support fewer kinds of plants and wildlife. Besides the changes described above, many scientists believe that changes in the forest may also cause the forest to function differently in the long-term. with uncertain consequences. For example. the forest may be less able to adjust to events like fires and floods. 15. Have you seen any examples of heavy deer browsing in the NW region? 0 Yes D No a I am unsure 144 0 There are no EXACT figures. but professionals ESTIMATE that approximately 30% of forest areas in the NW region experience heavy deer browsing. 0 This is an EsTTMATEo AVERAGE for the entire region — damage may be higher in areas where there are more deer and lower in areas with less deer. How concerned would you be if the percent of forest area in the NW region that experiences heavy deer browsing increased from the current level of 30% to 40% in future years? (Check only one) 0 Very ooncemed D Concerned E] Somewhat ooncemed E] Not concerned at all For each of the following statements, please circle the response that comes closest to your point of view. Strongly Agree Neither Disagree Strongly Agree Agree nor Disagree Disagree b. I am ooncemed about heavy deer browsing in the forest because of the potential long-term effects to the forest 145 F ( Section III. Deer Scenarios . The main purpose of this survey is to find out what aspects of the deer herd are important to you. To do this we will show you scenarios and ask you to choose the one you prefer for the NW region. The previous pages provided information about the current deer situation in the NW region. In the next few pages you will see the current deer situation and two altemative deer scenarios. Here is an example of what you will see on the following pages: Please Compare Deer Scenarios 1, 2, and the Cunent Deer Situation: Current Scenario 1 Scenario 2 Sltuatlon “J Number of Deer Current 20% more than 20% more than number current number cunent number nun-l Number of Mature Bucks Current 10% more than 10% more than a. number current number current number Percent of Deer With At Least One 25% 25% 25% 2 Characteristic of Poor Health Percent of Residential Properties 30% 30% 30% Experiencing Some Deer Damage >< Deer Damage per Acre of Cropland $6 $7 $8 m Number of Deer-Vehicle Collisions 3.562 3.700 3,600 Percent of Forest Areas 30% 30% 30% Experiencing Heavy Deer Browsing Following each table. you will be asked which scenario you prefer for the NW region. A series of these scenarios will be presented and the items in each scenario will vary. The aitemative scenarios may not always seem logical. but they are projections of wills deer scenarios for the NW region. It is very important that you carefully consider the scenarios MW when you make your choice. This will provide the best information to state deer management professionals to help them design deer management policies. 146 18. Please compare Deer Scenarios A. B. and the Current Deer Situation: Number of Deer Number of Mature Bucks Percent of Deer With At Least One Characteristic of Peer Health Percent of Residential Properties Experiencing Deer Damage Deer Damage per Acre of Cropland Number of Deer-Vehicle Collisions Percent of Forest Areas Experiencing Heavy Deer Browsing Which of these do you prefer? C] Current 0 A n B 19. How satisfied would you be If Scenario Awas the situation In the NW region? (Check only one) 0 Extremely u Somewhat 0 Neither Satisfied 0 Somewhat 0 Extremely Satisfied Satisfied nor Dissatisfied Dissatisfied Dissatisfied 20. If Scenario A was the situation in the NW region, which of the following would you likely do? (Check all that apply) 0 Do nothing because it would not change the situation 0 Do nothing because the situation would not be that bad a Contact someone with authority to get the situation changed El Take steps myself to change the situation 147 21. Please compare Deer Scenarios C. D. and the Current Deer Situation: Number of Deer Number of Mature Bucks Percent of Deer With At Least One Characteristic of Poor Health Percent of Residential Properties Experiencing Deer Damage Deer Damage per Acre of Cropland Number of Deer-Vehicle Collisions Percent of Forest Areas Experiencing Heavy Deer Browsing Which of these do you prefer? a Current u c u D 22. How satisfied would you be if Scenario C was the situation in the NW region? (Check only one) 0 Extremely CI Somewhat [1 Neither Satisfied 0 Somewhat 0 Extremely Satisfied Satisfied nor Dissatisfied Dissatisfied Dissatisfied 23. If Scenario C was the situation in the NW region. which of the following would you likely do? (Check all that apply) [3 Do nothing because it would not change the situation a Do nothing because the situation would not be that bad 0 Contact someone with authority to get the situation changed [3 Take steps myself to change the situation 148 24. Please compare Deer Scenarios E, F. and the Current Deer Situation: Number of Deer Number of Mature Bucks Percent of Deer With At Least One Characteristic of Peer Health Percent of Residential Properties Experiencing Deer Damage Deer Damage per Acre of Cropland Number of Deer-Vehicle Collisions Percent of Forest Areas Experiencing Heavy Deer Browsing Which of these do you prefer? El Current 0 E a F 25. How satisfied would you be if Scenario E was the situation in the NW region? (Check only one) El Extremely 0 Somewhat [3 Neither Satisfied 0 Somewhat 0 Extremely Satisfied Satisfied nor Dissatisfied Dissatisfied Dissatisfied 26. If Scenario E was the situation in the NW region. which of the following would you likely do? (Check all that apply) D Do nothing because it would not change the situation 0 Do nothing because the situation would not be that bad [3 Contact someone with authority to get the situation changed 0 Take steps myself to change the situation 149 27. For each of the following statements. please circle the response that comes closest to your point of view. b. People can avoid most deer-vehicle collisions. d. The effects of deer browsing in the forest are significant In the NW region of Michigan. f. Adearwithatleastonecharacteristicof poor health has a disease. h. There is a significant amount of deer damage to agricultural crops in the NW region of Michigan. I. When comparing scenarios. I accepted that 30% of residential properties In the NW region of Michigan experience some deer damage. 31'0le MM Agree SA 150 r i Section V. Background Information ’ In this section we ask a few questions about your background so that we can compare our results to the state population. Your responses are completely confidential and will not be linked to your name in any way. 28. How many years have you lived in the NW region (Marquette, Baraga. Iron, or Dickinson county)? years 29. Which best describes the area where you live? (Check only one) Cl Many neighborhoods in a town or city 0 Scattered neighborhoods outside the town 0 Few neighborhoods in a rural area 30. How many years have you lived in the state of Michigan? years 31. Do you hunt deer in the NW region (Marquette. Baraga. Iron. or Dickinson counties)? 0 No (If No. Please skip to question 34) 0 Yes #> if Yes. which counties 32. Approximately how many years have you hunted deer in these counties ? years 151 33. 34. 35. 37. About how many days did you spend hunting deer in these counties last year? days Do you own property in the NW region (Marquette, Baraga, Iron, or Dickinson county)? 0 No (If No. please skip to question 36) D Yes ‘> If Yes. approximately how many acres? What are the uses of your property? (Check all that apply) 0 Primary residence 0 Recreational residence El Farming 0 Forest products 0 Hunting C] Other 0 Wildlife viewing What was your gross household income in 2000? (Check only one) a $0 to $14,999 a $35,000 - $49,999 0 $15,000 - $24,999 [3 $50,000 - $74,999 [3 $25,000 - $34,999 [3 Over $75,000 Is any of your income derived from the activities beloW? (Check all that apply) 0 Farming D Forestry El Tourism Cl None of these What is the highest level of formal education that you have completed? (Check only one) 0 Some high school 0 Associates Degree (2 year degree) Cl High School Graduate or equivalent 0 College graduate (Bachelors or 4 year degree) 0 Trade or Vocational School Cl Graduate or Professional degree 0 Some college 152 Thank you for helping us with this project. That completes the survey, but if there is anything else you would like to share with us concerning deer management please use the remainder of this page, or feel free to attach additional sheets to this survey. Please use the enclosed sticker to seal your completed survey, and drop the survey in the mall. Postage has already been paid. 153 Public Version r LIVING WITH DEER IN THE NORTHWESTERN UPPER PENINSULA OF MICHIGAN A Survey Conducted by Michigan State University Marquette -, a)"; it: Jen; «is z" {{{ Marquette, Baraga, Iron, and Dickinson counties. Region 25 154 Michigan State University is conducting surveys about deer management in many areas of Michigan. We are asking you about deer management in the region of Marquette, Baraga, Iron, and Dickinson counties. These four counties are referred to as the Northwest (NW) region. Deer management in the NW region affects YOU because these decisions affect issues such as the number of deer in the region available for viewing and hunting, deer-vehicle collisions, or deer damage to forestry and agriculture. Because these issues affect you, it is important that your Opinions and concerns are heard by state deer management professionals. Taking a few minutes to fill out this survey will help ensure that your input about deer management in the NW region is heard. Please complete this survey if: 0 you live in Marquette, Baraga. Iron, or Dickinson county OR hunt deer in any of these counties If you do not live or hunt deer in Marquette, Baraga, Iron, or Dickinson county please check the box below and return the survey. D I am not a current resident and do not hunt deer in Marquette. Baraga. Iron. or Dickinson county. 155 ( Section I. Aspects of the Deer Herd . 1. During the last year, approximately how often did you . . . (circle only one) ....See deer in your own yard or in Never Rarely Sometimes Frequently your neighborhood ....Take a drive or a walk for the Never Rarely Sometimes Frequently specific purpose of viewing deer NUMBER OF DEER 2. In the diagram below. the letter E represents the number of deer last year in the NW region. Please circle the letter that best represents the number of deer that you would like to have in future years in the NW region. (Circle one letter below. If you are not sure or don't care. check the appropriate box below). None 4 150% less?" Same as f" 50% more I :{Twice as l . , 7_ last year "I manyor i: " , "Ii ‘ 7‘ 'l ‘3 more 'i i a: ,5: ‘L ' 1' 1 f 41 ~__ A y B -' C ' D ’ ‘E. . " F ' i H E; r Less More [3 I am unsure Didon't care 3. How concerned would you be If the number of deer in the NW region decreased in future years by 20%? (Check only one) 0 Very ooncemed l3 Concerned 0 Somewhat ooncemed 0 Not ooncemed at all 156 4. “FORMATION ABOUT HERD HEALTH Herd health refers to the physical well-being of the deer hard. The health of a deer herd may be considered excellent, even though a few individual deer may be in poor health. 0 When the number of deer in poor health increases. the health of the deer herd decreases. When a deer Is in POOR HEALTH it may have AT LEASTONE of these characteristics: Smaller body size than expected Low reproductive success Disease Lower chance of surviving long winters The percent of deer In the hard that have AT LEAsT ONE of the characteristics of poor health varies. depending on the region of the state. There are no EXACT figures, but professionals ESTIMATE that: In Michigan the percent of deer with AT LEAST ONE of the characteristics of poor health ranges from 5% to 50%. In the NW region approximately 25% of the deer have AT LEAST ONE of the characteristics of poor health. How ooncemed would you be if the percent of deer In poor health in the NW region increased from the current level of 25% to 35% in future years? (Check only one) 0 Very ooncemed Cl Concerned 0 Somewhat concerned 0 Not ooncemed at all Based on your own opinion and experience with deer. how would you rate the health of the deer herd In the NW region? (Check only one) 0 Excellent 0 Fair Cl Good 0 Poor 0 I am unsure 157 “mm ABOUT DEER DAMAGE TO REsIOENTIAL PROPERTY Residential property damage is caused by deer feeding on plants, trees. and shrubs that people plant in their yards. 0 Professionals EsTIMATE that approximately 30% of the residential properties in the NW region experience some deer damage. e This Is an EsTIMATEO AVERAGE for the W. Damage may be higher in areas where there are more deer and lower in areas with less deer. 6. Have you experienced residential property damage from deer? D Yes * 6a. If Yes, has the deer damage caused you to D No change the types of plants in your yard? DYes- D No D I am unsure 7. How Important would It be to you if the percent of residential properties In the NW region experiencing some deer damage decreased from the current level of 30% to 10% in future years? (Check only one) 0 Very Important 0 Important 0 Somewhat important 0 Not important at all 158 “summon ABOUT DEER ems To Aemcuuum. CROPS Deer damage to agriculture ls caused by deer feeding on agricultural crops. e There are approximately 47,000 acres of cropland in the NW region. 0 Professionals ESTIMATE that deer cause approximately $6 of damage per acre of cropland in the NW region. 0 This is an Es'nMATEO AVERAGE for ell egg in the entlg mun, Damage may be higher for certain crops and lower for others. Damage may also be higher in areas where there are more deer and lower in areas with less deer. Have you or has someone you know In the NW region experienced deer damage to agricultural crops? 0 Yes #> 8a. If Yes, has the damage resulted in: D N 0 (Check only one) 0 Major economic loss 0 Moderately Important economic loss 0 Relatively unimportant economic loss Glam unsure How ooncemed would you be if the average amount of deer damage to agricultural crops in the NW region increased from the current level of $6 per acre to $9 per acre In future years? (Check only one) [3 Very ooncemed CI Concerned Cl Somewhat ooncemed Cl Not ooncemed at all 159 “FORMATION ABOUT DEER-VEHICLE COLLISIONS Deer-vehicle collisions refers to the number of reported collisions between deer and automobiles during a given time period. a In 1999 in the STATE OF MICHIGAN there was a total of 67,669 reported deer-vehicle collisions In the state of Michigan. e In the NW region there were 3.562 reported deer-vehicle collisions. 9 About 65% of the collisions in the NW region occurred on county or local roads. 0 About 35% of the collisions in the NW region occurred on Interstates. US Routes, or State Routes. 10. Before reading about deer-vehicle collisions. did you know that the majority of reported deer- vehicle collisions In the NW region occurred on county or local roads? DYes 0N0 Cl I am unsure 11. Have you or anyone In your Immediate family been Involved in a deer-vehicle collision? 0 Yes 0N0 I3 I am unsure 12. How important would it be to you If the annual number of reported deer-vehicle collisions in the NW region decreased from the 1999 level of 3.562 to 2.000 collisions in future years? (Check only one) 0 Very Important 0 Important 0 Somewhat important 0 Not Important at all 160 Wm ABOUT DEER AND THE FOREST In a forest, deer generally feed, or browse, on the plants, shrubs. and tree seedlings. Over time, deer browsing may change the types of plants, trees. and animals that live in the forest. The extent of the changes depends on how much browsing occurs in the forest. In forest areas that experience heavy deer browsing for 5 - 10 years: 0 Sometypesofwlldfiowersmaybeelknlnated.whilesomegmssesandfernsmay increase. 0 Some trees, like white cedar and maples, may be eliminated, while trees like spruce and fir may increase. 0 The elimination of certain kinds of trees may cause losses In the commercial forest Industry. 0 Some birds, like warblers may be eliminated, while cardinals and bluejays may increase. e In general, the habitat will support fewer kinds of plants and wildlife. Besides the changes described above. many scientists believe that changes in the forest may also cause the forest to function differently in the long-term. with uncertain consequences. For example, the forest may be less able to adjust to events like fires and floods. 13. Have you seen any examples of heavy deer browsing in the NW region? [3 Yes D No D lam unsure 161 0 There are no EXACT figures. but professionals ESTIMATE that approximately 30% of forest areas In the NW region experience heavy deer browsing. 0 This is an ESTIMATED AVERAGE for the emifilegieg — damage may be higher in areas where there are more deer and lower in areas with less deer. How concerned would you be if the percent of forest area in the NW region that experiences heavy deer browsing Increased from the current level of 30% to 40% In future years? (Check only one) a Very concerned 0 Concerned E] Somewhat ooncemed 0 Not concerned at all For each of the following statements. please circle the response that comes closest to your point of View. Strongly Agree Neither Disagree Strongly Agree Agree nor Disagree Disagree b. I am concerned about heavy deer browsing in the forest because of the potential long-term effects to the forest 162 ( ‘ Section ill. Deer Scenarios ’ The main purpose of this survey is to find out what aspects of the deer herd are important to you. To do this we will show you scenarios and ask you to choose the one you prefer for the NW region. The previous pages provided information about the current deer situation in the NW region. In the next few pages you will see the current deer situation and two altematlve deer scenarios. Hereisanmrnglgofwhetyouwillseeonthefoliowing pages: Please Compare Deer Scenarios 1, 2, and the Current Deer Situation: Current Scenario 1 Scenario 2 Situation Lu Number of Deer Current 20% more than 20% more than I number current number current number n- Percent of Deer With N Least One 25% 25% 25% Characteristic of Poor Health 2 Percent of Residential Properties 30% l 30% 30% Experiencing Some Deer Damage g Deer Damage per Acre of Cropland $6 $7 $8 Number of Deer-Vehicle Collisions 3,562 3.700 3,600 Percent of Forest Areas 30% 30% 30% Experiencing Heavy Deer Browsing Following each table. you will be asked which scenario you prefer for the NW region. A series of these scenarios will be presented and the items in each scenario will vary. The allemative scenarios may not always seem logical. but they are projections of possible deer scenarios for the NW region. It is very important that you carefully consider the scenarios W when you make your choice. This will provide the best information to state deer management professionals to help them design deer management policies. 163 1S. PleasecornpareDeerScenariosA.B.endtheCurrentDeerSltuation: Number of Deer Percent of Deer With At Least One Characteristic of Poor Health Percent of Residential Properties Experiencing Deer Damage Deer Damage per Acre of Cropland Number of Deer-Vehicle Collisions Percent of Forest Areas Experiencing Heavy Deer Browsing Which of these do you prefer? 0 Current 0 A D B 17. How satisfied would you be If Scenario Awas the situation In the NW region? (Check only one) DExtremeiy uSomeahat oNeitherethflsd 08am oExtremer Satisfied Satisfied norDissstlsflsd Dissatisfied Dissatisfied 18. if Scenario Awas the situation in the NW region. which of the following would you likely do? (Check all that apply) D Do nothing because it would not change the situation a Do nothing because the situation would not be that bad 0 Contact someone with authority to get the situation changed 0 Take steps myself to change the situation 19. Please compare Deer Scenarios C. D. and the Current Deer Situation: Number of Deer Percent of Deer With At Least One Characteristic of Poor Health Percent of Residential Properties Experiencing Deer Damage Deer Damage per Acre of Cropland Number of Deer-Vehicle Collisions Percent of Forest Areas Experiencing Heavy Deer Browsing Which of these do you prefer? 0 Current D c a D 20. How satisfied would you be if Scenario c was the situation in the NW region? (Check only one) QExtremely QSomewhat uNeitherSatisfled DSomewhet uExtremeiy Satisfied Satisfied norDissatisfled Dissatisfied Dissatisfied 21. If Scenario C was the situation in the NW region. which of the following would you likely do? (Check all that apply) 0 Do nothing because it would not change the situation oDonothingbeceusethesltuatlonwouldnotbetl'ratbad 0 Contact someone with authority to get the situation changed Cl Take steps myself to change the situation 165 22. Please compare Deer Scenarios E. F. and the Current Deer Situation: Number of Deer Percent of Deer With At Least One Characteristic of Peer Health Percent of Residential Properties Experiencing Deer Damage Deer Damage per Acre of Cropland Number of Deer-Vehicle Collisions Percent of Forest Areas Experiencing Heavy Deer Browsing Which of these do you prefer? 0 Current 0 E a F 23. How satisfied would you be if Scenario E was the situation in the NW region? (Check only one) C] Erdrernely 0 Somewhat 0 Neither Satisfied C] Somewhat D Eidremely Satisfied Satisfied nor Dissatisfied Dissatisfied Dissatisfied 24. ifScenarioEwasthesituation lnthe NW region. which ofthefoliowingwouidyou likelydo? (Check all that apply) 0 Do nothing because it would not change the situation DDonothingbeceusethesltuationwouldnotbethatbad QContactsomeonewithauthoritytogetthesituationchanged 0 Take steps myself to change the situation 166 25. Please compare Deer Scenarios A, C. E. and the Current Deer Situation: Number of Deer Percent of Deer With At Least One Characteristic of Peer Health Percent of Residential Properties Experiencing Deer Damage Deer Damage per Acre of Cropland Number of Deer-Vehicle Collisions Percent of Forest Areas Experiencing Heavy Deer Browsing Which of these do you prefer? 167 26. For each of the following statements. please circle the response that comes closest to your point of view. Strongly Agree Neither Disagree Strongly MM Acres nor Disagree b. People can avoid most deer-vehicle collisions. d.TheeflectsofdeerbrowsInginthe forestaresignificantintheNWreglonof SA Michigan. f. Adeerwlthatleastonecharacterlsticof poorhealthhasadisease. h. There is a significant amount of deer damage to agricultural crops In the NW region of Michigan. ). When comparing scenarios. I accepted that 30% of residential properties in the NW region of Michigan experience some deer damage. 168 r l Section V. Background lnforrnatlon ’ in this section we ask a few questions about your background so that we an compare our results to the state population. Your responses are completely confidential and will not be linked to your name in any way. 27. How many years have you lived in the NW region (Marquette, Baraga. Iron. or Dickinson county)? years 28. Which best describes the area where you live? (Check only one) 0 Many neighborhoods in a town or city 0 Scattered neighborhoods outside the town 0 Few neighborhoods in a rural area 29. How many years have you lived in the state of Michigan? years 30. Do you hunt deer in the NW region (Marquette, Baraga. Iron, or Dickinson counties)? 0 No (If No, Please skip to question 33) 0 Yes * If Yes, which counties 31. Approximately how many years have you hunted deer in these counties ? years 169 32. 33. 35. 36. 37. About how many days did you spend hunting deer in these counties last year? days Do you own property in the NW region (Marquette. Baraga. Iron. or Dickinson county)? 0 No (it No, please skip to question 35) D Yes » it Yes, approximately how many acres? What are the uses of your property? (Check all that apply) 0 Primary residence 0 Recreational residence 0 Farming 0 Forest products 0 Hunting D Other 0 Wildlife viewing What was your gross household income in 2000? (Check only one) 0 $0 to $14,999 C] $35,000 - $49,999 0 $15,000 - $24,999 C] $50,000 - $74,999 Cl $25,000 - $34,999 [3 Over $75,000 is any of your Income derived from the activities below? (Check all that apply) 0 Farming D Forestry 0 Tourism 0 None of these What Is the highest level of formal education that you have completed? (Check only one) 0 Some high school 0 Associates Degree (2 year degree) 0 High School Graduate or equivalent 0 College graduate (Bachelors or 4 year degree) 0 Trade or Vocational School 0 Graduate or Professional degree 0 Some college 170 Thank you for helping us with this project. That completes the survey. but if there is anything else you would like to share with us concerning deer management please use the remainder of this page, or feel free to attach additional sheets to this survey. Please use the enclosed sticker to seal your completed survey, and drop the survey in the mail. Postage has already been paid. 171 APPENDIX B SURVEY CORRESPONDAN CE 172 Name Date, 2001 Address in a few days you will be receiving a questionnaire for a research project at Michigan State University. It will ask you about deer management issues that affect you as a resident of Marquette, Baraga, Iron, or Dickinson county. Your opinion and concerns are important to ensure that deer are managed to meet the needs of your area. We have found that many people like to know ahead of time that they will be receiving a questionnaire. As your time is valuable we will be enclosing a small token of our appreciation. We look forward to receiving your completed questionnaire. Sincerely, Kristy Wallmo Project Coordinator Michigan State University (printed on letterhead) 173 Name Date, 2001 Address We need your help with the enclosed survey about deer management in northwest Michigan. You may recall from a letter we sent you last week that the survey is part of an effort by Michigan State University to learn your opinions toward a variety of deer management issues. Results of the survey will provide guidance to the Michigan Department of Natural Resources in developing deer management policies that address the needs and concerns of residents of many different areas of Michigan. You might be wondering why we want your opinion, particularly if you are not a hunter or a wildlife enthusiast. Your input is vital because managing deer involves trade-offs that affect you. We realize that it takes time out of your day to fill out this survey, and have enclosed three first class stamps as a way of saying thank you for your help. Please take a few minutes to share your viewpoint by filling out this survey. All responses are completely confidential - your name and address will never be connected to your responses in any way. Rest assured, your privacy will be protected to the maximum extent allowable by law. if you have any questions or comments about this study feel free to call me at 517-432-5037. If you have any further questions concerning your rights as a survey respondent please contact Dr. David Wright, Chair of the MSU Committee on Research Involving Human Subjects, at (517) 355-2180. Thanks for participating in this study. Sincerely, Kristy Wallmo Project Coordinator Michigan State University (printed on letterhead) 174 Dear Sir or Madam: You were recently sent a questionnaire concerning deer management in your area. If you have returned the questionnaire, thank you. If you have not yet completed the questionnaire, please take a few minutes to do so now. Your input is important to ensure that deer are managed to meet the needs of your area. Sincerely, Kristy Wallmo Project Coordinator Michigan State University wallmokr@msu.edu (517) 432-5037 175 Name Date, 2001 We recently sent you a survey about deer management in your area of Michigan. Although we have received completed surveys from many of the residents that were selected from your area, to date we have not heard from you. I am writing to you again because your input is vital! Managing deer involves trade-offs that affect you and the people in your community. As a member of our scientifically designed study, we need to hear from you to make sure that our results are truly representative of residents in your area. Results of the survey will provide guidance to the Michigan Department of Natural Resources in developing deer management policies that address the needs and concerns of residents of many different areas of Michigan. Please take a few minutes to share your viewpoint by filling out this survey. We remind you that all responses are completely confidential - your name and address will never be connected to your responses. Rest assured, your privacy will be protected to the maximum extent allowable by law. If you have any questions or comments about this study feel free to call me at 517-432-5037. If you have any further questions concerning your rights as a survey respondent please contact Dr. David Wright, Chair of the MSU Committee on Research Involving Human Subjects, at (517)355-2180. Thank you for your contribution to the success of this study. Sincerely, Kristy Wallmo Project Coordinator Michigan State University (printed on letterhead) 176 Name Date, 2001 Address During the last two months we have sent you several mailings about deer management in your area of Michigan. Our study is drawing to a close, but we would like to make one final attempt to obtain your input. Managing deer involves many trade-offs that affect you and the people in your community, and that means your input is ESSENTIAL! By filling out the survey you are helping to provide guidance to the Michigan Department of Natural Resources in developing deer management policies that address the needs and concerns of residents in your area. Please take a few minutes to share your viewpoint by filling out this survey. We remind you that all responses are completely confidential - your name and address will never be connected to your responses. Rest assured, your privacy will be protected to the maximum extent allowable by law. If you have any questions or comments about this study feel free to call me at 517-432-5037. If you have any further questions concerning your rights as a survey respondent please contact Dr. David Wright, Chair of the MSU Committee on Research Involving Human Subjects, at (517) 355—2180. Thank you for your contribution to the success of this study. Sincerely, Kristy Wallmo Project Coordinator Michigan State University (printed on letterhead) 177 BIBLIOGRAPHY 178 BIBLIOGRAPHY Adamowicz, W., Louviere, J ., and Swait, J. 19983. Introduction to Attribute-Based Stated Choice Methods. Final Report submitted to the Resource Valuation Branch Damage Assessment Center, NOAA, US. Department of Commerce. Adamowicz, W., Boxall, P., Williams, M., and Louviere, J. 1998b. Stated Preference Approaches for Measuring Passive Use Values: Choice Experiments and Contingent Valuation. American Journal of Agricultural Economics, 80: 64-75. Alverson, W., and Waller, D. 1997. Deer Populations and the Widespread Failure of Hemlock Regeneration in Northern Forests. Pages 280-297 in W.J. McShea, H.B. Underwood, and H]. Rappole, eds. The science of overabundance: deer ecology and population management. Smithsonian Institution Press, Washington, DC. Alverson, W., Waller, D., and Solheim, S. 1988. Forests too deer: Edge effects in northern Wisconsin. Conservation Biology, 2: 348-58. Black, HC., Dimock, E.J, Evans, J ., Kao, C., and Rochelle, J .A. 1979. Animal Damage to Coniferous Plantations in Oregon and Washington: Part I. A Survey, 1963-75. Resource Bulletin 25, Forest Resources Laboratory, Oregon State University, Corvallis, OR. Bradley, M., and Daly, A. 1994. Use of logit scaling approach to test for rank order and fatigue effects in stated preference data. Transportation, 21: 167-184. Boxall, P., Adamowicz, W., Swait, J ., Williams, M., and Louviere, J. 1996. A comparison of stated preference methods for environmental valuation. Ecological Economics, 18: 243-253. Bull, P., and Peyton, RB. 1999. Stakeholder Preferences for Deer Management in DMU 15. Report submitted to Champion International Corporation, The Sand County Foundation, and Michigan Wildlife Bureau. Campa, H., Winterstein, SR, and Raymer, DP. 1996. Long-term Effects of Ungulate Browsing on Aspen Stand Regeneration and Structure. Michigan Dept. of Nat. Res., 1996. 230 pp. Final Report. Campa, H., Winterstein, S., Peyton, R.B., Dudderar, G., and Leefers, L. 1997. An Evaluation of a Multidisciplinary Problem: Ecological and Sociological Factors Influencing White-tailed Deer Damage to Agricultural Crops in Michigan. Transactions of the 62"” North American Wildlife and Natural Resource Conference. 179 Carpenter, L., Decker, D.J., and Lipscomb, J. 2000. Stakeholder Acceptance Capacity in Wildlife Management. Human Dimensions of Wildlife, 5: 5-19. Connelly, N.A., Decker, D.J., and Wear, S. 1987. Public Tolerance for Deer in a Suburban Environment: Implications for Management and Control. The 3rd Eastern Wildlife Damage Control Conference, Oct. 19-21, Gulf Shores, AL. Conover, M. 2002. Resolving Human-Wildlife Conflicts: The Science of Wildlife Damage Management. Lewis Publishers, CRC Press, Washington DC. Conover, M. 1997. Monetary and Intangible valuation of deer in the United States. The Wildlife Society Bulletin, 25(2): 298-305. Cristoffel, R., and Craven, S. 2000. Attitudes of woodland owners toward white-tailed deer and herbivory in Wisconsin. The Wildlife Society Bulletin, 28(1): 227-234. Curtis, J ., and Lynch, L. 2001. Explaining Deer Population Preferences: An Analysis of Farmers, Hunters, and the General Public. Agricultural and Resource Economics Review, 31(1): 44-55. deCalesta, D. 1997. Deer and Ecosystem Management. Pages 267-279 in W.J. McShea, H.B. Underwood, and H.J. Rappole, eds. The science of overabundance: deer ecology and population management. Smithsonian Institution Press, Washington, DC. deCalesta, D., and Stout, S. 1997. Relative deer density and sustainability: a conceptual framework for integrating deer management with ecosystem management. The Wildlife Society Bulletin, 25(2): 252-58. deCalesta, D. 1994. Impacts of white-tailed deer on songbirds within managed forests in Pennsylvania. Journal of Wildlife Management, 58: 711-718. Decker, D.J., and Brown, T. 1982. Fruit Growers vs. Other Farmers Attitudes Toward Deer in New York. The Wildlife Society Bulletin, 10: 150-155. Decker, D.J., and Gavin, T.A. 1985. Public Tolerance of a Suburban Deer Herd: Implications for Control. Eastern Wildlife Damage Control Conference 2: 192- 204. Decker, D.J., and Gavin, T. 1987. Public Attitudes Toward a Suburban Deer Herd. The Wildlife Society Bulletin, 15: 173-180. 180 Decker, D.J., and Purdy, K. 1988. Toward a concept of wildlife acceptance capacity in wildlife management. The Wildlife Society Bulletin, 16(1): 53-57. Decker, D.J., and Chase, L. 1997. Human Dimensions of Living with Wildlife — A Management Challenge for the 21“ Century. The Wildlife Society Bulletin, 25: 788-795. Diefenbach, D., Palmer, W., and Shope, W. 1997 . Attitudes of Pennsylvania sportsmen towards managing white-tailed deer to protect the ecological integrity of forests. The Wildlife Society Bulletin, 25(2): 244-251. Frawley, B. 2003. Quality Deer Management Survey: Deer Management Units 035 and A135. Michigan Department of Natural Resources, Wildlife Division Report 3388, Lansing, MI. Frawley, B. 2003. Quality Deer Management Survey: Deer Management Units 053. Michigan Department of Natural Resources, Wildlife Division Report 3389, Lansing, MI. Frawley, B. 2003. Quality Deer Management Survey: Deer Management Units 117. Michigan Department of Natural Rescurces, Wildlife Division Report 3391 , Lansing, Nfl. F rawley, B. 2003. Quality Deer Management Survey: Deer Management Units 059. Michigan Department of Natural Resources, Wildlife Division Report 3390, Lansing, MI. Frawley, B. 2003. Quality Deer Management Survey: Deer Management Units 118. Michigan Department of Natural Resources, Wildlife Division Report 3392, Lansing, MI. Frelich, L., and Lorimer, C. 1985. Current and predicted long-term effects of deer browsing in Hemlock forests in Michigan. Biological Conservation, 34(2): 99- 120. F ritzell, P. 1998. A survey of Michigan agricultural producers attitudes, perceptions, and behaviors regarding deer crop depredation to fruit, vegetables, and field crops. Thesis. Department of Fisheries and Wildlife, Michigan State University, East Lansing, MI. Gigliotti, L., Decker, D.J., and Carpenter, L. 2000. Developing the Wildlife Stakeholder Acceptance Capacity Concept: Research Needed. Human Dimensions of Wildlife, 5: 76-82. 181 Hanley, N., MacMillan, D., Wright, R., Bullock, C., Simpson, I., Parsisson, D., and Crabtree, B. 1998. Contingent Valuation versus Choice Experiments: Estimating the Benefits of Environmentally Sensitive Areas in Scotland. Journal of Agricultural Economics, 49(1): 1-15. Hanley, N., Wright, R., and Adamowicz, V. 1998. Using Choice Experiments to Value the Environment: Design Issues, Current Experiments, and Future Prospects. Environmental and Resource Economics, 1 1(3-4): 413-428. Healy, W. 1997. Influence of Deer on the Structure and Composition of Oak Forests in Central Massachusetts. Pages 249-266 in W.J. McShea, H.B. Underwood, and H.J. Rappole, eds. The science of overabundance: deer ecology and population management. Smithsonian Institution Press, Washington, DC. Hensher, D., Louviere, J., and Swait, J. 1999. Combining Sources of Preference Data. Journal of Econometrics, 89: 197-221. Lancaster, K. 1966. A New Approach to Consumer Theory. Journal of Political Economy, 74: 132-157. Langenau, E. 1994. 100 Years of Deer Management in Michigan. Michigan Department of Natural Resources, Wildlife Division Report 3213, Lansing, MI. Langeneau, E. 1993. 1993 Evaluation of the deer damage block permit program. Michigan Department of Natural Resources, Wildlife Division Report 3194, Lansing, MI. Lauber, T., Anthony, M., and Knuth, B. 2001. Gender and Ethical Judgements About Surburban Deer Management. Society and Natural Resources, 14: 571-583. Louviere, J ., Hensher, D., and Swait, J. 2000. Stated Choice Methods: Analysis and Application. Cambridge University Press, UK. Louviere, J ., and Timmermans, H. 1990. Stated Preference and Choice Models Applied to Recreation Research: A Review. Leisure Sciences, 12: 9-32. Marquis, DA. 1981. Effect of deer browsing on timber production in Allegheny hardwood forests of northwestern Pennsylvania. Northeast Forest Experiment Station, Broomall, PA. McCabe, T., and McCabe, R. 1997. Recounting Whitetails Past. Pages 11-26 in W.J. McShea, H.B. Underwood, and H.J. Rappole, eds. The science of overabundance: deer ecology and population management. Smithsonian Institution Press, 182 Washington, DC. McCreedy, C ., McCreedy, C., and McCreedy, P. 2001. Economic Impacts of White- tailed Deer in Indiana. FNR-160, Department of Forestry and Natural Resources, Purdue University, West Lafayette, IN. McShea, W., and Rappole, J. 2000. Managing the Abundance and Diversity of Breeding Bird Populations through Manipulation of Deer Populations. Conservation Biology, 14(4): 1 161-1170. McShea, W., and Rappole, J. 1997. Herbivores and the ecology of understory birds. Pages 298-309 in W.J. McShea, H.B. Underwood, and H.J. Rappole, eds. The science of overabundance: deer ecology and population management. Smithsonian Institution Press, Washington, DC. McShea, W., Underwood, B., and Rappole, J. 1997. Deer Management and the Concept of Overabundance. Pages 1-10 in W.J. McShea, H.B. Underwood, and H.J. Rappole, eds. The science of overabundance: deer ecology and population management. Smithsonian Institution Press, Washington, DC. McNew, K., and Curtis, J. 1997. Maryland Farmers Lose Bucks on Deer-Damaged Crops. Economic Viewpoints (December). Cooperative Extension Service, University of Maryland, College Park, MD. Michigan Department of Natural Resources. 1987. Deer-vehicle collisions in Michigan: A task force report. Wildlife Division Report 3072, Lansing, MI. Miller, S., Bratton, S., and Hadidian, J. 1992. Impacts of White-tailed Deer on Endangered and Threatened Vascular Plants. Natural Areas Journal, 12(2): 67- 74. Minnis, D., and Peyton, RB. 1993. Cultural Carrying Capacity: Modeling a Notion. Pages 19-34 in Jay B. McAninch, ed. Urban Deer: A Manageable Resource? Proceedings of a Symposium held at the 55th Midwest Fish and Wildlife Conference, December 12-14, St. Louis, MO. Morrison, G.C. 2000. WTP and WTA in repeated trial experiments: learning or leading? Journal of Economic Psychology, 21 : 57-72. Organ, J ., and Ellingwood, M. 2000. Wildlife Stakeholder Acceptance Capacity for Black Bears, Beavers, and Other Beasts in the East. Human Dimensions of Wildlife, 5:63-75. 183 Peyton, RB. 1984. A typology of natural resource issues with implications for resource management and education. Michigan Academician, 17(1): 49-58. Sayre, R.W., Decker, D.J., and Good, G. 1992. Deer Damage to Landscape Plants in New York State: Perceptions of Nursery Producers, Landscape Firms, and Homeowners. Journal of Environmental Horticulture, 10(1): 46-51. Stout, R., Knuth, B., and Curtis, P. 1997. Preferences of suburban landowners for deer management techniques: a step towards better communication. The Wildlife Society Bulletin, 25(2): 348-359. Stout, R., Stedman, R., Decker, D.J., and Knuth, B. 1993. Perceptions of Risk from Deer-related Vehicle Accidents: Implications for Public Preferences for Deer Herd Size. The Wildlife Society Bulletin, 21:237-249. Tilghman, N. 1989. Impacts of White-Tailed Deer on Forest Regeneration in Northwestern Pennsylvania. Journal of Wildlife Management, 53(3): 524-532. Train, K. 2003. Discrete Choice Methods with Simulation. Cambridge University Press, Cambridge, UK. Van Deelen, T., Pregitzer, K., and Haufler, J. 1996. A comparison of presettlement and present-day forests in two northern Michigan deer yards. American Midland Naturalist, 135(2): 181-194. Verme, L., and Johnston, W. 1986. Regeneration of Northern White Cedar Deeryards in Upper Michigan. Journal of Wildlife Management, 50(2): 307-313. Wisconsin Department of Natural Resources. 1999. Damage Abatement and Claims Program.Report prepared by Sarah Carter, Wisconsin Department of Natural Resources. Woolf, A., and Roseberry, J. 1998. Deer management: our profession’s symbol of success or failure? The Wildlife Society Bulletin, 26(3): 515-521. US. Fish and Wildlife Service. 1996. National Survey of Fishing, Hunting, and Wildlife-Associated Recreation. US. Fish and Wildlife Service, Washington DC. 184