1‘ Ant-(‘3 .. . xenunLuv.Vv..Ar: ”Fauna... .. 4’}; .M. , .55. a. a .J .fiV‘Mfl l .r I .. «awaits-"w...“- . HEAR. . .(L......1 £1133: 3......I\ .5... It!” . x... . . r £0... unmwwwkh. t..- .,._ at... «2...... . .81.... : ..~.. 53:... 1. .. .3. .. 1.1.1... ( 0.0.... I)!» v cl». 9\v 1v...“ .31.}. .I "n... In... .1! u (i. . . 139" ii “YEW: KEN! >33 [JIHRAJKY Michigan State University This is to certify that the dissertation entitled ANALYSES OF RECYCLING BEHAVIOR, RECYCLING DEMAND, AND EFFECTIVENESS OF POLICIES PROMOTING RECYCLING presented by SHAUFIQUE FAHMI SIDIQUE has been accepted towards fulfillment of the requirements for the Ph.D. degree in Agricultural Economics (7 - ” Major Professor’s Signature ‘1' / 2008 Date MSU is an affirmative-action, equal-opportunity employer -.n.-.-.--.-.--o-u-----~---‘- 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 AC1“ 1‘1 ‘1le "“' JN 1 7 2012 L..- 5l08 K:IProj/Acc&Pres/CIRCIDateDue.indd ANALYSES OF RECYCLING BEHAVIOR, RECYCLING DEMAND, AND EFFECTIVENESS OF POLICIES PROMOTING RECYCLING By Shaufique Fahmi Sidique A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics 2008 ABSTRACT ANALYSES OF RECYCLING BEHAVIOR, RECYCLING DEMAND, AND EFFECTIVENESS OF POLICIES PROMOTING RECYCLING By Shaufique Fahmi Sidique This dissertation analyzes the behaviors of recyclers, the demand for drop—off recycling and the effectiveness of policies promoting recycling. The first two essays in this dissertation are based on a survey of drop-off recyclers in the Lansing area in Michigan. The first essay studies the profile of people who utilize drop-off recycling sites and analyzes the factors influencing their site usage. The results show that the usage of drop-off recycling sites is influenced by demographic factors such as age, education, income and household size. Attitudinal factors are also found to affect site usage. Recyclers tend to use the drop-off sites more when they feel that recycling is a convenient activity and when they are more familiar with the sites. The second essay examines the demand for drop-off recycling sites as a function of travel costs and various site characteristics using the random utility model (RUM). The findings of this essay indicate that increased travel costs significantly reduce the frequency of visits to drop-off sites implying that the usage pattern of a site is influenced by its location relative to where people live. This essay also demonstrate that site specific characteristics such as hours of operation, the number of recyclables accepted, acceptance of commingled recyclables, and acceptance of yardwaste affect the frequency of visits to drop- off sites. The third essay addresses recycling rates. The effect of various recycling and waste management policy variables on recycling rate is assessed by utilizing a county level panel data from Minnesota. Our estimation procedure takes into account the potential endogeneity of these policy variables. The findings of this essay suggest that variable pricing of waste disposal increases the rate of recycling. Other policy variables such as the enactment of recycling ordinances and cumulative expenditures on recycling education are also found to be effective measures to increase recycling rate. To my beloved wife Husnita and my sons, lman and lhsan ACKNOWLEDGEMENTS There are many people I would like to thank for helping me during my doctoral study. I would like to express my greatest thanks and appreciation to Satish Joshi, my major supervisor for his support and guidance throughout my graduate program. His patience, kindness and friendship made my PhD experience an enjoyable one. I am extremely grateful to Frank Lupi, my co- supervisor who helped me a great deal in completing my dissertation and also for teaching me non-market valuation. I really appreciate his ‘willingness to accept’ my late night phone calls when I was writing my dissertation. I will be forever indebted to both of them. I am thankful to Susan Selke and Scott Loveridge for serving in my committee and for providing me valuable feedback on my dissertation. My gratitude to Eric Crawford for his advises and assistance on procedural and academic matters. I would also like to thank James Oehmke for serving in my committee during the initial stage of my dissertation and Scott Swinton for whom I had worked with at the beginning and towards the end of my study. I would like to thank the United States Environmental Protection Agency, Michigan State University Office of Vice President of Finance and Michigan State University Department of Agricultural, Food and Resource Economics for their financial support that contributed to the completion of my study. I would also like to thank Ann Bernstein and Mark Rust from the Minnesota Pollution Control Agency for providing me the waste and recycling data from Minnesota. I would like to express my appreciation to the professors in the Department of Agricultural, Food and Resource Economics from whom I have learned a lot about applied economics. I wish to also thank the departmental staff for their help throughout my time here. I would like to thank fellow graduate students in Cook Hall, particularly Elan Satriawan, Christopher Wright, Joshua Ariga, Athur Mabiso, Kwami Adanu, Abdoul Karim Murekezi, Sarma Aralas, Elliot Mghenyi and Yanyan Liu for their friendship and support. I am extremely grateful to my parents for their support and prayers throughout the years. I thank my brother Malique for his thoughts and wishes and my sister-in-law Haniza for her kindness to my family. I would like to express my deepest gratitude and love to my wife Husnita for all the sacrifices she made and for believing in me. Her love, support and prayers made this journey possible. Lastly -- to my little boys, I love both of you and thank you for being my source of everyday inspiration and bringing joy to my life. vi TABLE OF CONTENTS LIST OF TABLES ................................................................................................. ix LIST OF FIGURES ............................................................................................... xi INTRODUCTION .................................... ’ .............................................................. 1 ESSAY 1. THE EFFECTS OF BEHAVIORS AND ATTITUDES ON DROP-OFF RECYCLING ACTIVITIES ....................................................................... 3 1.1 Introduction ....................................................................................... 3 1.2 Literature Review .............................................................................. 4 1.3 Research Objectives and Hypotheses .............................................. 8 1.4 Methods .......................................................................................... 20 1.4.1 Questionnaire Design and Data Collection ............................. 9 1.4.2 Variables Description ........................................................... 10 1.4.3 Factor Analysis ..................................................................... 17 1.5 Analysis and Results ....................................................................... 22 1.6 Conclusions ..................................................................................... 28 References ............................................................................................ 29 ESSAY 2. ESTIMATING THE DEMAND FOR DROP-OFF RECYCLING SITES: A RANDOM UTILITY TRAVEL COST APPROACH .............................. 31 2.1 Introduction ..................................................................................... 31 2.2 Theoretical Framework .................................................................... 35 2.3 Survey Methods .............................................................................. 39 2.3.1 Questionnaire Design ........................................................... 39 2.3.2 Data Collection ..................................................................... 40 2.4 Model Estimation and Results ......................................................... 46 2.5 Scenario Analysis and Policy Implications ...................................... 53 2.6 Conclusions ..................................................................................... 58 References ............................................................................................ 72 ESSAY 3. FACTORS INFLUENCING THE RATE OF RECYCLING: AN ANALYSIS OF MINNESOTA COUNTIES ............................................. 63 3.1 Introduction ..................................................................................... 63 3.2 Literature Review ............................................................................ 65 3.3 Theoretical Framework .................................................................... 71 3.4 Data and Variable Descriptions ....................................................... 71 3.5 Econometric Methods and Results .................................................. 82 3.6 Conclusions ..................................................................................... 87 References .......................................................................................... 1 00 vii CONCLUSIONS ................................................................................................. 91 APPENDIX 1: Survey Instrument (Essays 1 & 2) ............................................... 93 APPENDIX 2: Tables (Essays 1 & 2) ............................................................... 102 APPENDIX 3: Probability Estimation of Variable Pricing Structure (Essay 3) .. 106 viii Table 1.1. Table 1.2. Table 1.3. Table 1.4. Table 1.5. Table 1.6. Table 1.7. Table 1.8. Table 2.1. Table 2.2. Table 2.3. Table 2.4. Table 2.5. Table 2.6. Table 2.7. Table 2.8. Table 2.9. Table 3.1. Table 3.2. Table 3.3. Table A2.1. Table A2.2. LIST OF TABLES Definition of variables ..................................................................... 11 Summary statistics of variables ...................................................... 12 Definition, distribution and descriptive statistics of Likert -scale variables ......................................................................................... 15 Total variance explained ................................................................. 17 KMO and Bartlett's test ................................................................... 19 Rotated factor matrix ...................................................................... 20 Factors, item variables and Cronbach’s alpha ................................ 22 Poisson regression ......................................................................... 25 Distribution of cars by sites from the effort survey .......................... 41 Distribution of respondents by sites from the on-site survey ........... 43 Demographic characteristics of drop-off recyclers .......................... 45 Definition of variables ..................................................................... 47 Random utility model results ........................................................... 50 Marginal implicit prices for changes in site characteristics .............. 53 Probability of site visitation ............................................................. 53 Probability of site visitation due to site closure ............................... 55 Probability of site visitation after changes in site attributes ............. 57 Definitions of variables .................................................................... 77 Summary statistics of variables ...................................................... 79 Determinants of the annual recycling rate in Minnesota counties... 84 Attributes of drop-off recycling sites ............................................. 103 Interview schedule ........................................................................ 104 Table A2.3. Distribution of cars according to sites ........................................... 105 Table A3.1. Logit estimation of the probability to implement variable .............. 107 LIST OF FIGURES Figure 1.1. Scree plot of eigenvalues and factors .............................................. 18 Figure 2.1. Locations and mean distance (miles) for drop-off recycling sites 57 Figure 3.1. Mean per capita municipal solid waste in Minnesota (1996-2004) .. 80 Figure 3.2. Mean per capita residential recycling in Minnesota (1996-2004) ..... 80 Figure 3.3. Mean per capita residential recycling rate in Minnesota (1996-2004) ......................................................................................................... 81 xi INTRODUCTION Recycling and municipal solid waste management are important environmental issues. Two hundred and fifty one million tons of municipal solid waste were generated in the United States in 2006 as compared to only 88 million tons in 1960 (USEPA, 2006). Concerns over this rising trend and increasing economic and social costs have prompted regulators and policymakers to introduce various policy initiatives aimed at reducing waste and increasing the amount of recycling. Among the policies and programs introduced are curbside recycling, drop-off recycling, source reduction efforts, recycling and waste legislation, and public education and campaign on recycling. Given the broad range of programs and policies available, there is a need to analyze the effectiveness of these practices. This dissertation contributes to recycling and waste management research by analyzing recycling behavior, demand and the effectiveness of several commonly adopted recycling policies and programs. The first and the second essay in this dissertation attempt to bridge the gap in the recycling and waste management research by studying drop-off recycling. Relative to curbside recycling research, there are a limited number of studies on drop-off recycling. The first essay analyzes the profile of recyclers utilizing drop-off recycling centers. This essay examines the relationships between drop-off recycling site visits and the socioeconomic and demographic characteristics of drop-off recyclers. This essay also examines behavioral aspects that influence drop-off site visits. The findings enhance understanding of the factors influencing participation and usage of recycling sites among drop-off recyclers. The second essay analyzes the demand for drop-off recycling sites in an urban area using the random utility model (RUM). The RUM model has been more commonly used in travel and recreation demand studies, and the essay is the first to apply the RUM method to assessing the demand for drop-off sites. The impact of different recycling site characteristics on the usage of drop-off sites is examined using the estimated model. The results from this essay provide policymakers a better understanding of the site characteristics that may influence the demand for drop-off recycling. The findings can also be used by policymakers and waste management companies to design, locate, and establish drop—off sites to increase site visits and collection of recyclables. The third essay analyzes the effectiveness of various recycling and waste management policy variables on recycling rate. This essay utilizes county level panel data from Minnesota covering the year 1996 to 2004. The policy variables examined include variable pricing for waste disposal, expenditure on recycling education, provision of curbside recycling services and drop-off centers, and enactment of recycling ordinances. This study accounts for the cumulative effects of the expenditure variable on recycling rate and also investigates whether different recycling programs such as curbside and drop-off recycling act as complements or substitutes in increasing recycling rates. This study also examines the effect of income and demographic characteristics on recycling rate. ESSAY 1. THE EFFECTS OF BEHAVIORS AND ATTITUDES ON DROP-OFF RECYCLING ACTIVITIES 1.1 Introduction In 2006, United States residents, businesses, and institutions produced more than 251 million tons of municipal solid waste. Recycling, including composting, was successful in diverting approximately 81 million tons of materials from disposal. However, more than 50 percent of solid waste generated is being Iandfilled (USEPA, 2007). To reduce the amount of waste entering landfills, policymakers and governments have implemented numerous recycling and waste reduction programs. Among the programs introduced are source reduction, curbside recycling and drop-off recycling programs. The success of a recycling program is however, largely dependent on household participation and sorting activities which are essentially behavior driven. A better understanding of recycling behavior will help us aid the design and improve the effectiveness of recycling policies. Various studies have been conducted on recycling behavior and areas examined include the motivational aspects that encourage people to recycle, the effect of socio-economic status and demographics on recycling, and the effect of knowledge and attitude on recycling (Vining and Ebreo 1990; Oskamp et al, 1991; Ebreo and Vining, 2001 ). These studies survey both recyclers and non-recyclers and draw conclusions based on the differences between these two groups. Studies have also examined only recyclers, concentrating on the behavior of this particular group of people. Speirs and Tucker (2001) studied the profile of recyclers utilizing drop-off recycling sites in Glasgow and across Ayrshire in south-west Scotland. They reported recyclers” travel distances, the weights and types of recyclables and demographic characteristics. The report was generally descriptive and did not establish relationships between drop-off site utilization and the profile of recyclers. There is also a growing interest in drop-off recycling research especially by policymakers and recycling and waste management service providers. In 2004, the Ohio EPA, for example, conducted a study of participation rates and usage patterns of recyclers at drop-off sites in Ohio. That study aimed to provide the empirical evidence required by the Ohio EPA to estimate the number of users utilizing recycling sites and to assess the percentage of population in a waste management distriCt with access to recycling facilities, to see whether it meets the regulated target percentage. We aim to add to this stream of behavioral literature by studying the profile of recyclers with a specific focus on individuals utilizing drop-off recycling sites. This is the first study that statistically analyzes the relationship between drop-off recycler characteristics and their number of visits to drop-off sites. 1.2 Literature Review There are many reasons and factors that influence recycling. Research findings suggest that attitudes, values and the extent of environmental knowledge or concern can be used to analyze recycling behavior (Domina and Koch, 2002). Studies have investigated the effect of factors such as convenience, motivations and general attitudes towards the environment, specific recycling attitudes, knowledge and social pressure on recycling involvement. Convenience is classified as an external facilitator that assists consumer recycling (Hornik et al, 1995). As recycling demands a significant amount of resources such as time, space, money and effort, making recycling convenient should increase household participation. In examining the differences between recyclers and non-recyclers, Vining and Ebero (1990) concluded non-recyclers were deterred by the inconvenience and the costs associated with recycling. Similarly, Domina and Koch (2002) in their study of textile recycling behavior reported that convenience is an important determinant distinguishing recyclers and non-recyclers. Saphores et al (2006) analyzed households’ willingness to recycle electronic waste at drop of centers and found that convenience factors such as proximity to the drop-of center would encourage recycling. Hornik et al (1995), based on a meta-analysis, concluded that frequency of recyclables collection, which is also convenience related, was a strong predictor of recycling behavior. Gonzalez-Torre et al (2003) examined selective waste collection systems that are frequently used in Europe and America and concluded that a system that requires less time effort to dispose and separate waste will result in a higher recycling rate. Concern for the environment is perceived to be important in encouraging recycling participation but empirical studies have shown mixed results. Domina and Koch (2002) found that people who have great concern for the environment were more likely to recycle. Meneses and Palacio (2005) in their study of the distribution of recycling tasks within the household reported that household members with positive attitudes towards ecology and who are motivated to protect the environment tend to bear a greater burden of the recycling role in the household. However, Vining and Ebero (1990) found that concern for the environment was indiscriminately expressed by both recyclers and non-recyclers. Similarly, Oskamp et al (1991) reported that general environmental attitude such as pre-ecological attitude and belief in the seriousness of environmental problems did not differ between recyclers and non recyclers. Knowledge about the availability of recycling programs and facilities is imperative for households to effectively participate in recycling. Studies have found that knowledge about recycling programs is a strong predictor of recycling involvement (Gamba and Oskamp, 1994; Hornik et al, 1995). Vining and Ebero (1990) found that recyclers are more aware of the publicity associated with recycling and more knowledgeable about the recycling facilities in the local area. Other studies have also tried to establish the importance of knowledge about the environment in encouraging recycling behavior. Oskamp et al (1991) reported that simple conservation knowledge predicts recycling participation. Studies have also investigated the effect social influence has on recycling behavior. Social influence in this context is defined as an individual’s concern about the perception of others, such as family and neighbors if they donot recycle (Vining and Ebreo, 1991 ). Oskamp et al (1991) and Do Valle et al (2005) reported that social influence is among the important factors that encourage people to recycle but Vining and Ebreo (1990) disagreed. Apart from behavioral aspects, numerous studies have also looked at the relationship between demographic and socioeconomic variables and recycling involvement. The most commonly examined variables are gender, age, education and income (Saphores et al, 2006). Meneses and Palacio (2005) argued that women bore a greater burden of recycling more often than men in the distribution of recycling tasks within a household. It has been argued that women are usually associated with recycling tasks because they traditionally have more authority as far as domestic tasks are concerned (Arcury, Scollay, and Johnson, 1987). Saphores et al (2006) found that women are more willing to recycle electronic waste at drop-off centers. However, other studies found no link between gender and recycling (Gamba and Oskamp, 1994; Werner and Makela, 1998). Other than gender, many studies have examined the relationship between age and recycling involvement. Some studies found age to be a significant factor influencing recycling involvement (Vining and Ebreo, 1990; Gamba and Oskamp, 1994; Margai, 1997; Scott, 1999; Saphores et al, 2006), but others did not (Werner and Makela, 1998; Meneses and Palacio, 2005). Contrary to popular expectation that younger people are likely to be more involved in recycling, researches have concluded that middle aged and older peOple are more likely to recycle (Vining and Ebreo, 1990; Meneses and Palacio, 2005; Saphores et al, 2006). The relationship between education and recycling is ambiguous. Saphores et al (2006) found that higher education increases the willingness to recycle but several other studies reported that education has no significant effect in influencing recycling behavior (Vining and Ebreo, 1990; Oskamp et al, 1991; Gamba and Oskamp; 1994; Meneses and Palacio, 2005). Some studies have also found a positive relationship between income level and recycling involvement (Vining and Ebreo, 1990; Oskamp et al, 1991; Gamba and Oskamp, 1994) but a study by Scott (1999) found no relationship. 1.3 Research Objectives and Hypotheses The main objective of this study is to analyze the influence of various factors such as socioeconomic, demographic and behavioral characteristics on drop-off site visits. The behavioral aspects examined are environmental affiliation, recycling activities, and perception and attitudes towards recycling and the environment. This study also analyzes the effect of drop-off site distance from home on site visits. We propose and test the following hypotheses: H1 : Distance to recycling sites from home reduces the number of visits. H2: Number of different types of recyclables brought to a site increases the number of site visits. H3: Time taken to sort recyclables increases the number of site visits. H4: Access to curbside recycling reduces the number of site visits. H5: Demographic factors such as age, gender, marital status, education and employment status influences the number of site visits. H6: Affiliation with an environmental organization increases the number of site visits. 1.4 Methods Since we seek to analyze the effects of recycler characteristics on the number of drop-off site visits, we conducted a survey of drop-off site visitors. This section describes the survey design and data collection process. This section also reports the descriptive statistics of the variables of interest. We also conducted factor analysis to reduce the number of our attitudinal variables into a few interpretable factors that were later operationalized as explanatory variables in our statistical model of drop-off site visits. 1.4.1 Questionnaire Design and Data Collection The data for this study was collected through in-person interviews conducted at eight drop-off recycling sites around the Lansing area in Michigan. The survey (see Appendix 1) included questions on the frequency of visits to drop-off sites in the past three months and one year. Respondents’ home address was elicited to allow calculation of respondents’ travel distances to the recycling site. The survey also contained questions soliciting demographic information of the respondent such as gender, education, employment status income and marital status. There were questions on the respondents’ recycling activities such whether they have a curbside recycling service at their residence, the types of recyclables they brought on-site, and the time they take to sort their recyclables. A question on whether the respondents are affiliated with any environmental organization was also included. The survey also included a set of questions assessing the respondent’s experience and attitude towards recycling. ln answering these questions, respondents were read statements and asked to indicate the extent to which they agree or disagree with the statements on a five- point Likert-scale ranging from strongly agree to strongly disagree. The questionnaire was pre-tested and further improved before conducting the actual survey. The questionnaire pretest was conducted by interviewing several recyclers at one of the drop-off sites. The pretest resulted in some wording refinements and the changes in the arrangement of questions in the instrument. The finalized survey was conducted for four weeks, from the last week of October 2006 to the last week of November 2006. Interviews were conducted at each site four times on a three hour interval each time throughout the four week period. The survey dates chosen for all the sites were randomly selected to avoid any potential bias. During the survey, recyclers visiting the sites were approached for interviews. At the end of the survey, we approached 527 recyclers and managed to complete 356 interviews for a 68% response rate. 1.4.2 Variables Description Table 1.1 lists and defines the demographic and other related variables that were utilized in our analysis and its definitions. Most of the variables do not require further elaboration except for a few. The variables THREEMTHS and ONE YEAR are the number of visits to the drop-off site where the respondent was interviewed in the past three months and one year. The variable DISTANCE represents the roundtrip distance from the respondent’s home to the recycling site where the respondent was interviewed. The roundtrip distance was computed using MapQuest (www.mapquest.com). The variable CURBSIDE is a 10 dummy variable indicating if the respondents have access to curbside recycling pickup at their home. Table 1.1. Definition of variables Variable Definition THREEMTHS Total frequency of site visits in the last 3 months ONE YEAR Total frequency of site visits in the last 1 year DISTANCE Total round-trip distance from home to site NUMREC Number of different types of recyclables brought on site SORT/ME Time taken (in minutes) to sort recyclables at home CURBSIDE Access to curbside recycling (yes=1, no=0) Educated with a bachelor's degree or higher (yes=1, CDEGREE no=0) INCOME Annual household income ($1 ,000’s) HSIZE Household size AGE Age MALE Male (yes=1, no=0) MARRIED Married (yes=1, no=0) FULLEMP Employed full-time (yes=1, no=0) Affiliated with an environmental organization (yes=1, ENVAFF no=0) The summary statistics of the variables (Table 1.2) indicate that the average visits of respondents to a drop-off site in the past three months and one year are approximately 4 and 15 times respectively. The average roundtrip distance traveled by the respondents to a drop-off site is around 19 miles. The respondents recycle on average 6 different materials each time they visit a drop- off recycling site, and they spend approximately 16 minutes sorting out the recyclables that they bring. Twenty five percent of the respondents reported that 11 they have curbside recycling service at their residence. The majority of the respondents (72%) had at least four years of college education. Sixty four percent of our respondents are fully employed and the mean household income is $77,935. Our sample was comprised of 56% male respondents indicating a balanced recycling participation between genders. Seventy percent of the respondents were married, and the average household size was 2.5 people per household. Only 26% of the respondents indicate that they are affiliated with one or more environmental organizations. Table 1.2. Summary statistics of variables Variable Obs. Mean SD THREEMTHS 348 4.330 3.455 ONE YEAR 348 1 4.652 1 3.804 DISTANCE 333 19.712 10.287 NUMREC 348 6.322 3.474 SORT/ME 344 1 6.1 66 27.337 CURBSIDE 345 0.252 0.435 CDEGREE 348 0.71 8 0.450 INCOME 348 77,935 52.791 HSIZE 346 2.520 1 .265 AGE 345 48.542 15.181 MALE 347 0.556 0.498 MARRIED 348 0.704 0.457 FULLEMP 348 0.641 0.480 ENVAFF 346 0.263 0.441 Table 1.3 describes the statements that are used in our survey to elicit the respondents experience, knowledge and attitude towards recycling along with the 12 respective distribution of Likert scale responses and descriptive statistics. The scale is defined as (1) strongly agree, (2) agree, (3) neither agree nor disagree, (4) disagree and (5) strongly disagree. Based on the mean score we can see that drop-off recyclers disagree that recycling is a difficult task (M=4.174, SD=0.825). They also disagree to both the statements of not having enough sorting time (M=4.285, SD=0.711) and storage space (M=3.797, SD=1.038) indicating that time and storage space do not deter their recycling activities. The recyclers also disagree that recyclables stored may attract pests (M=4.026, SD=0.825). Most of the recyclers agree that they are familiar with the recycling facilities (M=1.947, SD=0.847) and the materials accepted for recycling in their area’s facility (M=1.724, SD=0.595). The recyclers also agree that their family expects them to recycle (M=2.312, M=1.012). However, the recyclers are quite indifferent on the statements on whether their neighbors (M=3.303, SD=0.848) and friends (M=2.912, SD=1.012) expect them to recycle. Nevertheless, most of the recyclers feel good about themselves when they recycle (M=1.559, SD=0.579). The mean scores also show that the recyclers strongly feel that recycling is generally beneficial to the environment. The recyclers strongly agree that recycling is major way to reduce pollution (M=1.617, SD=0.653), to reduce landfill use (M=1.549, SD=0.591), to conserve natural resources (1.563, SD=0.628) and to improve environmental quality (M=1.575, SD=0.598). Additionally, these general perceptions on the benefits of recycling are strengthened by what the recyclers believe on the contributions of their activities. The recyclers strongly 13 believe that their recycling activities will actually contribute to reducing pollution (M=1.635, SD=0.680), reducing landfill use (M=1.553, SD=0.585), conserving natural resources (M=1.571, SD=0.636) and improving environmental quality (M=1.576, SD=0.622). 14 .2358... 29.9.52. «5.. N am $3 osmmm swam Seam .x. 5 0.9.8. o. o... .898 88... >5. oszm .w mam m... o 330 9...... 8m... .90.... $03 9&8 as...“ «8N 8.9.8. c. 2. 8.9...“ 9%.".me a.“ mzomz .35 >8 5 8.5.8. 9:962 mmmd emu; o\o fim 90m; o\om.mm Swan 9.. 5 05.9.8. .o. 3588 Em: <3. $9.65.... 6... 9...; 5:89 Em _ . . o . o . o . o . o . .33 >5 5 3.5.8. 3. o E. . S o S m S m \o. B s... 8 65.9.8. 9.. 5.; 5.58. Em . 25...... . . o . o . o . o . .9me mmm o omo 4 Sn mm \0. mm \08 m \00 o. 69...... 9% _ .m... 838.98. 2... Sun. . . o . o . o . o . o . 65.9.8. 5. mafia... 9.. 80 F 3n m 3 mm 2. mm Sm m \on 3 \oo N 99m 6. 8QO canoes 92 .0: on _ MOSHE. . . . . dc. 989 .o m «.55... m . E... m8... «8...... as... 8 .3 o Sm m 0% 0 tom 0. me... 385. La; .9. on... “.3: . . o . o . o . o . o . .xmm. .3055 mum o wt .V So no \0v 3 be m 3 m \o. m m m. 9.989 29.33.. .9: .6”. 0.995 $585 . 0203.0 .02 02m< o m :32 29.25 353.0 02 u< 02a< 29.25 EoEESm >625 03.2.; 3:955 IdIo>= romeo 8.3.9 28m- :9... .o 3.3.2» 9.3.88.3 as. 523.5... 58.5.8 .3 some 15 mmod mmmd mwmd owmd mmmd mmmd med mmmd mhmd Nmo. .. mum; Km... mmm. F mam... mum... mmm. F mvmé Eb... mmm. P N 5N o\om.o o\om.o o\om.o o\oo.o o\om.o o\om.o o\om.o o\om.o o\om.m o\om.o o\om. F o\om.o o\om. — o\om.o o\oN. P o\om.o o\oN.F o\ow.N— o\om.o o\ow. F o\oN. F o\om.m o\oN.P o\o 7N o\om.o o\o ...v o\ov.v o\om.®w o\om.om £0va o\oo.om o\on.mv o\om. Fm o\om.nv o\om.mv o\on.mv o\o fine o\oN.ov o\ow.ov 0&de o\om.nv $53.. o\om.mv o\oo.m.v o\ov.wv o\oh.mv o\om.wv o\oN. rm .>.._m=a .m...oE..o..>..m 96......5 96.. ....<. $52.2... m....o>oo. >5 6... 92.8 . 68.286. .939. 623.50 29. __.2. 33.2.0... m....o>oo. >E .m... 99.2. . .m_._......m. .0 mm: .2232. moans. 92. .._>.. «2.3.8.... m..._o>oo. >... .m... 92.8 . 22.2.8 8:8. 29. .__; 8:323 9:989 >8 5... 92.3 _ >556 .m...mE..o._>..m 90.9... _..2. 95056.". 6.8.386. .932 9598 2 >9... 8.9.. m m. 9:962”. .m....ccm. .0 mm: .3232. 8:8. 9 >9... .99... m m. m....o>owm 22.2.8 moans. 2 >9... BEE m m. m..__o>oom .o.o>om. _ cos... =mm>E Scam boom .3: 6.3.0.6.... 29.3.39. m.o>om. o. 6:. 9898 >..Em. >.2 O>2mm mum... <2m QZSQmIm quQmmm Ox. 2m mum... <2 QZSQmm .OQQMI QOOO > .5... (m 6.58. a. 2...... 16 1.4.3 Factor Analysis We use factor analysis with principal component analysis to group the Likert-scale variables into a small number of interpretable factors. Factor analysis will group the variables that are measuring the same construct in the same factor. This method is commonly used in social science research to reduce data into a smaller set of factors that can be used to linearly reconstruct the original variables (STATA 2003). We use the Kaiser eigenvalue criterion and the scree test to decide on how many factors to retain before proceeding with further analysis. According the eigenvalue criterion, factors with eigenvalues greater than one are retained and factors with eigenvalues less than one are considered insignificant and therefore excluded. Table 1.4 reports the initial factor extraction with the eigenvalues and percentage of variances for each successive factor. Using the eigenvalue criterion method, four factors are retained for further analysis. Table 1.4. Total variance eXplained Initial Eigenvalues Factor Total % of Variance Cumulative % 1 7.6666 42.59% 42.59% 2 2.0989 1 1.66% 54.25% 3 1 .6771 9.32% 63.57% 4 1 .1 681 6.49% 70.06% The scree test on the other hand, is a graphical method of determining the number of appropriate factors to retain. A scree test involves plotting the eigenvalue magnitudes on the vertical axis against the ordinal eigenvalue 17 numbers on the horizontal axis and noting the point at which the plot becomes horizontal. The number of factors corresponding to the horizontal point indicates the appropriate number to retain. In Figure 1.1, the point where the line becomes horizontal starts at factor 4. Thus, the scree test indicates that we should also retain four factors, similar to the result of the eigenvalue criterion method. Figure 1.1. Scree plot of eigenvalues and factors @— (o- (I) a: 2 9v- C d) .9 LU N— O I I I I I 0 5 1O 15 20 Number We also conducted the Kaiser-Mayer Olkin’s measure of sampling adequacy (KMO) test and Bartlett’s test of sphericity to assess the suitability of the survey data for factor analysis (Hair et al., 1998). The results are reported in Table 1.5. The KMO is a statistical test that indicates the proportion of variance in the variables which is common variance, while the Bartlett's test is a statistical test for the presence of correlations among the variables. The KMO index ranges 18 from 0 to 1, reaching 1 when each variable is perfectly predicted without error by the other variables. A small value (<0.05) of the Bartlett's test significance level indicates that the data do not produce an identity matrix and thus, are suitable for factor analysis. The results of the KMO and Bartlett's test show that the data meet the fundamental requirements for factor analysis. Table 1.5. KMO and Bartlett's test Kaiser-Meyer-Olkin Measure of sampling 0.841 adequacy Bartlett's test of sphericity Approx. Chi2 5075 Of 1 53 Significance <0.001 We use the Varimax rotation method (Kaiser, 1958) to rotate the four retained factors in our solution. Varimax is an orthogonal rotation method that maximizes the variance of the squared loadings for each factor. Varimax rotation will also ensure that the factors produced will be independent or unrelated to each other. The rotated factor matrix with its factor loadings is presented in Table 6. We consider variables with loadings greater than 0.4 as “highly loaded’ and are salient to the interpretation of a factor. Using this criterion, the variables are grouped together in the appropriate factor categories (refer to highlighted cells in Table 1.6). Each factor is described based on these variables and assigned descriptive names. We also compute the Cronbach’s coefficient alpha for each factor to test the reliability of scales of the item variables. There is no standard cut-off point for the alpha coefficient, but it is generally agreed upon that the lower limit for Cronbach alpha is 0.7, although it may decrease to 0.6 (Hair et al., 19 1998) or even 0.5 (Nunnally, 1978) in exploratory research. We then use the factor loadings to compute new variables called factor scores. These scores are composite measures indicating the degree to which an individual scores highly on a particular factor based on their responses to the variables included in that factor (Hair et al. 1998). Table 1.6. Rotated factor matrix Variable Factor1 Factor2 Factors Factor4 DIFFIC 0109 39%;»: -0.138 0115 TIME 0205 f' an- 0002 -0.036 we 1 SPACE 0145 “may -0.118 -0.015 . W ;" PEST 0231 59.31.91! 0.104 -0.286 FACILI 0.168 -0.003 0.027 :&8§,: l MATER/ 0.224 -0.198 "0.129 PW: NEIGHB 0.042 0005 ”1:323..." 0.086 FRIEND 0.126 -0.092 3‘ M1989, 0.020 FAMILY 0.112 -0.122 23.183, 0.042 GOOD g‘afiaa: -0.351 -0.041 0.190 REDPOL 49%. -0.050 0.067 0.105 REDLAND 301999,: 0106 0.015 0.101 NATRES : 927'- 0.112 0.007 0.089 ENVQ ”6.97: -0.113 0.048 0.116 .m _ BREDPOL “5.5751041 -0.065 0.103 0.057 .' ‘N BREDLAND go“. ,0. -0.130 0.074 0.108 BNATRES 1.801.924", -0.075 0.016 0.079 m “’7“- BENVQ Q8921", -0.144 0.106 0.064 The variables that load highly into Factor 1 are GOOD, REDPOL, REDLAND, NA TRES, ENVO, BREDLAND, BNA TRES and BENVQ. This factor 20 is labeled as “Attitude” and can be best described as attitude and beliefs of the environmental benefits of recycling activities. A low score for this factor indicates that the respondents have positive attitudes and firm beliefs that their recycling activities will lead to environmental benefits such as reduced pollution and landfill use, conserving natural resources and improving environmental quality. The Cronbachs’s alpha for these items is 0.96. The item variables with high loadings on Factor 2 are DIFFIC, TIME, SPACE and PEST. We labeled this factor as ‘Convenience’ as it relates to recycling being a convenient activity to undertake. A high score for this factor signifies that the respondents regard recycling as something that is convenient to them as they have no issues of it being difficult, time consuming, space consuming and inhibitive to pests. The Cronbach’s alpha for these variable items is 0.6964. We labeled factor 3 as ‘Social Pressure’ and the variables that load highly on factor 3 concern the social pressure on the recycler. The variables in this factor are NEIGHB, FRIEND and FAMILY. A low score for this factor indicates that the respondents feel that neighbors, friends and family expectations are important elements in encouraging them to recycle. The Cronbach’s alpha for factor 3 is 0.7015. The variables that load highly on Factor 4 are FACILI and MA TERI. We labeled this factor as ‘Familiar’ as it relates to the familiarity of recycling facilities. A low score for this factor demonstrates that respondents are highly familiar with the recycling facilities and the materials accepted in the recycling facilities in their area. The Cronbachs’s alpha for factor 4 is 0.579. 21 Table 1.7. Factors, Item variables and Cronbach's alpha Factor Item Variables Cronbach's 01 GOOD, REDPOL, REDLAND, (1) Attitude NATRES, ENVO, BREDLAND, 0.960 BNA TRES, BEN V0 (2) Convenience DIFFIC, TIME, SPACE, PEST 0.696 (3) Social Pressure NEIGHB, FRIEND, FAMILY 0.702 (4) Familiarity FACILI, MA TERI 0.579 Table 1.7 outlines the factors, their respective variables as extracted by the factor analysis, and their Cronbach’s alpha coefficient. The factors condense the experience, knowledge and attitude towards recycling of our survey respondents into four new interpretable variables; namely Attitude, Convenience, Social Pressure and Familiarity. We use the factor score of these four new variables to ascertain their relationship to the usage of drop-off sites. 1.5 Analysis and Results A key element of this paper is to analyze the variables that influence visits to drop-off recycling sites. This section develops a visitation model to relate the effects of demographics, recycling activities, environmental affiliation and the variables derived from the factor analysis to the number of trips taken to a drop- off site. The visitation model is developed using the Poisson regression method. Poisson regression is utilized because the data for our dependent variable, the trips an individual takes to a recycling site y,-, are classified as a count variable where y,- can only take discrete values ( y,- = 1,2,3,...). More specifically, we will use the endogenous stratified and truncated Poisson regression since we use do not observe zero trips for any of the sample members as our sample is 22 obtained via the on-site sampling method. Following Haab and McConnell (2002) the Poisson probability with on-site endogeneity and truncation is expressed as follows: -A- yr -1 e ’A. . . _ I Pr(}’I|YI > 0)" YI _1! where A, is both the mean and the variance of the distribution. Since it is necessary for A, > 0, it is commonly specified as an exponential function: A; = exp( XIB) (2) where x,- is a vector of explanatory variables. Equation 1 can be simplified by re- writing it as: 9"” Af'" Prlyilyi > 0)= y ., (3) i . where y,-'= y,- — 1. Using equation 3, we obtain the log-likelihood of a Poisson funcfion Inlral X. y)= ELLeXi" + XIBy. —In(y,-'!>l <4) and thus, the on-site endogenous and truncated Possion can be estimated by simply running a standard Poisson regression of y,-- 1 on all Xi’s. Table 1.8 presents the results of the Poisson regression models predicting the number of trips taken to a recycling drop-off site in the last one year‘. There are two models in this analysis; Model 1 is the basic model that uses distance, number of recyclables, sorting time, access to curbside recycling and 1 . . . . . We also ran the regressrons usrng the trips 1n the past three months as the dependent variable and found similar results but with less explanatory power. 23 demographic variables as dependent variables. Model 2 is the extended model that also includes all four Likert-scale variables derived from the previous factor analysis along with all of the basic variables in Model 1. The results show that both models 1 and 2 are statistically significant with the likelihood ratio statistics of 445.64 and 546.17 respectively. The coefficients in Model 1 are all statistically significant at 5% and 1% level except for CURBSIDE, MALE and ENVAFF. In this model, access to curbside recycling services, gender and environmental affiliation play no statistically significant role in increasing the expected number of site visits. The coefficients on NUMREC, INCOME, HSIZE, AGE and MARRIED in Model 1 are positive and the coefficients on DISTANCE, SORT/ME, CDEGREE and FULLEMP are found to be negative. The significance level of the coefficients in Model 2 after adding the four attitudinal variables did not change much except for CURBSIDE and MARRIED. The sign of the coefficient on access to curbside recycling remains negative but the variable is now statistically significant at the 5% level. The variable MARRIED is no longer significant. The coefficients on NUMREC, INCOME, HSIZE and AGE in the extended model remain positive. The coefficients on DISTANCE, SORT/ME, CDEGREE and FULLEMP remain negative and statistically significant. Three of the four attitudinal variables; CONVENIENCE, FAMILIAR and SOCIAL are significant at the 1% level. FAMILIAR and SOCIAL have negative signs and CONVENIENCE has a positive sign. 24 Table 1.8. Poisson regression (Dependent variable = ONE YEAR Le. number of visits in the past year) £92911 M2 Variable Coeff. Std. Error Coeff. Std. Error DISTANCE -0.010 0.001 ** -0.01 1 0.002" NUMREC 0.050 0.005" 0.050 0.005" SORT/ME -0.005 0.001 ** -0.004 0.001" CURBSIDE -0.061 0.040 -0.082 0.041 ** CDEGREE -0.1 29 0.035“ -0.105 0.036“ INCOME 0.001 0.0003” 0.001 0.0003“ HSIZE 0.067 0.014" 0.079 0.014“ AGE 0.007 0.001“ 0.005 0.001** MALE 0.026 0.032 0.045 0.033 MARRIED 0.1 18 0.041“ 0.058 0.043 FULLEMP -0.349 0.034" -0.378 0.035" EN VAFF 0.059 0.036 -0.018 0.039 CONVENIENC E 0.039 0.016“ FAMILIAR -0.148 0.017" SOCIAL -0.080 0.017“ ATTITUDE 0.019 0.017 CONSTANT 2.128 0.082“ 2.238 0.083“ Observations 329 322 Log-likelihood -231 5.49 -21 72.55 - 2 In (Ln/LU) 445.64 528.47 Pseudo R2 0.09 0.11 "a < 0.01 , *a < 0.05 25 The coefficients on DISTANCE in both models imply that the expected number of visits reduces by 1% as roundtrip distance from home to site increases by a mile. This result strengthens the findings by Saphores et al (2006) that improved proximity to recycling sites would encourage recycling behavior. The coefficients on NUMREC in both models indicate that the number of site visits is expected to increase when a recycler recycles a wider variety of recyclables. The time taken to sort the recyclables at home was found to reduce the expected number of site visits. The SORTTIME coefficient in Model 2 indicates that a 10 minute increase in sorting time reduces the expected number of site visits by 4%. It is also expected that people with curbside recycling service at their residents too have less frequent visits to drop-off sites as many of their recyclables are collected at the curb. The coefficient for CURBSIDE in Model 2 suggests that the availability of curbside recycling reduces the expected number of site visits by 8.2%. The results show that people with a bachelor degree or higher are expected to have fewer site visits when compared to less educated people. The CDEGREE coefficient in Model 2 indicates that having a bachelor degree reduces the expected number of site visits by approximately 10%. One possible explanation is educated people tend to have busier jobs and were likely to allocate less time on recycling activities. The negative coefficients on FULLEMP also indicate that people who are employed full time are likely to spend less time on recycling activities when compared to people who are employed part time or unemployed. The results demonstrate that an increase in annual household 26 income is expected to increase the expected number of site visit in both models. This result also confirms the findings by Vining and Ebreo (1990), Oskamp et al (1991), and Gamba and Oskamp (1994) that income level increases recycling involvement. The positive relationship between household size and the expected number of visits is very much anticipated as larger households tend to consume more goods. The positive relationships between age and number of site visits found in both models are also consistent with previous findings that older people have a higher tendency to recycle (Vining and Ebreo, 1990; Meneses and Palacio, 2005; Saphores et al, 2006). The positive coefficient of the variable CONVENIENCE in Model 2 indicates that the number of expected site visits increases when recycling is regarded as a convenient activity. This result confirms the previous findings that convenience is an important factor that encourages recycling behavior (Vining and Ebero, 1990; Hornik et al, 1995; Domina and Koch, 2002; Gonzalez-Torre et al, 2003; Saphores et al, 2006). The coefficient for FAMILIAR suggests that people who are more familiar with locations and materials accepted at the drop- off center in his or her vicinity are expected to make more visits to the centers than the less familiar people. The coefficient for SOCIAL implies that peers and family pressure has a positive effect on drop-off site visits. This result conforms to the findings of previous studies that indicate social pressure is an important factor motivating recycling behavior (Oskamp et al, 1991; Do Valle et al, 2005). 27 1.6 Conclusions The success of a drop-off recycling program, similar to other recycling programs, is largely dependent on the participation of the public. This study helps us understand the profile of people who utilize drop-off recycling sites as well as some of the underlying factors that influence their frequency of use. The results from this study demonstrate that drop-off site utilization is influenced by demographics factors such as age, education, income and household size. These results corroborate the findings of previous studies that include demographic effects in analyzing recycling behavior. This study also found that location plays a crucial role in influencing the usage pattern of drop-off sites. Recyclers are likely to use a drop-off site more frequently if the travel distance from home to site is shorter. Thus, the decision to establish a drop-off recycling program should factor in location to encourage its use. There are also perception and attitudinal factors that affect drop-off site visits. Recyclers tend to use the drop-off sites more when they feel that recycling is a convenient activity. This study also found that familiarity of the recycling sites and recyclables accepted is associated with increased usage of drop-off facilities. This suggests that promotion of drop-off recycling facilities and the materials accepted could be used to encourage drop-off recycling activities. 28 References Arcury, T. A., Scollay, S. J., and Johnson, T. P. 1987. “Sex Differences in Environmental Concern and Knowledge: The Case of Acid Rain.” Sex Roles 16:463-472. Domina, T., and K. Koch. 2002. “Convenience and Frequency of Recycling: Implications for Including Textiles in Curbside Recycling Programs.”Environment and Behavior 34(2) :21 6-238. Do Valle, P.O., R. Elizabeth, J. Menezes and E. Rebelo. 2004. “Behavioral Determinants of Household Recycling Participation: The Portuguese Case.” Environment and Behavior 36(4) :505-540. Ebreo, A., and J. Vining. 2001. “How Similar are Recycling and Waste Reduction? Future Orientation and Reasons for Reducing Waste as Predictors of Self-Reported Behavior.” Environment and Behavior 33(3) :424-448. Gamba, R.J., and S. Oskamp. 1994. “Factors Influencing Community Residents’ Participation in Commingled Curbside Recycling Programs.” Environment and Behavior 26(5) :587-612. Gonzalez-Torre, P.L., B. Adenso-Diaz, and A. Ruiz-Torres. 2003. “Some Comparative Factors Regarding Recycling Collection Systems in Regions of the USA and Europe.” Journal of Environmental Management 69(2):1 29—1 38. Haab, TC, and K.E. McConnell. 2002. Valuing Environmental and Natural Resources: The Econometrics of Non-Market Valuation. Northampton, MA: Edward Elgar Publishing. Hair, J.F. Jr., R.E. Anderson, R.L. Tatham, and WC. Black. 1998. Multivariate Data Analysis 5th Edition. Upper Saddle River, NJ: Prentice Hall. Hornik, J., J. Cherian, M. Madansky, and C. Narayana. 1995. “Determinants of Recycling Behavior: A Synthesis of Research Results.” Journal of Socio- Economics 24(1):105—127. ‘ Kaiser, HF. 1958. “The Varimax Criterion for Analytic Rotation in Factor Analysis.” Psychometrika 23(3):187-200. Margai, F. 1997. “Analyzing Changes in Waste Reduction Behavior in a Low- lncome Urban Community Following a Public Outreach Program.” Environment and Behavior 29:769-792. Meneses, 6.0., and AB. Palacio. 2005. “Recycling Behavior: A Multidimensional Approach.” Environment and Behavior 37(6):837—860. 29 Nunnally, JG. 1978. Psychometric Methods, 3rd. Ed. New York: McGraw-Hill Book Co. Oskamp, S., M.J. Harrington, T.C. Edwards, D.L. Sherwood, S.M. Okuda, and DC. Swanson. 1991. “Factors Influencing Household Recycling Behavior.” Environment and Behavior 23(4) :494-51 9. Saphores, J.M., H. Nixon, O.A. Ogunseitan, and AA. Shapiro. 2006. “Household Willingness to Recycle Electronic Waste: An Application to California.” Environment and Behavior 38(2) :1 83-208. Scott, D. 1999. “Equal Opportunity, Unequal Results: Determinants of Household Recycling Intensity.” Environment and Behavior 31 :267-290. Snyder, K.C., O.V. Kristel, B. Dhmammarungruang, and S. Sang. 2004. Report to the Ohio Environmental Protection Agency: Drop-off Recycling - Understanding Participation and Determining an Empirical/y Based Access Credit Model. Columbus, OH: Ohio Environmental Protection Agency, July. Speirs, D., and P. Tucker. 2001. “A Profile of Recyclers Making Special Trips to Recycle.” Journal of Environmental Management 62(2):201 -220. STATA Corporation. 2003. STA TA Base Reference Manual A-F Release 8. Vol. 1. College Station, TX: STATA Press. USEPA. 2007. Municipal Solid Waste. [Cited November 11 2007]. Available from m7/www.epagov/msw/recycle.htm. Vining, J., and A. Ebreo. 1990. ‘What Makes a Recycler? A Comparison of Recyclers and Nonrecyclers.” Environment and Behavior 22(1 ):55-73. Werner, C., and Makela, E. 1998. “Motivations and Behaviors that Support Recycling.” Journal of Environmental Psychology 18:373-386. 30 ESSAY 2. ESTIMATING THE DEMAND FOR DROP-OFF RECYCLING SITES: A RANDOM UTILITY TRAVEL COST APPROACH 2.1 Introduction The four primary methods to collect recyclables in the United States are curbside programs, drop-off centers, buy-back centers, and deposit or refund programs (USEPA, 2007). Drop-off recycling is a recycling program where designated sites are established to collect a range of recyclables and usually the recyclers themselves are required to deposit the sorted recyclables in specially marked containers. Drop-off recycling is also one of the most widely adopted recycling programs by local governments in this country. As of 1998, there were 12,000 recyclable drop-off sites and 9,000 curbside programs established in the United States (USEPA, 2000). Drop-off recycling centers are less costly to operate compared to curbside programs, and they are also faster to implement than take-back programs or other similar programs involving manufacturers (Saphores et al, 2006). Drop-off program operators are able to save on labor and transportation costs because these costs are transferred to the recyclers. Drop-off operations typically do not impose any charges to recyclers utilizing drop-off sites. Drop-off recycling is also considered to be the most financially viable recycling option in areas with low population density such as in rural areas or the countryside (Tiller, Jakus and Park, 1997). There is a need to understand community participation and effectiveness of drop-off recycling sites to assist policymakers in making better recycling and 31 waste management decisions. In 2004, Ohio EPA conducted a study to learn more about waste diversion amounts, recycling participation rates and usage patterns at drop-off recycling sites in Ohio. This information is required by the authority to more accurately grant population access credits to solid waste management districts in Ohio for the drop-off recycling programs in their jurisdictionz. Despite its wide implementation, relatively little literature addresses the demand for drop-off recycling. Curbside recycling as a waste management policy tool is the more popular research area in the field of recycling and waste management. Fullerton and Kinnaman (1996), Hong and Adams (1999), Van Houtven and Morris (1999), Kinnaman and Fullerton (2000), Jenkins et al (2003) analyze the effect of curbside recycling, together with other policy tools such as variable garbage pricing, on the amount of waste generation and recycling. Other curbside recycling research investigates the values consumers place on curbside recycling by computing their willingness to pay for the service (Lake et al, 1996; Aadland and Caplan, 1999; Aadland and Caplan, 2003; Blaine et al, 2005). One of the few exceptions of recycling research that is related to drop-off recycling is the stated preference study of a drop-off program conducted by Tiller, Jakus and Park in 1997. Their study analyzed the economic feasibility of establishing a drop-off recycling program in a rural and a suburban area of Tennessee by utilizing the contingent valuation method to calculate household 2 Access credit is the number of people with access to recycling facilities in a solid waste management district in Ohio. The access credit is divided over the population of a district to determine if the district complies with the minimum required percentage of the population with access to recycling facilities. 32 willingness to pay (WTP) for the program. The estimated WTP for the three different types of households controlled for respondents’ income, education level, age and attitudes toward the importance of recycling. They found that suburban recyclers, which consist of households with curbside recycling services, are willing to pay the most for drop-off recycling, with a mean WTP point estimate of $11.74 per month. Rural recyclers have a mean WTP of $7.07, and rural non- recyclers have the lowest mean household WTP of $4.05. Chang and Wei (1999) examine the strategic planning aspects of drop-off recycling centers in Kaohsiung, Taiwan. Their study analyzed the trade-off between the number and size of drop-off centers, walking distances to the drop- off centers, population covered in a service area and the driving distance of collection vehicles. The analysis was conducted by formulating a multi-objective mixed-integer linear programming model which balanced the following objectives: to maximize the population served by recycling centers, to minimize walking distance and to minimize total routing distance of collection vehicles subject to several physical constraints such as limit on drop-off centers in an area, service efficiency, capacity limitations, scheduling limitation and service area. There are also a few other studies that have indirectly looked into drop-off recycling. Folz (1991) in a study examining the success of recycling programs reported that solid waste management experience of recycling coordinators is a very important factor in maximizing participation in drop-off programs. It was argued that experienced coordinators make better decisions in choosing the best strategic locations for drop-off centers. Folz also reported that advertising and 33 promotion of recycling results in higher waste diversion to drop-off programs. In a descriptive study, Speirs and Tucker (2001) examined the profile of recyclers utilizing drop-off recycling sites in Glasgow and around Ayrshire in south-west Scotland. They reported on the recyclers’ travel distances, the weights and types of recyclables and demographic characteristics. They also found that people whose trips were solely for the purpose of recycling tend to be a shorter distance from the sites compared to people who combine their recycling trips with other activities. In a more recent publication, Saphores et al (2006) studied willingness to recycle electronic waste at drop-off centers by conducting a mail survey of households in California. The results from their multivariate analysis indicated that familiarity and convenience were very important factors in influencing willingness to recycle. People who are familiar or accustomed with glass, metal, paper or plastic recycling are more willing to recycle electronic waste. The study also found that people who lived more than 5 miles away from the nearest drop- ofi recycling center were less likely to recycle. In comparison to the broader literature on recycling, and specifically the attention paid to curbside recycling, there are relatively few studies that emphasize drop-off recycling. We address this gap by studying the demand for drop-off recycling sites in an urban area with several substitute sites using the random utility model (RUM). The main objective of this study is to examine the impact of location and different drop-off recycling site characteristics on drop-off recycling visits. We hypothesize that the travel costs incurred by recyclers to drop-off sites reduce site visits. We also hypothesize that site specific 34 characteristics such as operating hours, number of recyclables accepted, acceptance of commingled recyclables and acceptance of yardwaste affect recycling visits. This study also uses the RUM model to predict the changes in drop-off recycling patterns given the changes in site characteristics. This study improves our understanding on attributes of drop-off sites that may influence visitation demand. The study findings can be used by local governments and recycling and waste management companies to design and establish recycling drop-off centers that will increase site visitation and collection of recyclables. Our study utilizes the revealed preference approach which is different from the study conducted by Tiller et al (1997) that uses the stated preference approach. Unlike stated preference studies that rely on a respondent’s survey answers on monetary amounts, choices, ratings or other preference indications to establish a measure of value on non-market goods or services (Brown, 2003), a revealed preference study collects information on respondents’ actual behavior, such as number of visits and cost of traveling to particular sites, to establish the demand and value of these non-market goods or services. The RUM model which originates from the transportation field is widely used within the field of environmental economics to analyze the demand for recreational sites. We believe that our study is a novel application of the RUM travel-cost method to estimating the demand for local public services such as drop-off recycling sites. 2.2 Theoretical Framework The RUM model is widely used to analyze discrete choices in the face of many substitutes. In our case, the RUM is appropriate because it is able to 35 consider a household’s selection of a drop-off recycling site, chosen from a set of many alternative drop-off sites, on an occasion in which they have chosen to visit a drop-off site. While the decision to utilize a drop-off site has many elements of a cost minimizing decision, we posit that households also have preferences (and hence derive utility) from the convenience attributes and other attributes of recycling sites. Thus, when selecting a site the household is assumed to take into account the trip cost to arrive to the site as well as the site characteristics. The trip cost would mainly be the driving cost and time cost to travel to and from the site. Site characteristics are the features of each drop-off site such as operating hours and types of recyclables accepted. Hence, each recycling site will give households different utility levels and, after factoring in the travel costs, households are assumed to visit sites that yield them the highest utility. Specifically, to model the drop—off site selection process, we assume that households derive utility from the quality or characteristics of a particular drop-off site. Each household has a choice set of S number of sites that they could visit denoted by j = 1,2,...S. Let xi,- represent trips household i takes to site jwith a vector of M site characteristics [qj1,qj2,...qu]. In evaluating the utility household i derives from a trip to site j, we assume that qk = 0 for all k a: j. The utility function for household iis defined as follows U{Zisxij(Qj1:Qj21---q/'M)} (1) where z,- is a composite consumer good and the utility function is assumed to be increasing and strictly quasi-concave in all its arguments. Households maximize utility subject to a budget constraint (2) and time constraint (3). 36 n injcij +2,- s y,- (2) j=1 n qutg+h,+I,sT (3) j=1 where oil-is the return-trip driving cost to site j, z,- is purchased at a normalized price equal to 1, y,- is the household income which is further defined in (4), til-is the time taken for each round-trip to site j, h,-is the hours spent working, I,is the time spent for leisure and Tis the total time available. Income is given by . _ 0 -h- y I —y,- +WI I (4) where y? is fixed income and w,- is the wage rate. Solving the time constraint using equation (4) for h and substituting it into the budget constraint yields n inj(c,-j +w,-t,-j)+z,- Syi (5) j=1 By solving the utility maximization problem, we derive the Marshallian demand function for x,-,- and substituting the demand function in the utility function results in the indirect utility function that is expressed as follows: Vij = (YI -(Cij +Witij)in) (6) The indirect utility function can be represented in a linear form where 85 are the parameters to be estimated and e,-,- is the random error term Vij =BU’i "(Cij +Witij))+BqCIj +91] (7) The cost of visiting site] that consists of round-trip driving and time costs, c,-,- + w,-t,-,- is essentially the travel cost which will be simplified as tcij. We also 37 note that y,- will drop from the equation as it does not vary across sites. Thus, the indirect utility function can rewritten as follows vi]- = [31¢th + quj + 9,7 (8) On a given a choice occasion, a household decides to recycle at the site that yields the highest utility. A recycling site k is chosen by household iif: Btctcik + quk + eik 2 [3,0th + quj + eij for all j e S (9) We could express the choice to visit a recycling site in a probabilistic framework, where the probability of a household visiting site k is: Pr([3,ctc,-k + quk + eik 2 6,0th + [3qu + 9,; for all j e S) (10) The choice probability of household i visiting site kcould be expressed using the conditional logit form, where: Pr (ik) = :XMfitctcik + quk) (11) ZeXp(BtctCij + quj) [=1 The model estimators can be derived by maximizing the following log-likelihood function constructed from equation 11 f N N export to. +3 0 ) IOQLn(y.B)= ZIogP ’° ”‘ q " S "=1 Zexpwtctcij + qu j) \i=1 / (12) The equation in (12) provides the likelihood function for a random sample of drop-off recyclers. However, if our data is collected using a choice-based or an on-site sampling method, then we would need to correct for the potential 38 endogenous stratification in the data. With on-site sample data, if we know the population proportions of recyclers visiting the S sites, we can to use the Weighted Exogenous Sampling Maximum Likelihood (WESML) method to derive consistent estimates of the model parameters (Manski and Lerman, 1977). The WESML estimator is obtained by weighting the population proportions by the sample proportion and incorporating these weights in the likelihood function. From equation 12, the weighted exogeneous likelihood function is presented as follows 1 \ expwmtcik + quk) N Q. Ioanly.B) = Z—fi’jlogP n=1 8 ’ Z expwmtcij + quj) M=1 I where O,- is the proportion of the population selecting site j and H ,- is the analogous proportion for our choice based sample. To use the demand model to forecast changes, we use the WESML estimates for the parameters to predict the probability for the households across our sample. 2.3 Survey Methods 2.3.1 Questionnaire Design The questionnaire used in this study consists of questions pertaining to the respondent’s recycling activities. We include questions on the frequency of visits to drop-off sites in the past three months and one year to calculate site visits. Questions on the respondent’s income and home address are also included in the questionnaire to compute the travel costs. These questions occurred at the end of the interview because we felt respondents would be more 39 comfortable sharing this more personal information after going through a set of more general questions. On the home address question, we asked the nearest street intersection to the respondent’s home whenever they were reluctant to reveal their address. The questionnaire also consists of questions eliciting general demographic information of the respondent such as gender, education, employment status and marital status. The questionnaire was pre-tested and further improved before conducting the actual survey. The questionnaire pretest was conducted by interviewing several randomly selected recyclers at one of the drop-off sites. The pretest resulted in some wording refinements and changes to the arrangement of questions in the instrument. 2.3.2 Data Collection We define our population to consist of recyclers utilizing eight drop-off recycling sites in and around Lansing, Michigan. We used on-site interviews to collect our survey data. This survey method is chosen over a random population survey because we expect the percentage of population that recycles at drop-off sites to be low and we would require a large sample size to obtain sufficient number of drop-off recyclers in such a sample. Furthermore, with low proportions for the target population, the costs involved in conducting a detailed survey with such large sample size are also high. Manski and Lerman (1977) suggest that choice based survey can achieve the economies of scale not available with a population survey in circumstances where the respondents are physically clustered according to the alternative they select. Similarly, Haab and McConnell 40 (2002) noted that on-site sample surveys are a more cost effective approach for data collection for multiple site models when the proportion of the population participating in the activity is quite low. However, there is a problem associated with on-site sample surveys - the sampling scheme is often independent of the population proportions visiting the survey sites. The problem arises as model parameter estimates depend on the sampling proportions. If the sampling proportions differ from the population proportions, the model will suffer from inconsistent parameter estimates. In other words, sampling proportions that differ from population proportion will result in the parameter estimates capturing both the sampling plan and recycler’s behaviors, rather than behavior alone. Nevertheless, the on-site sampling endogeneity problem can be addressed if the population proportions are known. Specifically, consistent parameters can be estimated using the WESML method if the population shares are known (Manski and Lerman, 1977). There were two separate processes involved in obtaining our survey data: the first part was to measure recycling effort at each drop-off site and the second part was to conduct on-site interviews of recyclers. The measures of recycling effort at each site are used to estimate the population proportions of the eight drop-off sites to construct the WESML weights to correct for possible choice- based sampling bias. To measure recycling effort, we counted the vehicles visiting each of the drop-off sites. The counting exercise was conducted simultaneously at each site, except for Site 8 due to the site’s operating hours restriction (see Appendix: Table A1). We selected Saturday, October 21, 2006, 41 9am to 2pm for the car counting exercise for the seven drop-off sites to capture the busiest recycling period. The car counting for Site 8 was conducted on Monday, October 23, 9am to 2pm, as we expected the traffic at this time to be the busiest, thus equivalent to weekend traffic at other sites. Table 2.1 provides the distribution of cars according to recycling sites. Table 2.1. Distribution of cars by sites from the effort survey Site Name Cars Percentage Site 1 146 25.44% Site 2 202 35.19% Site 3 113 19.69% Site 4 29 5.05% Site 5 50 8.71% Site 6 26 4.53% Site 7 0.52% Site 8 0.87% For the second part, on-site survey interviews were conducted for four weeks starting from the last week of October 2006 to the last week of November 2006. Each site was randomly visited 4 times, on a three-hour interval each time throughout the survey period. For each visitation time, we randomly selected the sites. A site that was not open on a particular visitation time was excluded to ensure a zero probability of selection for that time. With the exception of Site 2 and Site 8, we excluded sites that had been selected on the same visitation time in the previous weeks. For example, if Site 1 had been selected for Week 1, Sunday 3-6pm, the site was excluded in the drawing for that time period for Week 2 at the same time. We made an exception for Site 2 and Site 8 because 42 both sites have limited operating hours and if we were to impose a duplicate time restriction on these sites, the sites would have been visited less than 4 times. A site was excluded from the draw once it had been selected 4 times (see Table A2 in Appendix for the detailed survey schedule). At the end of the survey duration, we approached 527 recyclers. Out of the total approaches made, 356 recyclers agreed to participate in the survey giving us a response rate of approximately 68%. The distribution of respondents according to recycling sites is given in Table 2.2. We obtained the highest number of respondents from Site 1 which represents 26.4% of the total sample. The second highest number of respondents was obtained from the Site 3 followed by the Site 2, representing 22.75% and 20.51 % of the total sample. The percentage distribution of respondents for Site 5, Site 6, Site 4 and Site 8 are 12.36%, 8.43%, 5.34% and 3.09% respectively. The lowest number of respondents is from Site 7 constituting only 1.12% of the total sample. Table 2.2. Distribution of respondents by sites from the on-site survey Site name Respondents Percentage Site 1 94 26.4% Site 2 73 20.51% Site 3 81 22.75% Site 4 19 5.34% Site 5 44 12.36% Site 6 30 8.43% Site 7 4 1.12% Site 8 11 3.09% 43 Table 2.3 presents the demographic profile of the recyclers in our sample. We had more males (55.62%) than female respondents (44.38%) in the survey. With regards to age, the majority of the respondents are 40 years or older, and the highest age group was between 50 to 59 years old (28.16%). The lowest age group was between 30 to 39 years old which accounts to only 10.92% of the respondents. As for household composition, about 71% of the respondents are either married or living with a partner. Approximately 80% of the respondents lived in a household comprised of 2 or more people. The respondents were mostly college-educated as approximately 70% of the sample have a bachelor’s degree or higher. The majority of the respondents were also employed full time (64.27%). In terms of household income, roughly 60% of the respondents reported a household income of $45,000 or more. 44 Table 2.3. Demographic characteristics of drop-off recyclers Variable Frequency Percentage Gender Male 154 44.38 Female 193 55.62 Age 18 to 30 years old 56 16.09 30 to 39 years old 38 10.92 40 to 49 years old 86 24.71 50 to 59 years old 98 28.16 60 years old or more 70 20.11 Education Some high school 3 0.86 High school or GED 54 15.56 Vocational or trade school 5 1.44 Two year degree 36 10.37 Four year degree 133 38.33 Graduate school 116 33.43 Employment Status Employed full time 223 64.27 Employed part time 29 8.36 Unemployed 3 0.86 Retired 63 1 8.16 At home parent 8 2.31 Student 21 6.05 Income Less than $25,000 43 12.36 $25,000 to $44,999 75 21.55 $45,000 to $74,999 78 22.41 $75,000 to $99,999 77 22.13 $100,000 or more 75 21.55 Marital Status Single 69 19.94 Married/Living with partner 245 70.81 Divorced/Widowed/Separated 32 9.25 Household Size 1 65 18.79 2 149 43.06 3 53 15.32 4 57 16.47 5 13 3.76 more than 5 9 2.60 2.4 Model Estimation and Results We estimated our basic model using the WESML estimation method and the model was specified as follows: Vij = B1TRA VELCOSTil- + 52 HOURS ,-j + [33 NUMREC ij 14 + fi4COMM/NG ij + H5 YARDWASTE I] + eij ( ) where v,-jis the indirect utility individual igets from visiting site j, and each j has the same independent, Type 1 extreme value distribution, F9191] ) = exp(— exp(— e,-,- j) (15) which, under maximization, yields the conditional logit model for the choice probabilities as in (11). We also examine an extension of the basic model by incorporating interactions between demographic variables and the site attributes. The extended model is specified as follows: 17,-,- = [31TRA VELcosn-j + I32 Houris,j + 83 NUMREC”- + I34 COMMINGij + 85 YA RD WASTE,-j + Be EHounsij +I37ENUMPEC,j +88 ECOMMINGyj +89EYARDWASTE,-j (16) + B10 HHouns-j + p, 1HNUMREC,-j + B12HCOMMINGj + B13HYARDWASTE,-j + e,- Table 2.4 provides a list of variables and their definition. For each site and for each respondent, the travel cost variable was calculated by adding up the roundtrip driving cost and the time cost to travel from the recycler’s home a drop- off site. The distance from home to the recycling site was obtained with Mapquest using the shortest distance option. The driving cost was assessed at 35 cents per mile (AAA, 2006). Driving cost consists of per mile vehicle operating cost plus depreciation per mile. Time cost was the opportunity cost incurred by 46 the recycler during the drop-off recycling activity calculated using the recycler’s income. The trip time is computed assuming that a recycler travels at 35 miles per hour on average. The recreation literature has generally accepted 1/3 of an individual’s wage as a lower bound and an individual’s full wage as an upper bound for the hourly value of time spent driving (Parsons, 2003). We use the recycler’s full wage in our time cost calculation and the wage is computed by dividing annual income by 2080 hours of work time (52 weeks at 40 hours per week). Table 2.4. Definition of variables Variable Definition TRA VELCOST Roundtrip travel and time cost from home to site j HOURS Total operating hours per week NUMREC Number of recyclables accepted COMMING Number of types of commingled recyclables accepted YARDWASTE 1 if site accepts yard waste (0 otherwise) EHOURS Interaction between full employment dummy (1 = fully employed, 0=otherwise) and HOURS ENUMREC Interaction between full employment dummy and NUMREC ECOMMING zigzag/1221; between full employment dummy and EY A RDW A STE lcgegaocfinstfgween full employment dummy and HHOURS Interaction between household size (= the number of household members) and HOURS HNUMREC Interaction between household size and NUMREC HCOMMING Interaction between household size and COMING HYARDWASTE Interaction between household size and YARDWASTE 47 The site characteristics data used in our model were obtained from the information given at the site and through our own observations. The operating hours for the recycling sites varied from as low as only 15 hours per week to 24 hours a day. We expect operating hours per week to increase site visitation because it increases flexibility and convenience for recyclers. The number of recyclables accepted also varies from site to site. There are sites accepting as few as only 5 types of recyclables to a site that accepts 17 different types of recyclables. We expect sites accepting a wider range of recyclables to receive higher visitation rates when compared to sites accepting a limited range of recyclables. There are sites that accept commingled plastic or papers and recyclers visiting these sites are not required to separate the different types of recyclable plastics or papers. This attribute is expected to increase site visits as it makes recycling easier and more convenient. We also included a dummy variable to represent sites that accept yardwaste. The sites that accept yardwaste are Site 1, Site 2, Site 3 and Site 8 and they charge fees ranging from $5 to $10 per cubic yard of yardwaste recycled. However, as Site 2 only. accepts yardwaste from its township residents, the yardwaste dummy variable for Site 2 will only take a value of 1 if the respondent lives in the township where Site 2 is located. EHOURS, ENUMREC, ECOMMING, EYARDWASTE are the interactions between the dummy variable for full employment and the attribute variables; operating hours, number of recyclables accepted, acceptance of commingled recyclables and acceptance of yardwaste. These variables are included to 48 ascertain the degree of influence of the site attributes on drop-off site visits for persons with a full-time employment. We also interact household size (the number of people in the household) with the same four attribute variables to form HHOURS, HNUMREC, HCOMMING, HYARDWASTE to determine if households with different sizes place a different weight on the site attributes. The estimation results are presented in Table 2.5. Model 1 is the basic model, and Model 2 is the extended model that includes interaction variables between demographics and site attributes. All the parameters in Models 1 are statistically significant, and all the parameters in Model 2 are statistically significant except for NUMREC. The travel cost variable is negative and highly significant as expected in both the models. In other words, the parameter indicates that by holding all other variables constant, it is expected that the probability of visiting a recycling site will decrease as the cost of traveling to the site increases. The parameter estimates for the travel cost variable in both models are also very similar. The estimates for YARDWASTE indicate that yardwaste acceptance has a large impact on increasing the probability of site visits. The parameter estimates for HOURS indicate that increasing site operating hours per week will increase site visits. However, this interpretation is only applicable to non 24 hour sites. 49 Table 2.5. Random utility model results (Dependent variable = Number of visits in the past year) M2d_6|_1 £29st Estimate Std. Error Estimate Std. Error TRA VELCOST -0.144 0.005" -0.145 0.005" HOURS 0.010 0.001** 0.002 0.001* NUMREC 0.196 0.011** 0.001 0.023 COMING 0.501 0.035" 0.766 0.087" YARDWASTE 2.260 0.082“ 1.041 0.171 ** EHOURS 0.010 0.001" ENUMREC 0.039 0.021 * ECOMMING 0.208 0.076" E YARDWASTE 0.767 0.152“ HHOURS 0.001 0.0005" HNUMREC 0.078 0.01 1 ** HCOMMING -0.1 52 0.028" HYARDWASTE 0.383 0.075“ Observations 343 Adj-R2 0.472 0.489 Log-likelihood -6607 -6306 "a < 0.01, “a < 0.10 The results in Model 1 also imply that increasing the number of recyclables accepted at a drop-off site will increase the probability of site visits. The parameter estimates for COMMING in Model 1 indicate that allowing for an additional type of commingled recyclables will increase the probability recyclers visit that drop-off site. NUMREC is no longer significant in Model 2 although its interaction with the fully employed and household size variables are both significant. The positive parameter estimates for EHOURS, ENUMREC, ECOMMING and E YARDWASTE in Model 2 suggests that site attributes such as operating hours, the number of recyclables, acceptance of commingled and acceptance of yardwaste have more impact in increasing site visits for recyclers 50 that are working full-time. This result is intuitive as one would expect a fully employed person to be more occupied and might place a higher value on convenience-related site attributes when compared to recyclers who are not employed. The positive estimates for HHOURS, HNUMREC and HYARDWASTE indicate that site operating hours, the number of recyclables and acceptance of yardwaste have more impact in increasing site visits for larger households. Perhaps not intuitive, the negative parameter estimate for HCOMMING suggests that the acceptance of commingled recyclables would reduce recycling visits of larger households. In interpreting the results for the site attribute variables, it is important to note that with only eight sites in the model, there are limitations on the independent variation in the site attributes. Indeed, with the exception of travel costs, we found some evidence of multicolinearity between the site attributes. Further, with only eight sites in our model, we can include a maximum of eight site attribute variables before our model is over identified. Thus, we found that due to multicollinearity between the site attributes, the parameter estimates are sensitive to the combination of attributes included in our model. Nevertheless, the travel cost parameter estimate was stable and consistent regardless of the site attributes included in our model. To further aid the interpretation of the results, we construct Table 2.6 that provides the marginal implicit prices for changes in site characteristics (using the estimates from Model 1). The marginal implicit price of a variable is calculated as a ratio of the variable’s parameter estimate to the travel cost parameter estimate 51 and represents the marginal rate of substitution between a site attribute and travel costs. Marginal implicit prices can be used to compare the relative Importance of different site characteristics to travel costs. The marginal implicit price of $15.69 for YARDWASTE indicates that yardwaste acceptance has the highest impact on recycling visits compared to a one unit change in the other site attributes. HOURS has the lowest marginal implicit price per unit. The results suggest that a 20 hour increase in operating hours per week for a non 24 hour site has almost the same impact as accepting an additional type of recyclable. Similarly a change from the lowest site operating hours, 15, to 24 hour operation would have about the same effect as accepting an additional eight kinds of recyclables (a change in NUMREC of eight). The results also suggest that a change from zero to two types of commingled recyclables accepted has approximately the same impact as accepting 5 additional types of recyclables or having an additional 100 operating hours. While YARDWASTE has the largest effect for a one unit change in the variable, since it is a dummy variable the one unit change is akin to a change from its lowest value to its highest value. Considering the effect of a change from the lowest to highest value of NUMREC, a change of 17, when multiplied by its marginal implicit price, reveals that it has the largest overall effect of any of the site attributes over their respective ranges. 52 Table 2.6. Marginal Implicit prices for changes In site characteristics Variable Marginal Implicit Prices HOURS $0.07 NUMREC $1 .36 COMING $3.48 YARDWASTE $15.69 2.5 Scenario Analysis and Policy Implications In this section, we use the basic model to impute the probability of site visitation to the respective eight sites. The probability of visitation for each site is calculated by substituting the parameter estimates derived from our weighted model into the household weighted site choice probability function (equation 11) and summing it up across all respondents. Table 3 presents the probability of visitations in descending order, to all the drop-off sites. Table 2.7. Probability of site visitation Slte name Predicted probability of trip Site 3 0.271 Site 2 0.209 Site 1 0.207 Site 5 0.101 Site 6 0.100 Site7 0.055 Site 8 0.044 Site 4 0.013 The model predicts that most recycling trips are taken to the Site 3 drop- off recycling site. The recycling site with the lowest probability of site visitation is 53 Site 4. The difference between the site with the highest and lowest probability of visitation is also very large. Given that a recycler makes a trip to a drop-off site, the probability of the recycler choosing to visit Site 3 is approximately 20 times larger than the probability of the recycler choosing Site 4. Using the model we predict changes in visitation rates when a recycling site is closed. Table 2.8 displays the probability of site visits to remaining sites when one particular site is closed. The closed site in the table is indicated by a zero probability of visitation. A site closure will result in recyclers resorting to alternative sites, and they will substitute the closed site with its next best alternative. The next best alternative site might be a site that is closest in distance to the recycler or a site with similar attributes to the site that has been closed. Since we cannot know which site is each person’s next best alternative, we report the predicted probability of site choice after the change. The best substitute for a closed site, on average, will experience the highest increase in probability of site visitation. The simulation results indicate that if we close Site 1, the site that experiences the highest increase in visitation probability is Site 5. This is probably because of the proximity between Site 5 and Site 1 although Site 5 lacks some of the attributes Site 1 has, such as the acceptance of yardwaste. Another site that receives an equally high increase in probability of visits when Site 1 is closed is Site 2. An explanation for this is the similarity in features between the two sites such as the acceptance of a wide variety of recyclables although the distance between the two sites is quite far. 54 55 - mvod mvod 95.0 306 50.0 95.0 mead m mam Bod - mmod mmod Bod «mod mood Rod n gm word 9 to .. w? to mono ammo 3 to mm; o mam veto mote 3 to - mono om; ammo 5:. m cum m5.o Bod m Ed mad - $5.0 «Nod 390 v 25 mmmd mmwd vomd nmmd mmmd - 3N6 9nd m 25.. mowd Sud mwmd 3N6 :md ENG - Bud N 95 o 5.0 ommd vwmd mvmd 08.0 Emd Rad - w 2m m 8.5 m 35 m mum. m 3.2m. v mum m. mum N mum. _ 3.6 $20 985 $20 $90 890 $20 820 986 flglom NEH. g m 8:8 4 Eon. 39.3 Mam 9% 33.0 m_ cgm n 22.3 at. B 32532.. @3052; 2:2. 25 658.0 23 2 can 5:27.; 98 8 3:3an .3 833 We also use the model to predict the changes in visits when we change the attributes of a recycling site. We create a hypothetical combination of good attributes for a recycling site: 24 hours of operating time, accepts 10 different types of recyclables, accepts two types of commingled recyclables and accepts yardwaste. One possible scenario is to improve the features of Site 6 to the hypothetical site attributes. Site 6 is a smaller and a less comprehensive site as compared to the popular sites such Site 1, Site 2 and Site 3. However, Site 6 is strategically located in the middle of all the eight sites, and it has the lowest mean distances for all respondents (see Figure 2.1). This simulation will indicate what happens to visitation patterns when we improve the attributes of Site 6. Alternatively, another possible scenario is to change the features of Site 8 to the hypothetical site attributes described above. Site 8 is one of the least popular sites mainly because of its location but also due to its attributes such as a low level of recyclables accepted and limited operating hours. Table 2.9 outlines the probability of site visitation to all eight sites under these two scenarios: changing the attributes of site 6, and changing the attributes of site 8. 56 Figure 2.1. Locations and mean distance (miles) for drop-off recycling sites _.—...a Walenown .1! cm Rd _ “:l +--t Family/1.11 ‘i. ’82. ~( on wwu . .. Table 2.9. Probability of site visitation after changes in site attributes Predicted probability of a Predlcted probability of a Site name trip If the attributes of trip if the attributes of Site 6 are improved Site 8 are improved Site 1 0.122 0.193 Site 2 0.178 0.205 Site 3 0.164 0.230 Site 4 0.007 0.012 Site 5 0.050 0.090 Site 6 0.421 0.090 Site 7 0.024 0.049 Site 8 0.034 0.131 57 By improving the attributes of Site 6, the probability of‘visits to the recycling site has substantially increased from 0.10 to 0.42. On the contrary, the probability of a visit to all other substitute sites has decreased with Site 1 having the greatest reduction from 0.27 to 0.12. This is anticipated as we have enhanced a substitute for Site 1 at a more convenient location. It is also interesting to note that Site 2 experiences a smaller decrease in probability of visits when compared to Site 3. This suggests that most of the recyclers using Site 2 are from the Site 2 township itself, and it would not be convenient, distance wise, to switch to Site 6. Furthermore, for a Site 2 township resident, the site offers similar attributes to the ‘new’ Site 6 such as the acceptance of yardwaste and wide variety of recycling materials. Subsequently, improving the site attributes of Site 8 also results in an increase in the probability of visit to the site. However, the increase is only from 0.03 to 0.13 which is not as large compared to our first scenario when we change the attributes of Site 6. Changing the attributes of Site 8 also did not result in large decreases in the probability of visits to other remaining sites. This implies that attributes alone are not enough to substantially increase visitation rates and a recycling site needs to be both conveniently located and equipped with comprehensive attributes to attract a large share of users. 2.6 Conclusions There are only a few studies on drop-off recycling despite the wide implementation of drop-off programs across this country. This study addresses this gap by proposing a novel method to assess the demand for drop-off 58 recycling sites. The use of a RUM model, which has been traditionally employed in transportation and recreation economics, is an appropriate and theoretically consistent way to analyze the demand for drop-off recycling sites in an urban setting with several substitute sites. The findings demonstrated that higher travel costs significantly reduce the probability of visiting a drop-off recycling site. This implies that the location of a site relative to where people live clearly affects site visitation. The findings also indicate that site-specific convenience-related attributes such as site operating hours, the number recyclables accepted, acceptance of commingled recyclables, and acceptance of yardwaste significantly affect which drop-off recycling sites get visited. However, some caution is warranted when interpreting the specific effects of individual site attributes (other than travel costs) due to our finding of potential multicollinearity between these attributes. Nevertheless, taken as a group, the site attributes were always highly significant in explaining site choices. Given the significance of all these factors, policy makers should consider the influence of site location and site attributes when planning and designing facilities in order to maximize the use of drop-off recycling sites. Drop-off sites should be located in areas that are relatively near and accessible to a majority of the population. The results from the scenario simulation demonstrate that a site that is conveniently located will attract a high usage if equipped with convenience-related attributes such as a accepting a wide variety of recyclables and yardwaste. Our scenario simulation results also demonstrate that improving 59 convenience-related attributes alone will not be sufficient to significantly increase site visits if the site is not strategically located. 60 References AAA. 2006. Your Driving Costs 2006. [Cited December 1 2006]. Available from http://www.aaaexchanqe.com/Main/ Aadland, D., and A.J. Caplan. 1999. “Household Valuation of Curbside Recycling.” Journal of Environmental Planning and Management 42(6) :781 -799. Aadland, D., and A.J. Caplan. 2003. “Willingness to Pay for Curbside Recycling With Detection and Mitigation of Hypothetical Bias." American Journal of Agricultural Economics 85(2) :492-502. Blaine, T.W., F.R. Lichtkoppler, K.R. Jones, and RH. Zondag. 2005. “An Assessment of Household Willingness to Pay for Curbside Recycling: A Comparison of Payment Card and Referendum Approaches.” Journal of Environmental Management 76(1 ):15-22. Brown, TC 2003. “Introduction to Stated Preference Methods.” In P.A. Champs et al, ed. The Economics of Non-Market Goods and Resources: A Primer on Nonmarket Valuation. Dordrecht: Kluwer Academic Publishers, pp. 99-110. Chang, N., and Y.L. Wei. 1999. “Strategic Planning of Recycling Drop-Off Stations and Collection Network by Multiobjective Programming.” Environmental Management 24(2) :247-263. Folz, DH and J.M. Hazlett. 1991. “Public Participation and Recycling Performance: Explaining Program Success.” Public Administration Review 51 (6):526-532. Fullerton, D., and TC Kinnaman. 1996. “Household Responses to Pricing Garbage by the Bag.” The American Economic Review 86(4) :971 -984. Haab, TC, and K.E. McConnell. 2002. Valuing Environmental and Natural Resources: The Econometrics of Non-Market Valuation. Northampton, MA: Edward Elgar Publishing. Hong, S., and RM. Adams. 1999. “Household Responses to Price Incentives for Recycling: Some Further Evidence.” Land Economics 75(4):505-514. Jenkins, R.R., S.A. Martinez, K. Palmer, and M.J. Poldolsky. 2003. “The Determinants of Household Recycling: A Material-Specific Analysis of Recycling Program Features and Unit Pricing.” Journal of Environmental Economics and Management 45(2) :294-31 8. Kinnaman, TC, and D. Fullerton. 2000. “Garbage and Recycling With Endogenous Local Policy.” Journal of Urban Economics 48(3):419-442. 61 Lake, I.R., |.J. Bateman, and J.F. Parfitt. 1996. “Assessing a Kerbside Recycling Scheme: A Quantitative and Willingness to Pay Case Study.” Journal of Environmental Management 46(3) :239-254. Manski, CF, and SR. Lerman. 1977. “The Estimation of Choice Probabilities from Choice Based Samples." Econometrica 45(8):1977-1988. Parsons, GR. 2003. “The Travel Cost Model.” In P.A. Champs et al, ed. The Economics of Non-Market Goods and Resources: A Primer on Nonmarket Valuation. Dordrecht: Kluwer Academic Publishers, pp. 269-329. Saphores, J.M., H. Nixon, O.A. Ogunseitan, and AA. Shapiro. 2006. “Household Willingness to Recycle Electronic Waste: An Application to California.” Environment and Behavior 38(2):183-208. Snyder, K.C., O.V. Kristel, B. Dhmammarungruang, and S. Sang. 2004. Report to the Ohio Environmental Protection Agency: Drop-off Recycling - Understanding Participation and Determining an Empirical/y Based Access Credit Model. Columbus, OH: Ohio Environmental Protection Agency, July. Speirs, D., and P. Tucker. 2001. “A Profile of Recyclers lVlaking Special Trips to Recycle.” Journal of Environmental Management 62(2):201 —220. Tiller, K.H., P.M. Jakus, and WM. Park. 1997. “Household Willingness to Pay for Drop-Off Recycling.” Journal of Agricultural and Resource Economics 22(2) :31 0- 320. USEPA. 2000. Municipal Solid Waste in the United States: 2000 Facts and Figures. [Cited November 11 2007]. Available from hm)://www.epa.qov/garbaqe/msw99.htm. USEPA. 2007. Municipal Solid Waste. [Cited November 11 2007]. Available from http://www.epa.qov/msw/recycle.htm. Van Houtven, G.L., and GE. Morris. 1999. “Household Behavior under Alternative Pay-As-You-Throw Systems for Solid Waste Disposal.” Land Economics 75(4) :51 5-537. 62 ESSAY 3. FACTORS INFLUENCING THE RATE OF RECYCLING: AN ANALYSIS OF MINNESOTA COUNTIES 3.1 Introduction Annual municipal solid waste (MSW) generation in the United States has increased from 88 million tons in 1960 to approximately 251 million tons in 2006 (EPA, 2006). The escalating waste generation combined with concerns over landfill costs and availability of space have prompted regulators and policymakers to reform MSW policies at all levels, from the community to the state level. Waste management practices such as source reduction, recycling, and composting have been instituted in order to reduce materials from entering the waste stream. A range of programs and policy instruments are available to policy makers for managing waste and recycling. Given the variety, analyses of the effectiveness of these practices are needed for improved policy decisions. Several studies have been conducted to understand the effects of various waste management policies on reducing waste and increasing the amount of recycling, and these studies have analyzed the impacts of policies on per capita waste generation and recycling demand. However, most policymakers evaluate the effectiveness of recycling and waste management programs by looking at improvements in the rate of recycling because the rate of recycling is able to capture movements in the amount of recycling and waste generation at the same time, as opposed to separate measurement of per capita quantities. In fact some 63 states like New Jersey have regulations mandating minimum recycling or diversion rates. One of the key features of this study is that we analyze the impacts of different policies on recycling rates. Another feature of this study is the use of time-series data on recycling rates. Most of the earlier studies on the effectiveness of recycling and waste management policies are based on cross sectional data analyses, and hence they are unable to incorporate dynamics of certain policy variables. Compared to these, the analysis in this paper uses county level data from Minnesota covering the period 1996 to 2004 and the panel nature of the data enables analysis of cumulative effects over time. The broad objective of this research is to examine the factors that affect recycling rates in Minnesota counties by utilizing a set of panel observations of recycling and waste management policies, along with income and demographic variables. Specifically, this research analyzes the effectiveness of various recycling and waste management policy variables on county rates of recycling. The policy variables that we examine include: variable pricing for waste disposal, expenditures on recycling education, provision of curbside recycling services and drop-off centers, and enactment of recycling ordinances. Unique to this study, we account for the cumulative effects of the expenditure variable on recycling rate. We also investigate whether different recycling programs such as curbside and drop-off recycling act as complements or substitutes in increasing recycling rates by considering the interactions between the variables. Lastly, this research 64 examines the effect of income and demographic characteristics such as age, education and population density on the recycling rate. 3.2 Literature Review Household recycling and waste management literature can be divided into two broad classes; articles that use more comprehensive models incorporating the behavior of governments, firms, and consumers and articles focusing on consumer reactions to various pricing schemes (Linderhoff et al, 2001). Articles of the first type are mostly theoretical while those of the second type are mostly empirical. Earlier theoretical studies in this area began analyzing waste generation without including recycling. Wertz (1976) for example, develops an economic model to explain household's decisions regarding waste production. Households are assumed to maximize utility, which is a function of goods consumed and waste generated, subject to a budget constraint incorporating the cost of waste disposal that increases with waste generated. The model analyzes, among other variables, the impact of waste disposal unit-pricing and income on the quantity of waste generated. The analysis suggests that the quantity of waste generated decreases when the waste disposal fee increases, and waste generation increases with income. ' The study by Wertz (1976) was extended by Jenkins (1993) by modeling both the residential and commercial sectors’ demand for solid waste services and most importantly by including recycling as a waste reduction option. The household utility maximization model suggests that the level of household income, the prices of goods consumed, payments from recycling deposit items 65 and waste disposal user fees will affect the demand for solid waste services. A firm profit maximization model is developed to derive the demand for commercial solid waste services. The model assumes that the cost of production increases with recycling activities but is mitigated by the increase in revenue from selling recyclables. The final analysis suggests that the demand for commercial solid waste services is a function that is decreasing in costs of production, increasing in the revenue from sales of recyclables, and the decreasing in the user fee for commercial solid waste services. Saltzman, Duggal and Williams (1993) developed a theoretical framework for analyzing household waste generation and recycling. They adopted the model by Wertz (1976) and introduced recycling explicitly in the utility function. Unlike the study by Jenkins (1993), the recycling function of their study is more elaborate because it includes factors such as household’s degree of ecological consciousness and the amount of time required to sort the recyclables from waste. Their study also derives comparative statics to examine the mixed impact of income on the household recycling effort. Hong (1999) adopted the household production function framework by Becker (1965) and Pollack and Wachter (1975) to develop a household solid waste generation and recycling function. He incorporated the time spent recycling in the utility function and derived the reaction functions for solid waste generation and recycling. He then developed a system of structural equations to investigate the interactions between household waste generation and recycling and found out that household recycling efforts are expected to increase as waste collection fee increases. 66 Kinnaman and Fullerton (2000) developed a theoretical model to derive the recycling and waste disposal demand function for a community, and then empirically estimated the impact of waste disposal fees and curbside programs on recycling and waste generation. Their paper established that the demands for waste disposal and recycling are functions of curbside recycling, price of waste disposal, mandatory recycling policies, deposit refund system for recyclables, ban on yardwaste disposal, income and demographic characteristics. The empirical studies on recycling and waste management can be classified into two categories; the studies that utilize household level data (Fullerton and Kinnaman, 1996; Van Houtven and Morris, 1999; Hong, 1999; Hong and Adams, 1999; Linderhof et al, 2001) and the studies that utilize macro level data such as a community or county level data (Podolsky and Spiegel, 1998; Kinnaman and Fullerton, 2000; Johnstone and Labone, 2004; Callan and Thomas, 2006). Most of the empirical work using household level data studies the impact of garbage pricing on waste generation. Fullerton and Kinnaman (1996) utilize individual household data in Virginia to estimate the effect of a garbage unit pricing program on the weight of the garbage, the number of containers, the weight per can and the amount of recycling. They found that in response to unit pricing households reduced the number of garbage bags but not necessarily the weight of the garbage. They also found that there was an increase in the weight of recycling, but there was also an evidence of illegal dumping. They concluded the incremental benefit of the unit pricing program was small and outweighed by the administrative cost. 67 Van Houtven and Morris (1999) examined the implication of unit-based pricing on household waste generation in a waste demonstration project in Marietta, Georgia. Instead of paying a fixed collection fee, half of the participating households paid a fee for a reusable trash can and the other half paid for each trash bag collected. The data collected indicate that both the programs significantly reduced waste generation even after taking into account the increase in recycling. The bag program was also found to result in a larger reduction in waste generation compared to the can program. Hong (1999) studied the impact of unit pricing and aggressive recycling programs on solid waste generation and recycling by employing data from a large household survey from Korea. The results indicated that an increase in waste collection fee paired with aggressive recycling programs was more effective in reducing the waste disposal than an increase in collection fee without any recycling program. They also found that a weight-based unit pricing system was more effective in reducing waste compared to volume-based unit pricing as households tend to compact wastes to reduce charges in the latter form of pricing. Hong and Adams (1999) studied the effect of disposal fee and household characteristics on recycling and waste generation by analyzing household level data from Portland, Oregon. They reported that an increase in the price of solid waste collection increased the demand for recycling, and they also found that an increase in the price of collection had a negative effect on the demand for waste disposal. Linderhof et al (2001) utilized a large household panel data from a Dutch municipality to analyze the effects of weight-based pricing on household 68 waste collection. They concluded that weight-based pricing has a strong negative effect on the amount of waste collected. Podolsky and Speigel (1998) developed a theoretical model to derive the demand for municipal solid waste disposal and examine the effect of unit-pricing and household recycling on municipal solid waste disposed by using community-level data for 149 New Jersey municipalities. Unique to their study, the model assumed that households derive utility from the consumption of goods and incur disutility from recycling waste. The model suggests that the optimal level of waste generated will be at level where the marginal disutility of recycling equals the unit price of disposal. Their empirical evidence indicates that unit price significantly reduces the amount of solid waste disposed. They also hypothesized that recycling affects waste disposal through two components. The first is the non-price component which is the direct effect of recycling reducing the amount of solid waste as the two activities are substitutes. The second component captures the indirect effect of unit-pricing on waste disposal through changes in recycling. However, they found that only the non-price component of recycling significantly reduces waste disposal. Most of the macro-level studies that we reviewed utilize cross-sectional recycling and waste management data. As mentioned above, Podolsky and Speigel (1998) in a study using community-level data for 149 New Jersey municipalities found that unit pricing significantly reduces the amount of solid waste disposal. Kinnaman and Fullerton (2000) derive a similar conclusion by analyzing cross-sectional data of more than 900 US communities. They also 69 found that curbside recycling programs encourage recycling. The main contribution of the Kinnaman and Fullerton paper, besides having employed a relatively large dataset was to recognize the endogeneity of local government policy decisions. Unlike most of the previous studies that assume that policy variables such waste disposal pricing and provision of recycling facilities are exogenous; they argued in that these variables tend to vary with community attributes and household characteristics. Since public policies are usually responsive to the conditions in the local community, Kinnaman and Fullerton treat disposal pricing and implementation of curbside recycling as endogenous variables and correct the endogeneity by using a two-stage least square (ZSLS) estimation method. Johnstone and Labonne (2004) is one of the few waste management studies that has used a panel dataset in its analysis. This study analyzes the determinants of solid waste generation using the municipal solid waste, demographic and economic data of 30 Organisation for Economic Co-operation and Development (OECD) countries from 1980 to 2000. Adopting the framework proposed by Kinnaman and Fullerton (1997), their study models the demand for municipal solid waste services as a function of waste pricing and demographic characteristics such as average data on household size, number of children in a household, the number of working adults in a household and proportion of urban population. Due to in availability of the price data, they used population density and country dummies as proxies for waste disposal prices. It was argued that waste disposal pricing was reflected in the cost of disposal which was closely 70 related to population density and the country itself. Their study found that household waste generation was relatively inelastic with respect to income and that urbanization increases waste generation. In a more recent study, Callan and Thomas (2006) examine the demand for disposal and recycling services by utilizing cross-sectional data of 351 municipalities in Massachusetts. Similar to Kinnaman and Fullerton (2000), their study allowed for the endogeneity of garbage unit-pricing and the provision of curbside services, but their study differed by also allowing the demand for waste disposal and recycling to be simultaneously determined. In addition to variables for demographics, garbage pricing and curbside recycling, this study also incorporates an expenditure variable which measures state funded grants for recycling equipment and education. The results demonstrate that unit-pricing indirectly effects garbage disposal through increased recycling and that availability of curbside services reduces disposal demand. The results also indicate that communities with grant allocation for recycling education or equipment recycle significantly more than communities without any allocation. It was also found that household size and age are significant determinants for the demand for disposal service and education is a significant determinant of recycling. 3.3 Theoretical Framework We adopt the utility maximization model proposed by Kinnaman and Fullerton (2000) but with some slight modifications to explain waste generation and recycling in order to derive the function for the rate of recycling. Similar to 71 their approach, we begin by explaining household choices and then apply this framework to the county’s policy choices. Assume an economy in a county comprised of N identical households with utility functions represented by Ui -"‘~ U(Xini) (1) where x,- represents composite goods and services consumed by household i and E,- is the environmental quality that the household i perceives to enjoy. The environmental quality function is specified as follows Er = hIQia’iidi) (2) where g,- is the amount of waste disposed and r,- is the amount of materials recycled by household i. The choice of the g,- and n in E,- is also dependent on a set of demographic characteristics, d,-. Substituting equation (2) into (1 ), we rewrite the utility function as Ur = U(Xiwh(gisri;di)) (3) Each household maximizes utility subject the following budget constraint Xi +ngi + Pr’i = mi (4) where the price of composite goods x,- is normalized to unity, pg is the price of waste disposal, p,is the price of recycling and m,- is the total household income. Solving the utility maximization problem yields the following demand functions for waste and recycling: gi =g(pg,pr,mi,di) (5) fl = rIPgspramiidi) (6) 72 The following equations outline the functions for the price of waste disposal and the price for recycling. Pg =Pg(P.L.misai) (7) pr = pr(F.S.mi.d,-) (8) The price of waste disposal for a household will factor in the disposal fee (P) imposed by the waste service provider as well as other costs associated with the time and effort required by the household to handle the waste. The time and effort costs are assumed to be functions of household income and demographic characteristics. The price of waste disposal is also affected by regulations that enforce recycling (L). Presence of regulations such as the ban of certain recyclables in the waste stream will directly increase disposal costs because of possible penalty. The price of recycling for a household is mainly the time and effort costs of separating, storing, transporting and depositing the recyclables. Similar to waste, time and effort costs can be functions of income and demographic characteristics. Time and effort costs can also be affected by policy variables such as the types of recycling programs (F) and recycling education related expenditures (8). Recycling programs such as curbside recycling services and drop-off recycling centers will reduce the time and effort costs faced by households to recycle. Recycling education expenditures will help increase recycling awareness and will educate households on the importance and benefits of recycling. Recycling education will also expose households to the efficient 73 methods of recycling and the availability of recycling facilities and this will indirectly help reduce the time and effort costs associated with recycling. Substituting equations (7) and (8) into (5) and (6), we rewrite the waSte and recycling demand functions as 9i = QIP: L, F13: misdi) (9) r, = r(P, L, F, s, m,-,d,-) (10) We are able to derive the rate of recycling by using the information on waste and recycling, and the theory implies that the recycling rate function will have the same explanatory variables as in the waste and recycling demand functions. Recycling rate is defined as a percentage ratio of the weight of recycled waste to the total solid waste collected for disposal and incineration. (Lund, 2001). Since we assume that households within each county have the same utility function, results aggregated at the county level will represent aggregate household level decisions. Thus, the recycling rate function of a county is written as follows Hi = f(F’isLi:I'-i:8iamisdi) (11) We expect variable rate pricing to have a positive effect on recycling rate because in a variable pricing system, households will have to pay higher disposal charges when they produce more waste. This pricing system is expected to provide an incentive for households to increase recycling in order to reduce the amount of waste disposal. We also expect the presence of regulations that enforce recycling to increase the recycling rate. Mandatory recycling regulation forces people to recycle to avoid penalties, and this will increase recycling and 74 reduce the amount of recyclables in the waste stream. County recycling programs are expected to increase recycling rate; Recycling programs such as curbside recycling pickup and drop-off centers facilitate recycling activities by making recycling more convenient for households. We expect programs such as curbside recycling and drop-off centers to have a higher impact on recycling rate when implemented together than when they are implemented separately. We expect recycling education expenditures to be positively associated with the recycling rate because they increase awareness and reduce costs. We expect income to have a negative association with recycling rate. People with higher income generally consume more and tend to generate higher amounts of waste which leads to a lower rate of recycling. Also, when income increases the opportunity cost of recycling goes up and this can lead to a reduction in recycling. The county demographic variables are represented by age, education and population density. We expect age to be positively related to recycling as people who are older, especially retirees, tend to have more time to spend on activities such as recycling. This view is supported by the findings of previous studies that middle aged and older people are more likely to recycle (Vining and Ebreo, 1990; Meneses and Palacio, 2005; Saphores et al, 2006). We expect education to be positively associated with the rate of recycling. People with higher education are expected to be more aware of environmental issues which would encourage them to recycle. We expect population density to be negatively associated with recycling rate. Residences in high population density areas are usually smaller than those in low population density areas. 75 Since lack of space is one of the major inconveniences that discourage people to recycle, it is probable that high population density will lead to reduced recycling activities and increased waste generation, which results in an overall decrease in the rate of recycling. 3.4 Data and Variable Descriptions The data used for this research are drawn from two different sources. The recycling and waste management data for this study are obtained from the Minnesota’s SCORE database. Minnesota is among the pioneers of waste management reforms in this country. It began a statewide recycling effort as early as 1989 after the adoption of legislation based on the recommendation by the Governor's Select Committee on Recycling and the Environment (SCORE). The legislation provides state funding on waste reduction, recycling program management and household hazardous waste management. The waste management and recycling program in Minnesota is deemed to be one of the most successful state level programs in the United States considering the local government investments and public participation. According to a report in BioCycIe magazine’s survey 2006, Minnesota has a recycling rate of 43% and is ranked second in the nation after Oregon with a recycling rate of 45%. The SCORE database compiles data from annual surveys of waste management and recycling in all Minnesota counties. The SCORE survey is administered by the Minnesota Pollution Control Agency (MPCA) and is completed by county solid waste officers. The survey collects information on MSW generated, materials collected for recycling, solid waste collection system, 76 recycling programs and management, waste and recycling revenue and expenditure, source reduction programs and other MSW policy initiatives. We augment the recycling and waste management data with income and demographics data for Minnesota counties from the US Census Bureau database covering the period from 1996 to 2004. The two data sources combine to create a balanced panel of complete variables for 774 observations representing 86 counties in Minnesota. The variables selected for our analysis are listed and defined in Table 1. Table 3.1. Definitions of variables Variable Definition Rate Residential recycling rate per annum (percentage) VarP 1 if county implements variable rate pricing structure (0 otherwise) Ordin 1 if county has an ordinance that requires residences to recycle (0 Curb Drop EduExp Inc Age Educ Den otherwise) Percentage of population with access to curbside recycling Number of drop—off recycling centers per 1,000 persons in the population Cumulative (3 years) recycling education expenditure per capita ($) Income per capita (35) Median age Percentage of population with 4 or more years of college education Population density per square mile Recycling rate is computed by dividing the amount of residential recycling by the amount of total waste generated in the countya. VarP is our waste disposal pricing variable which is a dummy variable representing whether the county had 3 Total waste generated is the summation of amount of recycling and MSW Iandfilled. 77 a variable pricing scheme or not. The survey did not collect information on the actual unit prices charged. In contrast to the studies that use community level data, we use county data which with aggregate information for several different townships or communities making it difficult to deduce the county waste price as the prices usually vary across communities. Furthermore, in most cases variable pricing meant two or three tier pricing for different container sizes, and this type of pricing is not a true unit pricing such as “by the bag” pricing of waste disposal4 because the marginal costs of an additional unit of waste in a tiered pricing scheme are zero within a tier. Given these considerations, it is reasonable to operationalize variable pricing as a dummy variable. We expect the presence of VarP to be endogenously determined in a county. Our recycling regulation variable is represented by Ordin which indicates whether a county enacts recycling ordinances to make recycling compulsory for the residents. We use Curb and Drop to represent recycling programs in a county. Curb measures the percentage of county’s population with access to curbside recycling, and Drop measures the number of drop-off recycling facilities available per 1,000 persons in the county. We also expect both Curb and Drop to be endogenous. The county’s recycling education expenditure variable is represented by EduExp which is the three-year cumulative recycling education expenditure for each county. Unlike previous studies, we use cumulative expenditures instead of current expenditure because we believe that expenditure on education has cumulative effects on recycling rate. The recycling awareness This Information on the types of vanable pncrng was obtained through telephone mtervrews wrth Minnesota EPA officers and waste management officers from several counties. 78 created from previous year’s education will still have an impact on the current year’s recycling activities. We are also able to avoid the potential endogeneity problem of an expenditure variable by using cumulative expenditure in our model. The income and demographic variables are represented by Inc, Age, Educ and Den. Table 2 provides the descriptive statistics of the variables in our model. Table 3.2. Summary statistics of variables Variable Mean Std. Deviation Min Max Rate 16.00 6.71 1.55 40.18 Inc 24704 451 8 1 6379 45565 Age 38.10 3.69 29.00 48.10 Educ 11.73 0.05 4.44 39.61 Den 121.77 417.63 3.21 3348.21 VarP 0.82 0.39 0.00 1.00 EducExp 0.37 0.52 0.00 4.77 Curb 53.42 27.27 0.00 1 00.00 Drop 0.67 0.68 0.00 4.19 Ordin 0.23 0.42 0.00 1 .00 n=774 The SCORE survey data indicates that the mean recycling rate for Minnesota is 16% and the rate varies from 1.55% to 40.18%5. Figure 3.1 and 3.2 illustrate the trend in waste generation and the amount of residential recycling for Minnesota from 1996 to 2004. The calculated mean is lower than mean published in the BioCycle magazine because we excluded recyclables such as used tires that were based on estimates 79 950 1 000 1 050 1 100 Mean Pounds Per Capita Municipal Solid Waste 900 1 60 1 80 200 220 Mean Pounds Per Capita Residential Recycling 140 l L l 1 Figure 3.1. Mean per capita municipal solid waste In Minnesota (1996-2004) I 1996 I l I 1 998 2000 Year I 2002 2004 Figure 3.2. Mean per capita residential recycling in Minnesota (1996-2004) 1 996 I 1 998 2000 Year 80 I 2002 I 2004 Figure 3.3 illustrates the residential recycling rates from 1996 to 2004. It is evident from Figure 3.3 that generally there is an increasing trend in mean recycling rate with slight fluctuations from 1996 to 2002. The rate of recycling then decreases from 17% in 2002 to approximately 15% in 2004. This occurrence can be explained by the steadily increasing trend in per capita waste generation (Figure 3.1) and the slight decrease in per capita recycling from 2002 to 2004 (Figure 3.2). Figure 3.3. Mean residential recycling rate in Minnesota (1996-2004) 17 I 16.5 L Residential Recycling Rate Per Annum (Percent) 15.5 16 1 I I l l f 1 996 1 998 2000 2002 2004 Year Eighty-nine percent of the counties in our data adopt variable pricing structure for waste disposal and 23% have enacted ordinances making recycling compulsory for residents. The mean expenditure for recycling education spent on a person is 37 cents and it varies from zero to $4.77. The mean percentage of 81 population access to curbside recycling services is 53.42% and the mean number of drop-off recycling centers per 1,000 persons in the population is 0.67. The US Census Bureau data indicates that the income and demographic characteristics of the counties in Minnesota vary considerably. The mean per capita income for Minnesota counties is $24,704, and the county average per- capita income ranges from $16,379 to $45,565. The median age per county varies from 29 years old to approximately 48 years old. The mean percentage of population with a bachelor’s degree or higher is approximately 11.74%, and it varies from 4.44% to 39.61%. The mean population density is 121.77 persons per square mile, and the density varies from as low as 3.41 persons per square mile to 3,348.21 persons per square mile. 3.5 Econometric Methods and Results We model the county recycling rate as a function of waste management policy, income and demographic variables. The linear econometric specification of the county recycling rate function is specified as follows Rate” = 80 + VarP,-,B1 + Ordinitfiz + Curbitfi3 + Drop,-,B4 + CurbDropitfi5 + EducExpitfis + Incit87 + Ageitfig (12) + EdUCitBQ + Dem-m1 0 + TimetB1 1 + a,- + “it Time, is the dummy variable for each year except first year, and a,- and u,-, are the components for the unobserved disturbance for county iat time t. Since we have a panel data, we are able to exploit the repeatability of the data by decomposing the unobserved disturbance to allow for a county-level effect. Thus, a, represents the unobserved county-level effect and u], is the idiosyncratic error 82 that changes over time for each county. The dummy variables for years will allow for exogenous statewide changes in recycling rate over time. We estimate the model using the random effects method as we assume that the unobserved effect is uncorrelated with each of our explanatory variables. We use the Hausman’s (1978) specification test to confirm that random effects is the appropriate specification for the county specific unobserved effects in our model as opposed to fixed effects. The test result indicates that unobserved effects are adequately modeled by random effects, and the model produces efficient and consistent estimators. Table 3 outlines the results of our regression models; pooled-OLS, random effects model, random effects model with instrumental variables (IV). The pooled OLS model is estimated for comparison basis as the random effects model assumes strict exogeneity between the explanatory variables and the disturbance term. If this assumption fails, pooled OLS will produce more consistent estimators. 83 9.6 v P. .mod v 6:. .56 v at: MK to mm 3 bogmzcmi wk 2 888 8.8- :88 $88 858 «88 2:98:80 mm> mmLEESQ 59» :88 588- 588 588- 588 588- :3 3 8 mm 8 :8 828 .3888 5 8 88m 55:8 838 838 8888 £838 88 88¢. .588 888- 588 588- 2.5888 888- 85 .888 858 5.8 2K8 89.8 488- 8388 :858 ~88 .858 888 288 588 8698 3.88 «88- 83 388 1.888 :3 86 «58 888 88 :88- 5:88 :88 9:0 :88.— 883. :88; 888 :888 BS. 520 :8 88 82. .888 mum. F 838 m 88 ES. 3am 85 28.0580 3am 85 53580 BE 85 586580 285.58%". 2 285.585". 90.8.8“. AEacca Log 88.. 95262 32:023.". u 038:? E35809 33:33 «83:55. 5 3a.. 95268.. .355 9: Co £558.33 66 03¢... 84 The pooled OLS results suggest that age and higher education have significant positive contributions to the rate of recycling. The results also suggest that an increase in income marginally reduces the rate of recycling where a 1,000 dollar increase in annual income per capita will reduce the rate of recycling by 0.2 percentage point. Our pooled OLS model did not address the endogeneity problem of any of the policy variables. The policy variables that are significant in this model are Curb, Drop and Ordin. For this model, an increase in access to curbside recycling improves the rate of recycling -- a 1 percentage point increase in access increases recycling rate by 0.04 percentage point. Adding one recycling drop-off center per 1,000 populations was also found to increase recycling rates by 1.28 percentage points. The results also indicate that having a county recycling ordinance will increase the rate of recycling by 4.16 percentage point. The random effects model without the endogenous policy variables has a lower explanatory power than the pooled model, and we find that income, age and education are not significant in this model. However, in this model VarP significantly affects the rate of recycling. The implementation of a variable pricing (VarP) structure on waste disposal charges is suggested to increase recycling rate by 1.62 percentage points. Ordin remains significant in this model but Curb and Drop are no longer significant. Nevertheless, the interaction variable CurbDrop is significant in this model. This indicates that curbside and drop-off recycling are able to increase the rate of recycling when implemented together. 85 Similar to the pooled OLS, our random effects model did not account for potential endogeneity of the policy variables. The third model corrects for endogenous policy variables. We identify endogenous variables in our model by conducting endogeneity tests. The test is conducted by first estimating a reduced form OLS regression for the potentially endogenous variables. We regress these variables against all exogenous variables in our main equation (equation 12), together with other exogenous variables that do not appear in the main equation but we believe are correlated with the endogenous variable. We then obtain the residuals from our reduced form regression and use it as explanatory variable in an OLS regression of our main equation. We conclude that a variable is endogenous if the reduced form residual is found to have a significant coefficient in the main regression (Wooldridge, 2006). The test found that VarP, Curb and Drop are endogenous. We correct for endogeneity of VarP by using the predicted value from its reduced form regression (see Appendix 3: Table A3.1). We instrument for Curb and Drop with its one-year lagged variables. The choice of one-year lagged variables for Curb and Drop as the instruments is appropriate because we believe that the previous years’ curbside recycling services and drop-off centers are exogenous to the current year’s decision on the amount of recycling and waste generated in a county. Correcting for endogeneity slightly improves the explanatory power of our random effects model, evident from the increase in R-squared from 0.167 to 0.173. VarP, Ordin and EducExp are significant in this model. We found variable 86 price (VarP) to have a bigger impact on increasing the rate of recycling after correcting for endogeneity. Implementation of variable pricing structure on waste disposal was found to increase the recycling rate by 4.19 percentage points in the lV-random effects model as opposed to only 1.62 percentage points in the random effects model without endogeneity corrections. The coefficient for EducExp suggests a dollar increase in per capita cumulative education expenditure will increase recycling rate by 0.82 percentage points. Income and age are found to be significant in this model with the expected signs. 3.6 Conclusions The rate of recycling is an important indicator that is widely used by policy makers to assess the level of recycling activities in a community, county or state. This study examines the effect of policy, income and demographic variables on the rate of recycling in Minnesota counties. After accounting for random effects and endogenous policy variables, our results demonstrate that variable pricing of waste disposal significantly increases the rate of recycling. This confirms the previous findings from cross-sectional studies that variable pricing is an effective policy tool for increasing the amount of recycling and reducing waste generation. Our findings also indicate that regulations can be an effective means of increasing recycling. We found that the enactment of recycling ordinances making residential recycling mandatory increases the rate of recycling. It is also interesting to note that curbside recycling services and drop-off centers are effective in increasing the rate of recycling when implemented together as a recycling program. Curbside and drop-off recycling were found to 87 be insignificant in improving the rate of recycling if they are implemented separately. Educating the public on recycling was also found to be an increase recycling rate. The findings showed that the cumulative expenditure on recycling education increased recycling rate, at the 10% level of significance. Spending one dollar per person per year will increase the rate of recycling by approximately 2 percent. 88 References Becker, G. 1965. “A Theory of the Allocation of Time.” Economic Journal 75:493- 517. BioCycle. 2006. “Why Minnesota Ranks Second in the Nation in Recycling.” July Vol. 47, No. 7, pp. 14. Callan, S.J., and J.M. Thomas. 2006. “Analyzing Demand for Disposal and Recycling Services: A Systems Approach.” Eastern Economic Journal 32(2):221 - 240. Fullerton, D., and TC. Kinnaman. 1996. “Household Responses to Pricing Garbage by the Bag.” The American Economic Review 86(4):971-984. Hong, S. 1999. “T he Effects of Unit Pricing System upon Household Solid Waste Management: The Korean Experience.” Journal of Environmental Management 57(1):1-10. Hong, S., and RM. Adams. 1999. “Household Responses to Price Incentives for Recycling: Some Further Evidence.” Land Economics 75(4):505-514. Jenkins, R. 1993. The Economics of Solid Waste Reduction: The Impact of User Fees. London: Edward Elgar Publishing. Johnstone, N. and J. Labonne. 2004. “Generation of Household Solid Waste in OECD Countries: An Empirical Analysis Using Macroeconomic Data.” Land Economics 80(4):529-538. Kinnaman, TC, and D. Fullerton. 1997.”Garbage and Recycling in Communities with Curbside Recycling and Unit-Based Pricing." NBER Working Papers 6021, National Bureau of Economic Research, Inc. Kinnaman, TC, and D. Fullerton. 2000. “Garbage and Recycling With Endogenous Local Policy." Journal of Urban Economics 48(3):419-442. Linderhof, V., P. Kooreman, M. Allers, and D. Wiersma. 2001. “Weight-Based Pricing in the Collection of Household Waste: The Oostzaan Case.” Resource and Energy Economics 23(4):359-371. Lund, HF. 2001. The McGraw-Hill Recycling Handbook. New York: McGraw-Hill Book Co. Meneses, G.D., and AB. Palacio. 2005. “Recycling Behavior: A Multidimensional Approach.” Environment and Behavior 37(6) :837-860. 89 Podolsky, M.J., and M. Spiegel. 1998. “Municipal Waste Disposal: Unit Pricing and Recycling Opportunities.” Public Works Management and Policy 3(1 ):27-39. Pollack, RA, and ML. Wachter. 1975. “T he Relevance of the Household Production Function and its Implications for the Allocation of Time.” Journal of Political Economy 83(2):255-277. Saltzman, C., V.G. Duggal, and ML. Williams. 1993. “Income and the Recycling Effort: A Maximization Problem.” Energy Economics 15(1):33-38. Saphores, J.M., H. Nixon, O.A. Ogunseitan, and AA. Shapiro. 2006. “Household Willingness to Recycle Electronic Waste: An Application to California.” Environment and Behavior 38(2) :1 83-208. USEPA. 2006. RCRA Orientation Manual 2006. . [Cited November 30 2007]. Available from http://www.epa.qov/epaoswer/qeneraI/orientat/romggdf. Van Houtven, G.L., and GE. Morris. 1999. “Household Behavior under Alternative Pay-As-You-Throw Systems for Solid Waste Disposal.” Land Economics 75(4):515-537. Vining, J., and A. Ebreo. 1990. “What Makes a Recycler? A Comparison of Recyclers and Nonrecyclers.” Environment and Behavior 22(1):55-73. Wertz, KL. 1976. “Economic Factors Influencing Households' Production of Refuse.” Journal of Environmental Economics and Management 2(4) :263-272. 90 CONCLUSIONS The first two essays in this dissertation contribute to the current body of recycling and waste management literature by studying drop-off recycling behaviors and demand. The success of a drop-off recycling program is largely dependent on the participation of the public, and the first essay helps us to understand the profile of people who utilize drop-off recycling sites including the underlying factors influencing their usage. The results from this essay indicated that dr0p-off site utilization was influenced by demographic factors such as age, education, income and household size. This essay also concluded that attitudinal factors affect drop-off site visits. Utilization of drop-off sites increased for recyclers that feel recycling is a convenient activity and also for those that are more familiar with a recycling site. These findings signal the importance of promoting recycling facilities such as drop-off sites to encourage recycling activities. The results from the second essay suggest that the demand for drop-off recycling sites is affected by the site location and the site specific characteristics. Drop-off sites that are relatively near and accessible to a majority of the population are found to receive higher visits than sites that are less strategically located. Convenience-related site characteristics such as site operating hours, the number of recyclables accepted, acceptance of commingled recyclables and yardwaste are found to increase site visits. However, improving convenience- related characteristics alone is not sufficient to increase recycling site visits if the 91 site is not strategically located. Therefore, to increase utilization, policymakers should consider both location and facility characteristics before constructing a drop-off recycling facility. The third essay in this dissertation examined the effects of policy, income and demographic variables on the rate of recycling in Minnesota counties. The findings demonstrate that variable pricing of waste disposal and the enactment of a mandatory recycling ordinance significantly increase the rate of recycling. The results suggested that curbside recycling services and drop-off recycling sites were effective in increasing the rate of recycling only when implemented together. This essay also found some evidence that expenditures for public education about recycling increased a county’s recycling rate. 92 APPENDIX 1: Survey Instrument (Essays 1 & 2) ' 93 HOUSEHOLD RECYCLING DEMAND SURVEY Interview Version Site Name: Date: Enumerator: Introduction My name is [NAME] and I am a student at Michigan State University. I am currently working on a survey to study household recycling. In this survey we would like to know your experience as a recycler and your views with regard to recycling and the environment. Your input is important because it will help us learn more about household recycling demand and behavior as well as understand what can be done to improve household recycling activities. This interview should take around 10 minutes to complete. Consent Statement By continuing with this interview, you indicate your voluntary consent to participate in this study and have your answers included in the project data set. Your participation is voluntary. Your refusal to participate in or to withdraw from the study carries no penalty or loss of any benefits. You are free to not answer any of the questions that we will ask you. However, we hope that you will agree to answer the questions, as your answers are very important to this study. Answers are anonymous, and we will keep your individual views entirely confidential. Your privacy will be protected to the maximum extent allowable by law. If you have any questions or comments please contact me and if you have any questions concerning your rights as a survey participant, please contact the Director of the Human Research Protection Program in MSU. Hand out contact details. 94 Your Recycling Activities Please check lZl the appropriate box or write in your answers where appropriate. 1. About how many times have you been to this site in the past NE YEAR? (skip Question 2 if answer is 0) 2. About how many times have you been to this site in the past THREE M NTH ? 3. Is your driving today for the sole purpose of recycling? D YES (skip Question 4) III NO 4. What percentage of your driving on this trip today would you attribute to recycling? % 5. (Record the make and model of the respondent’s vehicle. If unsure, ask the respondent.) Make Model What is your vehicle year? 6. What are the materIaI(s) that you recycle at this site? (Check the relevant box(es) and write down the material(s) that is(are) mentioned but not listed) [I Newspaper CI #2 Plastic Milk Jugs Ci Cardboard Ci #2 Colored Plastic Jugs Ci Steel Cans I] Magazines E] Aluminum Ci Office Papers Ci Clear Glass Ci Mixed Papers Ci Brown Glass L—J Junk Mail I] #1 PETE Plastic Other Materialls) 7. Approximately how much time did you spend sorting the recyclables you brought in today? 95 8. Why did you choose to recycle at this site? Is it ..... (read all the options below and check the answer - the respondent can have more than one reason) Ci Because it is located near my residence Ci Because of the operating hours L__| Because of types of recyclables accepted E] Because the site is well maintained and organized Do you have any other reasons for visiting this site: 9. Are you aware of other drop-off sites besides this one? El YES Ci NO (skip Question 11) 10. Can you name the site or sites that you are aware of? (Use map to help respondents identify the sites (ignore the site respondent is currently at). Ask the respondents about the number of visits for the sites they have identified) 2%?sz Number or Site Name Check if in the vrsrts In Aware past the past ME THREE YEAR? M ”T” '7 (1) Site 1 (2) Site 4 (3) Site 5 4) Site 6 (5) Site 2 6) Site 3 7) Site 7 (8) Site 8 Other (specify): 96 11. Do you recycle at your residence? For eg. curbside recycling services at your property or other types of recycling services at your dormitory or apartment [I YES Ci NO (go to Question 15) 12. What are the material(s) you recycle at your residence? (Check the relevant box(es) and write down the material(s) that is(are) mentioned but not listed) i:i Newspaper Ci #2 Plastic Milk Jugs [:1 Cardboard Ci #2 Colored Plastic Jugs Ci Steel Cans CI Magazines El Aluminum El Office Papers C] Clear Glass Ci Mixed Papers CI Brown Glass CI Junk Mail D #1 PETE Plastic Other Materialis) 13. How satisfied are you with the recycling service at your residence? (Read the options and check the answer) El Very Satisfied Ci Satisfied D Indifferent El Unsatisfied I] Very Unsatisfied 14. How are you charged for the waste collection service at your residence? (Read the options and check the answer) I] Fixed Fee for eg. a flat charge for any volume of trash E] Incremental Fee for e.g. a variable charge according to the number of trash bags or the size of trash bin Ci Included in Taxes or Rent Ci No Charge CI Don't Know 97 Your Opinions on Recycling and the Environment 15. Please indicate the extent to which you agree or disagree with the following statements, taking into consideration your recycling experiences. The scale ranges from Strongly Agree, Agree, Neither Agree nor Disagree, Disagree and Strongly Disagree. 5 § 3 g 0 a 3 e§ g .15 > 5205) a _ O 2 :0 D c a s e z a For me, household recycling is a difficult task. I do not have enough time to sort the materials for recycling. I do not have enough space to store the materials for recycling. The recyclables that I store attract pests. I am familiar with the recycling facilities in my area. I am familiar with the materials accepted for recycling in the recycling facilities in my area. My neighbors expect me to recycle household materials. My friends expect me to recycle household materials. My family expects me to recycle household materials. I feel good about myself when l recycle. 98 16. Based on your knowledge and opinion about recycling and the environment, please indicate your level of agreement with each of the following statements. The scale ranges from Strongly Agree, Agree, Neither Agree nor Disagree, Disagree and Strongly Disagree. a gg ti 28 < 3 (a a. Da- 2‘ 5:" 26 ii £12 2 #03 0 (DO :75 2‘ Recycling is a major way to reduce pollution. Recycling is a major way to reduce wasteful use of land for landfills. Recycling is a major way to conserve natural resources. Recycling will improve environmental quality. 17. Based on what you believe about the contribution of your recycling activities to the environment, please indicate your level of agreement with each of the following statements. The scale ranges from Strongly Agree, Agree, Neither Agree nor Disagree, Disagree and Strongly Disagree. 3' £3 a a .>.' E “that 9226.323 2 géowo a 2 I believe that my recycling activities will help reduce pollution. I believe that my recycling activities will help reduce wasteful use of land for dumps. I believe that my recycling activities will help conserve natural resources. I believe that my recycling activities will help improve environmental quality. 99 Questions about You 18. (Check the respondent's gender) [:1 Male [:1 Female 19. In what year were you born? 19 20. Which of the following best describes the highest level of education you have completed? Would you say. ...(Fiead the options and check the answer) 1:] some High School El High School or GED 1:] Vocational or Trade School [:1 Two Year Degree 1:] Four Year Degree [:1 Graduate School (9.9. MS, PhD, MD) 21. Which of the following best describes your employment situation? (Read the options and check the answer) El Employed Full Time 1:] Employed Part Time 1:] Unemployed E] Retired 1:] At Home Parent 1:] Student 22. What was your annual household income before taxes in 2005? $ (If respondent is reluctant to reveal exact income proceed to question A.) A. The median household income in Michigan is around $45,000. Which of the following best describes your income level? El below $45,000 (proceed to 8) [:1 $45,000 and above (proceed to C) B. Is your household income less than $25,000? El YES El ND C. Is your household income more than $75,000? El YES (proceed to D) El NO D. Is your household income more than $100,000? El YES DNo 100 23. What is your marital status? (Read the options and check the answer) 1: Single El Married/ Living with partner [:1 Divorced/ Widowed/ Separated 24. How many persons are living in your household (including yourself)? (go to Question 29 if answer is 0) 25. How many children under 18 years old are living in your household? (go to Question 29 if answer is 0) 26. How many children under 6 years old are living in your household? 27. What extent would you agree or disagree with the following statement: Your child or children play an important role in influencing you to recycle. 1:] Strongly Agree C] Agree 1:] Neither Agree nor Disagree 1:] Disagree 1:] Strongly Disagree 28. Do you or any member of your household belong to any environmental organization? E] YES 1:] NO In this study, we are trying to measure your driving distance to a recycling site. It is important that we know where you live (or the nearest street intersection from your residence) so that we could accurately measure your driving distance. Would you be willing to share your address with us? 29. Where do you live? Street City Zip Code ~ End of the survey ~ 101 APPENDIX 2: Tables (Essays 1 & 2) 102 Table A2.1. Attributes of drop-off recycling sites Site name HOURS NUMREC COMMING YA RD WASTE Site 1 168 9 O 1 Site 2 15 20 2 1 Site 3 35 12 1 1 Site 4 168 3 1 0 Site 5 168 15 2 0 Site 6 168 10 2 0 Site 7 168 12 o 0 Site 8 35 6 1 1 103 Table A2.2. Interview schedule Visitation Time Recycling Site Week 1, Sat 3 to 6 Week 1, Sun 9 to 12 Week 1, Sun 3 to 6 Week 1, Mon 3 to 6 Week 1, Tue 9 to 12 Week 1, Tue 3 to 6 Week 1, Wed 3 to 6 Week 1, Thu 9 to 12 Week 1, Thu 3 to 6 Week 1, Fri 9 to 12 Week 1, Fri 3 to 6 Week 2, Sat 3 to 6 Week 2, Sun 9 to 12 Week 2, Sun 3 to 6 Week 2, Mon 3 to 6 Week 2, Tue 9 to 12 Week 2, Tue 3 to 6 Week 2, Wed 3 to 6 Week 2, Thu 9 to 12 Week 2, Thu 3 to 6 Week 2, Fri 9 to 12 Week 2, Fri 3 to 6 Week 3, Sat 3 to 6 Week 3, Sun 9 to 12 Week 3, Sun 3 to 6 Week 3, Tue 9 to 12 Week 3, Wed 3 to 6 Week 3, Thu 9 to 12 Week 4, Tue 9 to 12 Week 4, Wed 3 to 6 Week 4, Thu 9 to 12 Site 5 Site 1 Site 6 Site 6 Site 7 Site 7 Site 4 Site 6 Site 3 Site 8 Site 5 Site 1 Site 4 Site 5 Site 5 Site 6 Site 4 Site 1 Site 3 Site 4 Site 1 Site 3 Site 3 Site 7 Site 7 Site 2 Site 2 Site 8 Site 2 Site 2 Site 8 104 Table A2.3. Distribution of cars according to sites Site Name Cars Percentage Site 2 202 35.19% Site 1 146 25.44% Site 3 113 19.69% Site 5 50 8.71% Site 4 29 5.05% Site 6 26 4.53% Site 8 5 0.87% Site 7 3 0.52% 105 APPENDIX 3: Probability Estimation of Variable Pricing Structure (Essay 3) 106 Table A3.1 presents the results for the logit estimation of the probability to implement variable pricing for waste disposal services in Minnesota counties. The exogenous variables not estimated in the main recycling rate equation (Chapter 3: Equation 12) are Yard and RecExp respectively. Yard represents a dummy variable for the ban on landfilling of yardwaste and RecExp represents the county recycling capital and operating expenditure. Table A3.1. Logit estimation of the probability to implement variable pricing structure (Dependent variable = VarP) Coefficient Std. Error Inc -0.00003 0.00004 Age 01 1 2 0.038*** Educ 0.020 0.031 Den 0.005 0.003* Ordin 0.650 0.262" EduExp 0.262 0.237 Yard 2.585 0.500*** RecExp 0.075 0.021 *** Year Dummies Yes Constant 2.653 1 .630* n 774 Pseudo Fi-squared 0.11 Log-likelihood -329.553 ***u < 0.01, **a < 0.05, *0 < 0.10 107 ,4, Ht. ,.._, I ' lllltllllt