Citation
Essays in the economics of solid waste management and recycling

Material Information

Title:
Essays in the economics of solid waste management and recycling
Creator:
Martinez, Salvador
Publication Date:
Language:
English
Physical Description:
ix, 101 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Bottles ( jstor )
Fees ( jstor )
Households ( jstor )
Landfills ( jstor )
Municipal solid waste ( jstor )
Prices ( jstor )
Pricing ( jstor )
Recycling ( jstor )
Solid wastes ( jstor )
Yard waste ( jstor )
Dissertations, Academic -- Economics -- UF
Economics thesis, Ph. D
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2004.
Bibliography:
Includes bibliographical references.
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Salvador A. Martinez.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
022825245 ( ALEPH )
880637195 ( OCLC )

Downloads

This item has the following downloads:


Full Text










ESSAYS IN THE ECONOMICS OF SOLID WASTE MANAGEMENT AND RECYCLING















By

SALVADOR A. MARTINEZ


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2004






























This dissertation is dedicated to my parents, Manuel and Laurell, and my siblings, Manjo
and Felice, for their encouragement, support, and inspiration.














ACKNOWLEDGMENTS

I thank my undergraduate economics professors at Weber State University for stimulating my interest in economics early in my studies. Also, I thank the economics faculty at the University of Florida for imparting knowledge of tools and methods for undertaking economic research. I would like to thank researchers at the Public Utility Research Center (UF), Bureau of Economic and Business Research (UF), Resources for the Future, and the University of Florida for providing opportunities for undertaking economic research and policy analysis.

I thank my supervisory committee members Steven Slutsky and David Figlio for their keen insights into public economics and guidance in research. I am grateful to have Donna Lee as an outside member on my committee. I give special thanks to my dissertation chair, Larry Kenny, for his insights into economic research, his financial support toward of my dissertation expenses, and his patience and understanding.















TABLE OF CONTENTS
Page

ACKNOWLEDGMENTS....................................................................1..1i

LIST OF TABLES ............................................................................................................. vi

LIST OF FIGURES .......................................................................................................... vii

ABSTRACT ..................................................................................................................... viii


CHAPTER

1 INTRODUCTION .................................................................................................... 1

2 ADOPTION OF STATE SOLID-WASTE AND RECYCLING POLICIES .......... 4

Introduction ........................................................................................................ 4
Solid W aste and Recycling Policies .................................................................... 7
M ethodology and Explanatory Variables ........................................................... 10
Results .................................................................................................................... 17
Conclusion ........................................................................................................ 21

3 DETERMINANTS OF LANDFILL TIPPING FEES

Introduction ...................................................................................................... 31
Landfill Operations and Costs under Regulation ............................................. 33
M ethodology and Estimation ............................................................................. 34
Results .................................................................................................................... 40
Conclusion ......................................................................................................... 47

4 DETERMINANTS OF HOUSEHOLD RECYCLING: A MATERIAL-SPECIFIC
ANALYSIS OF RECYCLING PROGRAM FEATURES AND UNIT PRICING

Introduction ...................................................................................................... 54
Prior Research and a Conceptual Framework ................................................. 57
Data Description ............................................................................................... 60
M odel Specification .......................................................................................... 65
Results .................................................................................................................... 67
Recycling Program Features .................................................................. 69








Unit Pricing Policy V ariables ............................................................... 74
Socioeconom ic Factors ........................................................................... 75
Data Lim itations ................................................................................... 77
Conclusion and Policy Im plications .................................................................. 79

5 CON CLU SION ................................................................................................. 92

REFEREN CE LIST ...................................................................................................... 96

BIO G RA PH ICA L SKETCH ........................................................................................... 101














LIST OF TABLES


Table Page

2-1. Summary statistics for recycling policy regressions .......................................... 26

2-2. States adopting requirement for local government recycling programs .............. 26

2-3. Total landfill material bans OLS and fixed effects regressions ........................... 27

2-4. Recycling grants and loans probit regressions ................................................... 28

2-5. Recycling tax incentives probit regressions ........................................................ 29

3-1. Summary statistics for tipping fee regressions ................................................... 50

3-2. Tipping fee regressions using small-market definition ....................................... 51

3-3. Tipping fee regressions using medium-market definition ................................... 52

3-4. Tipping fee regressions using large-market definition ........................................ 53

4-1. Previous studies on effects of unit pricing and recycling programs on effort .......... 84

4-2. Metropolitan Statistical Areas sampled ............................................................... 85

4-3. Unit pricing program s .......................................................................................... 86

4-4. Summary statistics for independent variables in recycling logit regressions ..... 88 4-5. Proportions of materials recycled ........................................................................ 89

4-6. Recycling participation ordered logits ................................................................. 90

4-7. Marginal effects of significant policy variables from logits ............................... 91














LIST OF FIGURES

Figure Page

2-1. Total number of landfill material bans ............................................................... 23

2-2. States with recycling grants or loans ................................................................... 24

2-3. States with recycling tax incentives .................................................................... 25

3-4. M ean tipping fees by state ................................................................................... 49

4-1. Distribution of unit prices .................................................................................... 87















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

ESSAYS IN THE ECONOMICS OF SOLID WASTE MANAGEMENT AND RECYCLING

By

Salvador A. Martinez

August 2004

Chair: Lawrence Kenny
Major Department: Economics

My study examined the determinants of the implementation of state solid waste

and recycling policies (landfill material bans, "market development" initiatives, recycling program availability) and how solid waste management and recycling policies and other factors influence the prices set by landfills to accept solid waste and the percentage of various materials recycled by households. I examined state adoption of landfill material bans, recycling grants and loans, recycling tax incentives, and requirements for local government to offer recycling using state panel data from 1988 to 1999 when the U.S. Environmental Protection Agency implemented stricter guidelines on landfills. Less Republican control of state government, greater environmental organization membership, and a greater percentage of the state population in metro areas positively affected the adoption of these pro-environmental policies.

A national panel dataset on landfills is used (supplemented with other data) to

examine the effects of government regulations, cost factors, and competition on tipping








fees while investigating economies-of-scale effects. Implementing landfill material bans for automobile tires and motor oil reduces tipping fees; automobile battery bans, white good bans, and yard waste bans raise tipping fees. Hence, the cost savings from avoided treatment of the different materials and monitoring costs due to implementing the bans are different. Landfills in coastal counties and areas with higher leachate costs experience higher tipping fees. Surprisingly, increased competition does not result in lower tipping fees. Finally, a large percentage of the landfills could continue to accept more solid waste per year and decrease costs.

My study then uses a household-level dataset (covering 20 MSAs in the U.S.) to examine the impact of recycling and unit pricing program availability on the percentage recycled of five different materials: glass bottles, aluminum, plastic bottles, newspaper, and yard waste. The availability of curbside recycling has a stronger effect than drop-off recycling across all materials. Effects from mandatory recycling on the percentage recycled is inconclusive. Also, there is no evidence that unit pricing (price of household waste disposal varying with amount discarded) increases the percentage of materials recycled by households.













CHAPTER 1
INTRODUCTION

Chapter 2 uses a variety of econometric models to examine the determinants of the following state solid waste management policies: landfill material bans, state provision of recycling grants and loans, state provision of recycling tax incentives, and state requirements for local governments to implement recycling programs. Using statelevel data from 1988 to 1999, I use a set of covariates to explain the adoption of these various pro-environmental policies in a median voter framework. Results indicate that Republican control of state government results in less-strict pro-environmental standards being implemented. There is evidence that states with more pro-environmental preferences (proxied by the percentage of the state population belonging to the National Audubon society) implement stricter solid waste and recycling policy. In addition, states with older populations provide more landfill bans, and have a higher probability of having recycling grants or loans. As a consistently strong voting group, retirees could be influential in shifting the position of the median voter. Finally, higher concentrations of the state population in metropolitan areas increase the probability of states implementing requirements for local governments to offer recycling programs. Transaction costs are lower for grassroots environmental lobbying in more densely populated areas, facilitating support for local recycling policies such as local recycling availability.

Chapter 3 examines the impacts of three factors on tipping fees (the prices set by U.S. solid waste landfills to accept waste) while looking at scale effects of operation. Regulatory controls, operating costs, and competition affect tipping fees in different





2

ways. In terms of regulatory controls, landfill bans put in place by state government restrict certain materials from being deposited in landfills. Having automobile tire bans and motor oil bans in effect reduces tipping fees, while automobile battery bans, white good bans, and yard waste bans result in higher tipping fees. This suggests the tire and motor-oil bans result in net cost savings for the landfill where the labor costs of monitoring the bans are lower than the costs of handling toxic materials and dealing with increased risks (such as tire fires and vermin problems). Landfills located in areas with greater precipitation must use more resources in controlling leachate and protecting groundwater sources. A major portion of long-run operating costs, including leachate collection (the handling and disposal of sludge-type material from the landfill) are proxied by weather variables. Increasing the amount of precipitation increases the tipping fees set by landfills. Other operating costs (such as labor wages) increase tipping fees. Also, an increase in the shadow cost of expanding landfills, agricultural land prices, results in higher tipping fees. Regarding scale effects, there is evidence that a substantial percentage of landfills could accept greater amounts of waste and experience economies of scale.

Chapter 4 examines the percent recycled of five materials (glass bottles, plastic

bottles, aluminum, newspaper, and yard waste) by households using a dataset covering 20 metropolitan statistical areas. The households are located in communities with various recycling program regimes and unit pricing programs. Access to curbside recycling has a significant and positive effect on the percentage recycled of all five materials. The length of a recycling program's existence has a significant effect on two of the recyclable materials. There is no significant evidence that mandatory recycling increases the






3

percentage recycled for any of the materials. Finally, the level of the marginal cost of disposal facing the households (the unit price) is insignificant in the regressions.













CHAPTER 2
ADOPTION OF STATE SOLID WASTE AND RECYCLING POLICIES Introduction

Many state governments are implementing solid-waste management and recycling policies (to increase recycling activity, and reduce the amount of municipal solid waste entering landfills) to address one of America's continuing environmental challenges, municipal solid-waste disposal.' The concerns surrounding solid-waste disposal have escalated as several landfills are reaching capacity and many older landfills have leaked hazardous waste into groundwater (Menell, 1990). Building new landfills is not a popular option among various stakeholders, especially those citizens residing in areas near proposed landfills.2

A variety of factors may influence the state adoption of solid-waste management and recycling policies. Citizens who value recycling3 may be represented by environmental organizations that generate political action with lobbying, grassroots pressure, public advocacy, and research. State legislators are concerned about their state becoming a regional dumping ground for solid waste imported from other states.

The aim of my study is to analyze the factors influencing state governments in adopting solid-waste management and recycling policies when the Environmental Protection Agency (EPA) implemented stricter landfill standards. Landfill material bans, Kinnaman and Fullerton (The economics of residential solid waste management, NBER working paper, 1999) review literature on residential solid-waste management.
2 Nelson et al. (1992).
3 Aadland and Caplan (1999) and Jakus et al. (1996) discuss residential valuation of curbside and drop-off recycling, respectively.








state grants and recycling loans, and state requirements for local government to develop recycling are the solid-waste management and recycling policies investigated in the paper using econometric analysis with data from 1988 to 1999. 1 use various explanatory variables to capture political and socioeconomic forces within the states that affect the adoption of pro-environmental solid-waste management and recycling policies.

The Resource Conservation and Recovery Act (RCRA) has largely left solid-waste management and recycling policy to be determined by state governments. Surprisingly, the economics literature investigating the adoption of solid-waste and recycling policies at the state level is non-existent. To my best knowledge, the only existing empirical research on solid-waste management and recycling policy adoption is at the community level. The state's reasons for implementing solid-waste and recycling policies are quite different than those at the local level.

Determining state policies can be considered more important because states set the rules, and local governments operate within those guidelines. Unlike local governments developing solid-waste and recycling policy at the local level without state guidelines, local governments in states with solid-waste and recycling policies are constrained in their decision making. Any losses or gains resulting from operating recycling programs are absorbed by the local governments. However, the programs can easily become more costly than anticipated, as there is no cost sharing between local governments and state governments.4 Also, states are able to solve externality problems among communities by implementing solid-waste and recycling policies.



4 Kinnaman (2000) looks at recycling revenues and costs within a single community. He found the local economy paid an average of $102.45 per ton to recycle the material, an amount just over $55 per ton more than the cost of disposing in the landfill.








Feiock and West (1993), Tawil (On the political economy of municipal curbside programs: Evidence from Massachusetts, working paper, 1995), Callan and Thomas (1999), and Mrozek (2000) examine the adoption of solid-waste management and recycling policies at the local level. Except for Feiock and West (1993), who use a crosssectional dataset, the studies use data from communities within a particular state. In addition, only Mrozek (2000) uses data from more than one time period. Unlike theirs, my study uses a national dataset from 1988-1999, thus considering variation across states and over time. For example, the panel data set allows me to estimate state responses to stricter landfill operating guidelines imposed by the EPA.

Previous authors examine only one type of solid-waste or recycling policy (such as adoption of some type of recycling program, curbside recycling, and unit-pricing for residential waste disposal). I examine state landfill material bans, state provision of recycling grants or loans, state provision of recycling tax incentives, and requirements for local governments to implement recycling programs; and thus cover two of the four major categories of the EPA's agenda of integrated solid-waste management (source reduction, landfilling, combustion, and recycling).

Finally, my analysis includes a richer set of covariates to explain the adoption of solid-waste and recycling policies. The impact of regional spillover effects from solid waste are examined. With states unable to prevent interstate shipments of solid waste, some state governments may be implementing pro-environmental solid-waste and recycling policies to limit the growth and size of their home landfills and deter imports of solid waste. I also am able to include proxies for the size and strength of environmental groups (League of Conservation Voter scores averaged for each state, and membership in








the National Audubon Society); and for political party control of state government (including governor office and state legislature).

Results from including a measure of political party control suggest that

Republican control of state government results in less-strict standards, which is consistent with Democrats supporting greater environmental protection then Republicans. Results from including a measure of National Audubon membership confirm that states with a larger environmental movement implement stricter solid-waste and recycling policies. Other results indicate that states with older populations provide more landfill bans, and are more likely to have recycling grants or loans. This is consistent with retirees having a greater desire to leave a legacy, in terms of environmental protection. Higher concentrations of the state population in metropolitan areas increase the probability of states implementing requirements for local governments to offer recycling programs, thus supporting the hypothesis that metropolitan areas are conducive to pro-environmental lobbying.

Solid Waste and Recycling Policies

The Resource Conservation and Recovery Act of 1976 was amended by the

Hazardous and Solid Waste Amendments of 1984 (HSWA), directing the Environmental Protection Agency to develop a set of minimum criteria for solid-waste management facilities that receive either household hazardous waste or small quantities of exempt hazardous waste. The criteria were not put forth until October 1991. Both existing and new municipal solid-waste landfills were affected by the rules, which became effective in October 1993. The criteria included minimum standards regarding location restrictions, operations, design, groundwater monitoring and corrective action, closure and postclosure care, and financial assurance. In addition, the EPA required states to develop








and implement permit programs to ensure compliance no later than April 1993. This increase in federal intervention in solid-waste management policy brought attention to the choice to implement other policies that make it easier to comply with federal regulations.

The first solid-waste management policy to consider is the landfill material ban. A growing concern over landfill expansion and the desire to increase recycling opportunities (coupled with the new changes to RCRA) has supposedly led state governments to pass legislation that bans particular materials from the landfill. The landfill material bans I use in the econometric analysis are motor oil, vehicle batteries, white goods, tires, and yard waste. Environmental concern regarding vehicle batteries and motor oil is directly related to the contaminants from these materials escaping from the landfill area to the surrounding areas, especially to those areas with water sources and residential populations. White goods and tires are bulky items that take up larger amounts of space in the landfill, and can be "remanufactured" into other materials. White goods, a class of appliances including stoves, refrigerators, and clothes dryers, may contain components in their electrical wiring that can be harmful if leaked into water supplies. Finally, I consider landfill material bans on yard waste. Yard waste is estimated to be the second largest contributor to national municipal solid-waste generation, behind paper and paperboard (Kreith, 1994).

Since the landfill material bans collectively aim at extending the capacity of

landfills and lowering health risks, the landfill material bans are measured as the number of these five materials that are banned (TOTAL BANS). In 1988, about 96% of the states had passed landfill bans for two or fewer of the five mentioned materials. This figure has slowly decreased to 37% in 1999. The mean number of TOTAL BANS each year for all








states has steadily increased over time but only nine states had passed landfill material bans for all of the materials by 1999. Vehicle batteries are the most popular material to be banned, with over 80% of the states having the material banned from 1993 to 1999. Fewer states passed bans for white goods (only 31% of the states had a white good landfill ban in 1999). The greatest push for landfill bans seems to be concentrated in the Midwest and on the western and eastern coasts. Some of the states in the Rocky Mountains have not aggressively implemented landfill material bans. Figure 2-1 maps state values for TOTAL BANS in 1992 and 1999.

Next, I consider the adoption of state recycling grants, loans, and tax incentives. Several states provide grants and loans to both local governments and businesses, to implement recycling programs and stimulate demand for recycled products. Some grants are given on a competitive basis, while others are "block" grants. States like New Jersey and Pennsylvania give grants to local jurisdictions, as a reward tied directly to the amount of a material recovered from the solid-waste stream. Tax incentives include such items as tax breaks or exemptions on equipment or materials used in manufacturing final products using post-consumer recycled content or on machinery used in recycling facilities. Collectively, recycling grants, loans, and tax incentives are referred to as recycling "market development" initiatives in the recycling industry.

The dependent variables constructed from information on these initiatives include GRANTLOAN and TAX INCNTV, which are used only for 1992 and 1999, because of data availability. The first variable takes on the value 0 or 1; a positive value indicates that the state offered recycling grants or loans in the particular year. Likewise, the variable TAX INCNTV takes on the value 0 or 1, reflecting whether a state offered








recycling tax incentives in the particular year. Half the states provided no types of grants, loans, or tax incentives in 1992. This figure decreased to six states in 1999. Some states, that provided market development initiatives in 1992, did not do so in 1999. Figures 2-2 and 2-3 show states with positive values for GRANTLOAN and TAX INCNTV in 1992 and 1999.

The final state recycling initiative I consider is the adoption of state legislation

requiring local government units to develop recycling programs (LOCAL RECYCLING). The local government units may be counties, cities and counties, cities, or solid-waste management districts. Adoption of these policies is concentrated in the western states, and a major part of the east coast. Unlike landfill bans and market development initiatives, states are not reversing the legislation on the development of these local recycling programs. So I estimate hazard regressions, to examine the determinants affecting adoption of these requirements, using the same covariates as used in other regressions.

Data for TOTAL BANS, GRANTLOAN, TAX INCNTV, and the adoption of state legislation requiring local government units to develop recycling programs come from Glenn (1992a, 1992b, 1998a, 1998b, 1999), Goldstein (1997a, 1997b, 2000a), Goldstein and Madtes (2000), Kreith (1994), Raymond Communications (State recycling laws update, 1999), Steuteville (1994a, 1994b, 1995a, 1995b, 1996a, 1996b), Steuteville and Goldstein (1993), and Steuteville et al. (1993).

Methodology and Explanatory Variables The decision to adopt state solid-waste and recycling policies can be examined in a median voter framework. State solid-waste and recycling policies determined by the median voter, where the number of voters preferring a stricter amount of policy (P) is








equal to the number of voters preferring a weaker policy. The policy P* is preferred by the median voter to any other amount of solid-waste and recycling policy.

With state solid-waste and recycling policies being implemented by elected

officials, households vote for the candidate yielding higher utility levels. Citizens vote for state legislators within local voting districts across the state, and over time, in elections. Thus, voters select candidates (in various elections) whose choice of P is equal to their own preferred level of P. The amount of P implemented by state policy makers is expected to represent the amount of P desired by the median of the various district median voters. To explain the choice of solid-waste and recycling policies, a variety of socio-economic and political variables are used in the econometric analysis for the three classes of solid-waste and recycling policies.

Since landfills pose more of a health risk in densely populated areas, landfill regulation should be more likely in densely populated areas. The population density variable POP DENSITY is the number of persons per square land mile (in thousands) in the state. The variable is constructed from data on state land and water area size in 1987 defined by the U.S. Geological Survey and U.S. Census Bureau data on state population from 1988-1999. States with higher POP DENSITY are expected to be more likely to implement pro-recycling initiatives.5

Landfills in areas with less expensive land can expand landfill space at a

relatively low cost. To counteract the low land prices, policy makers may be more inclined to pass landfill bans and other pro-recycling policies to counteract potential landfill growth. Using data from the U.S. Agricultural Censuses in 1992 and 1997, AG


5 Mrozek (2000) expects higher urban density to explain economies of scale in the local government decision to implement curbside recycling. His results are mixed.








LAND equals the market value of land and buildings per acre. The coefficient for AG LAND is expected to be negative; that is, the higher price of land should result in fewer pro-recycling policies.

Previous economic research investigated whether environmental goods are

normal. Evidence supports the notion that individuals with higher socioeconomic status are more likely to support pro-environmental activities. Using per capita income data from the Bureau of Economic Analysis and the personal consumption price expenditures implicit price deflator, I calculate real personal income (in thousands) per capita (INCOME) from 1988 to 1999 to determine whether solid-waste and recycling policies are "normal." The expected sign for the INCOME coefficient is positive, but Mrozek (2000) finds no effect, Callan and Thomas (1999) find a negative effect, and Feiock and West (1993) find a positive effect.

Previous studies investigating household-level recycling behavior use age as a characteristic to explain recycling quantities. Using annual estimates from the U.S. Census Bureau, AGE65 is defined as the percentage of the state population that is age 65 or over for years 1988-1999. Illinois, Florida, and Pennsylvania had the highest percentage of residents age 65 and older in 1999. Small (Preserving family lands, land and people, Trust for Public Land, San Francisco, CA, 1996) notes that several elderly are holding onto land, as compared to selling, to prevent development. Older individuals could be supporting pro-environmental protection in recycling to leave a legacy.6 With older populations likely to be more concerned regarding environmental protection, the


6 Kahn (2000) investigates demographic change and the demand for environmental regulation using demographic covariates in California environmental ballots, local government environmental expenditures across the U.S., and Congressional voting on the environmental law regressions. Results on the variable percentage of the population 65 years or older are mixed.








coefficient for AGE65 is expected to be positive.7 Also, retirees are attracted to amenity rich areas with scenic beauty. Thus, AGE65 is also a measure of the stock of natural amenities.

On the national, level policy makers have debated whether states should be able to restrict interstate shipments of solid waste. State politicians are concerned about their states becoming the dumping ground in their region. Thus legislators may be passing pro-recycling initiatives to reduce the growth of waste products generated in the state, and also may be legislating to deter imports of solid waste. The greatest amount of interstate waste shipments occur in the Midwest and East. For example, Feliciano and Worth (1997 summary of Indiana solid waste facility data, IDEM Office of Solid and Hazardous Waste Management) report that 27% of annual waste disposal in Indiana came from out of state. Most of the out-of-state waste Indiana receives is from Illinois.

Data on municipal solid-waste generation at the state level from BioCycle's annual survey8 and state population data from the U.S. Census Bureau are used to construct the explanatory variable WASTE, which is the weighted average waste generation per capita of the states contiguous to the particular state for years 1988-1999. The more waste generated by neighbors, the more likely the home state will implement the solid-waste management and recycling initiatives.

To capture the size and strength of environmental groups in the state, I include three different measures. First, using membership data from the Audubon Society and

7 Callan and Thomas (1999) use a community's median age and its square as covariates explaining the adoption of unit-pricing, a program where households pay for each unit of trash disposal. They suggest that waste levels are low at the ends of the life cycle and the need for unit-pricing follows waste generation over the life cycle.
8 Glenn (1 992a, 1992b, 1998a, 1998b, 1999), Goldstein (1 997a, 1997b, 2000a), Goldstein and Madtes (2000), Steuteville (1 994a, 1994b, 1995a, 1995b, 1996a, 1996b), Steuteville and Goldstein (1993), and Steuteville et al. (1993).








U.S. Census Bureau state population estimates, I calculate the percentage of the state population belonging to the National Audubon Society from 1988-1999. I was not able to obtain complete membership data from the Sierra Club or National Wildlife Federation. In states with a higher percentage of members in environmental groups, it is expected the median voter would prefer a stricter standard. In addition, legislators are expected to represent the preferences of a median voter in a median voter model. The League of Conservation Voters' (LCV) average rating for each state's U.S. Senators and Representatives for 1988-1999 is another measure of the median voter preferences. The LCV rating ranges from 0 to 100, and measures support for environmental initiatives (100 implies full environmental support). Higher Audubon membership percentages and LCV scores should result in more states implementing pro-recycling initiatives and landfill material bans.

Transaction costs of organizing and promoting pro-environmental legislation should be reduced when individuals and groups are in close proximity. With reduced transaction costs and pro-environmental members finding greater ease in disseminating information, the percentage of the populace with pro-environmental ideology would be expected to increase. Hence, the position of the median voter would change, resulting in a stricter P*. I use the percentage of the state population living in metropolitan areas (METRO POP) to capture this occurrence.9 The U.S. Statistical Abstract contains values available in even years from 1988 to 1998; interpolation is used for the odd years. A


9 Callan and Thomas (1999) use a rural/non-rural indicator in their regression explaining the adoption of unit-pricing and find a negative and statistically significant effect using an economies of scale hypothesis for its inclusion.








higher percentage of the state population living in metropolitan areas is expected to result in more pro-environmental solid-waste management and recycling legislation.

Finally, I define a political variable REPUBLICAN CONTROL from the Council of State Governments (1989, 1991, 1993, 1995, 1997, 1999, 2001), which takes on the value 0 if both houses of the state legislature and the governor's position are controlled by Democrats, the value 2 if Republicans are in control of both houses and the governorship, and the value 1 if the political power in the state is shared by both parties for 1988-1999. The hypothesis is that state government controlled by the Republicans should be less likely to implement environmental protection (or solid-waste management and recycling initiatives) given the ideological positions of the party. Their ideology is anti-regulation and thus anti-environment. However, interpretation can be complex, as Republicans represent higher-income individuals who favor stricter standards.

To capture the effects of the changes in RCRA, dummy indicators are included for two time periods: 1991-1993 and 1994-1999. The default time frame is 1988-1990. These three breakdowns capture 1) the period when the EPA was formulating its requirements for landfills; 2) the transition period when states had to develop and implement permit programs, with landfills complying with EPA regulations; and 3) the period after the transition.

The variables are collected for all 50 states and the District of Columbia with

some exceptions. AG LAND is not available for DC. With Alaska and Hawaii not part of the continental U.S., the variable WASTE is not calculated for those states. REPUBLICAN CONTROL is not available for DC or Nebraska. So a maximum of 49 observations is possible using WASTE or using REPUBLICAN CONTROL. Table 2-1








gives summary statistics for these variables. Also, Table 2-2 list the states adopting local recycling requirements.

The econometric analysis uses a variety of models to examine the robustness of the effects of the independent variables in explaining the policies. First, Table 2-3 shows results from regressing TOTAL BANS on the previously described explanatory variables using Ordinary Least Squares (OLS) and Fixed Effects (state fixed effects) regressions for the period 1988 to 1999 with robust standard errors at the state level. Specification 1 is the base specification. Specification 2 adds the political variables REP CONTROL, AUDUBON, and LCV to the base specification, and drops WASTE. Finally, Specification 3 adds the WASTE variable to Specification 2. Two different tables show market-development initiatives. Table 2-4 gives probit regression results, showing whether the state provided recycling loans or grants for two time periods: 1992, and 1992 & 1999. The variable AG LAND is included in the 1992 regressions, using the same specification framework as for the landfill material bans. In a similar fashion, Table 2-5 presents the probit regressions for the dependent variable TAX INCNTV for years 1992 and 1992 & 1999. Finally, Table 2-6 shows results from Weibull proportional hazard regressions (investigating the adoption of a law requiring local government to offer recycling with the same specification framework used for TOTAL BANS).

Overall, the fit of the regressions for the various solid-waste and recycling

policies is reasonable, with many hypotheses supported. Low R-square values in the OLS regressions indicate that more variation of TOTAL BANS can be explained by variables not in the regression. Relative to the landfill material ban regressions, probits for GRANTLOAN and TAX INCNTV do not perform as well. The overall fit is good in








two of the three Weibull regressions explaining the adoption of a requirement for local government to implement recycling.

Results

AGE65 has a significantly positive impact on the number of landfill material

bans, and on provision of recycling grants; but is unrelated to use of tax incentives, and requiring local governments to implement recycling. The positive effect is consistent with the hypotheses of older populations considering environmental protection as a legacy, and areas with a greater percentage of retirees being rich in natural amenities and thereby warranting greater environmental protection. An increasing percentage of individuals over 65 (coupled with the fact that a large percentage of this population is voting), older individuals could be instrumental in shifting the position of the median voter. A one standard-deviation rise in AGE65 results in a. 12 to .50 increase in the total amount of landfill material bans, where significant. The estimated coefficients for AGE65 are relatively the same in Table 2-4 under various specifications. In Column 3-b, the marginal effect of AGE65 implies that increasing the percentage of the state population age 65 and over by 1% results in a 4% increase in the probability of providing recycling grants or loans. AGE65 is not statistically significant in Table 2-5 or 2-6. Age variables covering the other age groups were tested in the regressions. F-tests and Likelihood Ratio tests did not indicate that the group of other age classifications should be included.

The estimated coefficients on METRO POP have the hypothesized sign in all but two regressions, suggesting that metropolitan areas enable grassroots environmental groups to easily organize and lobby for pro-recycling policies. Also, the cost of recycling may be lower in states with more people living in metro areas. The coefficients are








significant in just one column in Table 2-4 and three columns in Table 2-5. The marginal effect for METRO POP in Column 2-b in Table 2-4 is .01 indicating that a 1% increase in the percentage of the state population living in metropolitan areas results in a .01 increase in the probability of providing recycling tax incentives. METRO POP is significantly positive in the first two specifications in Table 2-6. These specifications imply that in each year, a state with a higher percentage of the population living in metropolitan areas has a higher probability of adopting a requirement for local governments to offer recycling.

Where the coefficients for POP DENSITY are significant in the regressions for landfill material bans, grants and loans, and tax incentives, the coefficients are negative and don't provide support for the initial hypothesis that policy makers in regions with higher population density would implement these pro-environmental policies to combat the growth of landfills. The result for POP DENSITY is only positive and significant in the first specification in Table 2-6, for the recycling regressions.

The coefficients for INCOME are positive and significant in the fixed effects regressions in Table 2-3 (for landfill material bans and Table 2-4 for a couple of specifications in the short model); these results are consistent with the hypothesis on environmental protection being a normal good. A one standard-deviation rise in per capita income results in an increase of .38 to .43 landfill material bans. It could be that INCOME is picking up different effects in the various dependent variables. There is a time cost with recycling; so as income increases, the opportunity costs of devoting time to recycling increase. The coefficients on INCOME are negative but not significant in








the recycling regressions. This suggests that the income and time cost effects roughly cancel each other out.

Waste generation in neighboring states is expected to result in stricter solid-waste and recycling policies. Coefficients on WASTE are significant and positive only in the OLS regressions in Table 2-3, with a one standard-deviation increase in WASTE resulting in .24 and .23 increases in landfill material bans, for columns 1-a and 3-a, respectively. It could be that the state fixed effects are collinear with WASTE.

AG LAND was only available in years 1992 and 1997. The variable was

included in regressions for TOTAL BANS using the two years of data, but AG LAND was not statistically significant in a variety of specifications, which are not reported. When including the variable in the probit regressions, the estimated coefficient is negative in all three specifications in Tables 2-4 and 2-5, and is significant in two of the three specifications in Table 2-6. The negative coefficient is consistent with the hypothesis that policy makers are less likely to implement the recycling grants and loans as the price of agricultural land increases, because it is more costly for landfills to expand. The incentive to stimulate recycling among local governments and firms is decreased as the shadow price on waste disposal increases.

With the Republican party favoritism toward business and opposition to big government, there should exist significant differences between Republicans and Democrats regarding environmental concerns. The first political variable, REP CONTROL, is a measure of both partisanship and party ideology. Recall it takes the values of 0, 1, and 2 for Democratic control of both state legislative bodies and the governorship, mixed control, and Republican control of both state legislative bodies and








the governorship, respectively. The estimated coefficient in negative as predicted, and is and significant in all but one table.

The coefficient is negative in all specifications in Table 2-3, meaning more

Republican control of state government results in less landfill material bans being passed. Changing political control of the governorship and the legislative bodies from Democratic to Republican control results in a decrease of from .48 to .51 landfill material bans. The sign results for REP CONTROL are reversed between Tables 2-4 and 2-5, positive for grants or loans and negative for tax incentives. The former result is unexpected, but the grants and loans may be interpreted as subsidies for business and hence pro-Republican.

Next, AUDUBON is the percentage of the state population having membership in the National Audubon Society. The coefficients are not significant for the TOTAL BANS regressions or the recycling requirement hazards. However, the sign is significant in one of four specifications in the recycling grants and loans probits and is significant in three out of the four tax incentives probits in Table 2-5. Part of this difference between Table 2-3 and Tables 2-4 and 2-5 may be due to there no being enough year to year variation in the data. So there is some slight evidence that environmental activism may result in states adopting pro-environmental solid-waste and recycling policies.

As a result of electoral pressures, the environmental sentiment of the state's

Congressional delegation is expected to mirror that of the electorate. Thus, the average League of Conservation Voters (LCV) score for the state's Representatives and Senators is used as another measure of environmental sentiment in the state. The hypothesis that states with more pro-environmental sentiment are more likely to adopt pro-environmental








policies is supported in three of the four tables. Coefficients are positive and significant in Tables 2-3, 2-4, and 2-6, but are negative and significant in Table 2-5. Environmentalists may be more likely to support recycling grant and loan programs that provide local benefits than to support tax incentives which result in more "free financing" opportunities for businesses.

There is evidence more solid-waste and recycling policies are passed as the EPA implemented its landfill regulations. The dummy indicators for 1991-1993 and 19941999 are positive and statistically significant in all specifications in Table 2-3. Compared to 1988-1990, there were 1.31-1.46 more landfill material bans in 1991-1993 and 1.612.1 more in 1994-1999. When looking the at the 1994-1999 period indicator in the tax incentives and recycling grant and loan regressions, the probability of adoption was higher in 1999 than in the transition period.

Conclusion

This study is an investigation into the determinants of adoption of solid-waste management and recycling policies by states. It extends the literature by using state panel data covering several solid-waste management and recycling policies accounting for half of the EPA's platform for integrated solid-waste management, using a richer set of covariates (such as more political variables and a measure of interstate solid waste spillovers) and explaining state policies instead of local policies. I find that states with higher percentages of the population 65 or older tend to adopt more landfill bans and have a higher probability of having recycling grants or loans. Also, there is evidence that states with greater concentrations of population in metropolitan areas provide recycling tax incentives, and are more likely to adopt a requirement for the development of recycling programs at the local level probably indicating lower costs of implementing








recycling. When the waste generation around a home state is high, there is some evidence a state will implement more landfill material bans. Finally, the results indicate that Republican control results in less strict standards. The party effects are greater than those for any other variable. State membership in environmental groups or proenvironmental records of the state's Congressional delegation reflect the environmental preferences of the state's populace. I find that states with a more pro-environmental populace are more likely to adopt pro-environmental policies.

I initially stated the optimal amount of state solid-waste and recycling policies,

P*, is determined by the median voter and have included covariates to try and explain the selected pro-environmental policies. When examining the results from the different state policies, it seems the results of the landfill material bans regressions and the Weibull regressions on the requirement for local recycling have the most significant results consistent with initial hypotheses. One possible explanation for the more significant results in these regressions compared to other regressions in the tables is that the benefits from reduced disposal of potentially harmful materials and opportunities to recycle are likely to directly affect citizens more than the provision of recycling grants, loans, and tax incentives. A more comprehensive data set containing additional years of data for GRANTLOAN and TAX INCNTV, including actual amounts of grants and loans provided, would allow estimation techniques that might provide more insight into the determinants behind the adoption of state solid-waste management and recycling policies.

























1992 Landfill Material Bans
5 (7)
4 (6)
*3 (9)
LI 2 (12) 1 (9) 0 0 (8)


1999 Landfill Material Bans
[ 5 (9) * 4 (10) * 3 (13) ] 2 (10) 1I 0 (5)
[]0 (4)


2-1. Total number of landfill material bans A) In 1992. B) In 1999.


~ZZ~


Lr-7r--

















































1999 Recycling Grants or Loans
1 Recycling Grants or Loans (37) D No Recycling Grants or Loans (14)


Figure 2-2. States with recycling grants or loans. A) In 1992. B) In 1999.



















A




1992 Recycling Tax Incentives
* Recycling Tax Incentives (19) ] No Recycling Tax Incentives (32)














B


D No Recycling Tax Incentives (23)


Figure 2-3. States with recycling tax incentives. A) In 1992. B) In 1999.









Table 2-1. Summary statistics for recycling policy regressions Variables States Mean Standard Minimum Maximum Time Deviation Period TOTAL BANS 51 2.16 1.70 0 5 1988-99 GRANTLOAN 51 0.49 0.50 0 1 1992, 1999 TAX INCNTV 51 0.46 0.50 0 1 1992, 1999 AGE65 51 12.86 2.59 3.69 25.98 1988-99 METRO POP 51 66.92 21.56 20.04 100 1988-99 POP DENSITY 51 320.61 1146.19 0.92 9136 1988-99 INCOME 51 23.12 3.96 14.91 37.76 1988-99 WASTE 49 1.16 0.27 0.24 2.70 1988-99 AG LAND 50 1615.94 1543.64 184.02 7558.80 1992, 1997 REP CONTROL 49 0.91 0.63 0 2 1988-99 AUDUBON 51 0.19 0.10 0.07 1.05 1988-99 LCV 50 46.80 21.90 0 94.42 1988-99


Table 2-2. States adopting requirement for local government recycling programs 1988 1989 1990 1991 1992 1993-1999
MD, NY, PA MN, NC, VA, AZ, CT AR, NV, OR, NE, SD none
WA, AL, CA, SC, WV DC
Note: RI adopted in 1986 while NJ and VT adopted in 1987.






27


Table 2-3. Total landfill material bans OLS and fixed effects regressions Independent OLS Fixed OLS Fixed OLS Variables Effects Effects


AGE65 METRO POP

POP DENSITY INCOME

WASTE REP CONTROL


AUDUBON LCV

1991-1993 1994-1999

Constant


1-a
0.1885***
(0.0483) 0.0024 (0.0089)
-0.0003*** (0.0001)
0.0076
(0.0539)
0.9014** (0.4657)


1.3224*** (0.1950)
1.6741***
(0.2433)
-2.7087**
(1.0533)


1-b
0.0358
(0.0255) 0.0306 (0.0307)
0.0015**
(0.0005)
0.0964*** (0.0455) 0.1380 (0.3859)


1.3788*** (0.1977)
1.6390"**
(0.2225)
-4.3704** (1.9545)


2-a
0.1920*** (0.0563) 0.0090 (0.0083)
-0.0024** (0.0010) 0.0193 (0.0551)


-0.2088 (0.1313)
-0.7905 (1.5711)
0.0180** (0.0074)
1.4689***
(0.2124)
2.0117*** (0.2364)
-2.8062** (1.2184)


2-b
0.0587* (0.0322) 0.0288 (0.0301) 0.0050 (0.0150) 0.0977* (0.0513)


-0.2535** (0.1189)
0.6816
(0.6985) 0.0021 (0.0063)
1.3090***
(0.2124)
1.6623***
(0.2325) 4.7070*
(2.4073)


3-a
0.1777*** (0.0530) 0.0069 (0.0082)
-0.0023** (0.0011)
0.0223
(0.0652) 0.8527* (0.4456)
-0.1500
(0.1324)
-1.0409 (1.5754)
0.0196**
(0.0080)
1.4580***
(0.2189)
1.8221*** (0.1931)
-3.5279*** (1.2071)


Number of states Number of observations Adjusted R2 Root MSE


0.3728 1.3358


* 10% level of significance
** 5% level of significance
*** 1% level of significance


0.7602
1.4674


0.4081 1.4674


0.7641 0.8095


0.2531 1.4259


Robust standard errors are in parentheses.


Fixed Effects 3-b
0.0453* (0.0239) 0.0339 (0.0299)
0.0001 (0.0136)
0.1096**
(0.0558) 0.0441 (0.3553)
-0.2422** (0.1200) 0.4215 (0.6750)
-0.0016 (0.0058)
1.3180*** (0.2199)
1.6089***
(0.2267)
-4.1270* (2.2123)


0.7891
0.7978










Table 2-4. Recycling grants and loans probit regressions
Independent 1992 1992 & 1999 1992 1992 & 1999 1992 1992 & 1999
Variables
1-a 1-b 2-a 2-b 3-a 3-b
AGE65 0.0853 0.1130* 0.1521* 0.1382** 0.1240 0.1068* (0.0664) (0.0601) (0.0907) (0.0553) (0.0871) (0.0594) METRO POP -0.0135 -0.0079 0.0389* 0.0136 0.0320 0.0062 (0.0175) (0.0106) (0.0227) (0.0100) (0.0261) (0.0108) POP DENSITY 0.0040 -0.0002* 0.0080 -0.0011 0.0085 -0.0011 (0.0040) (0.0001) (0.0054) (0.0012) (0.0061) (0.0012) INCOME 0.2907*** 0.1237* -0.0322 -0.0153 -0.0038 0.0236 (0.1031) (0.0674) (0.1282) (0.0826) (0.1435) (0.0940) WASTE 0.5401 0.7912 0.9775 0.9714 (1.3157) (0.6727) (1.5060) (0.7088) AG LAND -0.0006 -0.0013* -0.0014* (0.0005) (0.0007) (0.0008) REP CONTROL 0.9869* 0.0016 1.0064 0.0077 (0.5498) (0.2215) (0.6254) (0.2395) AUDUBON 7.3819** 3.9903 6.5406 4.0359 (3.7593) (2.4915) (4.4308) (2.8178)
LCV 0.0564 0.0118 0.0615** 0.0134 (0.0241) (0.0100) (0.0257) (0.0117) 1994-1999 0.6621 1.4847*** 1.0989** (0.4273) (0.4767) (0.5554) Constant -7.6193*** -5.1799** -9.1998*** -4.2417*** -10.1064"** -5.3253***
(2.4491) (1.5332) (2.6783) (1.5803) (2.8575) (1.9389)

Number of states 48 49 49 49 47 47 Number of obs. 48 98 49 98 47 94 Prob > X 2 0.0125 0.0000 0.0059 0.0001 0.0032 0.0004 Pseudo R2 0.2216 0.2318 0.5116 0.2747 0.5065 0.2753
* 10% level of significance Robust standard errors in parentheses.
** 5% level of significance
*** 1% level of significance










Table 2-5. Recycling tax incentives probit regressions Independent 1992 1992 & 1999 1992 1992 & 1999 1992 1992 & 1999
Variables


AGE65 METRO POP POP DENSITY INCOME WASTE AG LAND REP CONTROL


1-a
0.0187 (0.0788) 0.0080 (0.0136) 0.0021 (0.0034) 0.0200 (0.0907)
-0.8501 (1.0904)
-0.0004 (0.0004)


AUDUBON

LCV

1994-1999


Constant


-0.3055 (2.2509)


1-b
-0.0154 (0.0665)
0.0227** (0.0094) 0.0003
(0.0006)
-0.1035* (0.0599)
-0.4092 (0.6449)


1.0057*** (0.3858) 1.1687 (1.5194)


2-a
0. 1200 (0.0921) 0.0209
(0.0153) 0.0052 (0.0040)
-0.0226 (0.1228)


-0.0004 (0.0005)
-0.7609** (0.3875) 6.8812* (3.7461)
-0.0622*** (0.0164)


-1.1758 (-2.3544)


2-b
0.0574 (0.0785)
0.0295*** (0.0107) 0.0011
(0.0012)
-0.1136 (0.0802)


-0.1749 (0.2375) 3.9170 (2.4272)
-0.0314*** (0.0090)
1.0921***
(0.4009)
-0.0377 (1.8343)


3-a
0.0694 (0.0782) 0.0090 (0.0204) 0.0064 (0.0047) 0.1618 (0.1369)
-1.2322 (1.3096)
-0.0005
(0.0005)
-1.0453** (0.4806) 11.2264** (4.5388)
-0.1018"** (0.0226)


-1.3204 (2.4902)


3-b
0.0307 (0.0748)
0.0279** (0.0117)
0.0012 (0.0013)
-0.092 1 (0.0835)
-0.7918 (0.6490)


-0.2662 (0.2387)
4.5331* (2.7504)
-0.0379***
(0.0106)
1.2756*** (0.4581) 1.0469 (1.8110)


Number of states 48 Number of obs. 48 Prob > X 2 0.7662 Pseudo R2 0.0399
* 10% level of significance
** 5% level of significance
1% level of significance


49 98
0.0777
0.0925


49 49
0.0120 0.2743


49 98
0.0003
0.1647


0.73014 4008 01798
Robust standard errors in parentheses.


47 47
0.0061


47
94
0.0008










Table 2-6. Local recycycling requirement Weibull proportional hazard regressions

Independent 1 2 3 Variables
AGE65 -0.0024 -0 A4 5n n-a1


METRO POP POP DENSITY INCOME WASTE REP CONTROL


AUDUBON

LCV


Constant


/In p

Number of states Number of adoptions Number of obs. Prob >X2
* 10% level of significance
** 5% level of significance
* 1% level of significance


(0.0745) 0.0258* (0.0156)
0.0002** (0.0001)
-0.0368 (0.1022)
-1.1914 (1.0175)


(0.0914) 0.0403* (0.0222)
-0.0009
(0.0022)
-0.1894 (0.1363)



-0.5085 (0.4264) 1.2071 (3.2458) 0.0327* (0.0168)
-2.2665 (2.2512)

-0.057 (0.1444) 46
17
394
0.1249


(0.0820) 0.0335 (0.0239)
-0.0022 (0.0020)
-0.0749 (0.1556)
-1.2014 (1.1562)
-0.7972*
(0.4322)
-1.9413 (4.1530)
0.0442** (0.0185)
-2.8530 (2.2992)


0.0588 (0.1579) 44
17
370
0.0184


-2.4755 (1.8772)

-0.0807 (0.1388) 46
19
377
0













CHAPTER 3
DETERMINANTS OF LANDFILL TIPPING FEES Introduction

Landfills continue to be the most widely used method of managing solid waste in the United States and are necessary in most any type of solid waste management system. The quantity of waste generated continues to increase while existing capacity of existing landfills continues to decrease. Siting new landfills is becoming more difficult because of increased environmental concerns from citizens and government related to location and operation of landfills. In addition, a variety of federal and state requirements for municipal solid waste landfills have been implemented in the past decade.

Tipping fees, the prices set by landfills to accept solid waste, have increased

steadily over the past decade. However, tipping fees are not the same across and within states. This analysis examines the determinants of tipping fees at municipal solid waste landfills covering years from 1992 to 1999 using appropriate controls for regulatory policies, competition, and cost factors. To the best of the my knowledge, there is no existing published, empirical research examining the determinants of tipping fees through econometric analysis. Understanding the factors influencing waste disposal costs are important to policy makers implementing integrated solid waste management plans.

Existing research on the theory of optimal pricing of municipal solid waste

disposal in landfills is limited. Ready and Ready (1995) develop a theoretical model for pricing a depleteable asset that can be replaced at some cost. For example, the landfill space essentially remains fixed but is depleted as more municipal solid waste is tipped in








the landfill. Space in the landfill is replaced with expansion of the landfill or construction of a new landfill. The authors find a component of the optimal fee grows at a real interest rate as space in the landfill is depleted and drops when a new landfill is built. So the shadow value of the resource increases as it becomes more scarce in the authors' model.

Also, limited survey and empirical research exists in examining tipping fees and related costs. Sheets and Repa (1990 landfill tipping fee survey, National Solid Wastes Management Association) surveyed over 200 municipal solid waste landfills in 1990 for information on tipping fees, size, intake amounts, and other characteristics. Based on prior monitoring of landfills in relation to the 1990 survey, the authors note the tipping fees have increased in different rates between various regions in the United States. The northeastern states, whose landfills have the least average remaining capacity, tend to have the highest tipping fees. Furthermore, Sheets and Repa suggest environmental standards on landfills are possibly raising the average tipping fee prices. Clayton and Huie (1973) use synthesized cost data to estimate annual total cost functions for various sizes of landfills using 1970 data, but the effects of individual factors on cost are not determined.

This chapter examines the effects of various forces broadly classified as operational, regulatory, and competitive factors on tipping fees set by solid waste landfills. I identify the effects of several cost factors and find there may be some benefits from locating moderate-sized landfills in drier areas away from coastal counties to the interior of states. Effects from operation costs and regulation factors appear to override competition effects in terms of magnitude.








Landfill Operations and Costs under Regulation

Typical costs involved in landfill operations include pre-development,

construction, operating, and closure costs. Major pre-development costs include site selection and design. Groundwater sources must be protected based on understanding the surface and subsurface geology of the area. Areas with higher precipitation have greater groundwater monitoring concerns. Locating landfills in remote areas usually results in lower prices paid for land space, but transportation costs are usually higher in the construction process. Construction costs include such items as excavation, leachate collection and treatment, groundwater monitoring, drainage controls for surface water, fencing, structures, and scales. The various costs related to each type of construction cost will vary depending on the geography and climate of the landfill. Operating costs include such items as labor, equipment and maintenance, administrative costs, and fuel. Closing a landfill requires one-time closure costs plus post-closure costs that can last over two decades. Gas control systems and cover are the major one-time closure costs; but the landfills need to provide leachate control, groundwater monitoring, land surface care, and other items far into the future.'

The Environmental Protection Agency was instructed to implement a set of minimum criteria for solid waste management facilities as part of the Hazardous and Solid Waste Amendments of 1984. After much consideration, the Agency put forth these criteria in October 1991. The criteria cover location restrictions, operations, design, groundwater monitoring and corrective action, closure and post-closure care, and financial assurance. Compliance from existing landfills was required by no later than


1 Kreith (1994) provides more in-depth discussionon landfill operations.








April 1993 as required by state permitting programs. The large increases in national average tipping fees in the 1990s as compared to the 1980s are said to be due to the new federal regulations.

Methodology and Estimation

This paper analyzes the effects of three broad categories on the tipping fees set by landfills: landfill operation costs, regulatory controls, and competition. A variety of variables and proxies are used to control for the effects of the various categories on tipping fees using yearly data on U.S. landfills from 1992 to 1999 from Chartwell Information. As can be seen from the Figure 3-1, tipping fees vary greatly across the United States. Tipping fee values are in real values.

When examining the mean levels of tipping fees in 1992 and 1999, waste disposal fees appear to be highest in the Northeast, West Coast, and upper Midwest. In 1992 the mean for landfill tipping fees was $28.30 per ton. Tipping fees generally increased from 1992 to 1999. The largest increases in mean tipping fee levels seemed to occur in states with the lower tipping fee levels in 1992. Thirty-seven of the forty-eight continental states had higher mean tipping fees in 1999 compared to 1992.

With landfills likely pricing to at least cover long-run average costs, landfill

operation costs are expected to account for a significant portion of tipping fee variation across the U.S. and over time. Regardless of where landfills are located, managers need to account for leachate collection and groundwater monitoring. Leachate refers to the liquids going toward the bottom of the landfill carrying dissolved and suspended contaminates. Precipitation and moisture in the landfill waste contribute to the amount of liquids in the landfill. Once the leachate is collected, it must be shipped to be treated further or additional facilities must be constructed on-site to treat it. Together, leachate








and groundwater monitoring costs are in the construction, operating, and closure costs of operating a landfill.

Since data on the costs of various leachate collection and groundwater monitoring systems are not available, these costs are proxied with weather data. The National Oceanographic and Atmospheric Administration collects weather data from numerous weather stations across the United States. In the county where a particular landfill exists, the mean July temperature (JULY TEMP in degrees Fahrenheit) and mean January temperature (JAN TEMP in degrees Fahrenheit) are calculated as an average for the weather stations in the county and for the years between 1970 to 1999 using the TD3220 report. The county's average annual precipitation (PRECIP in inches) is calculated.2 The averages are calculated over a long time period to capture the long-term effects of climate on long-run costs, not year to year changes in costs due to random weather. So tipping fees reflect the long-run costs of leachate collection. It is important to explain the effects of differences in long-run weather between different locations on tipping fees. Landfills leachate costs are based on long-run weather patterns, not today's weather.

Leachate collection costs are expected to be greatest in warmer areas with high precipitation, as contaminants can more easily migrate to the landfill bottoms. Hence, tipping fees set by the landfills should be higher, reflecting these increased costs. On the other hand, extreme cold weather makes daily landfill operations more difficult with extra machinery precautions and maintenance, labor force productivity losses, and higher costs to cover material. Decreases in winter temperatures are expected to increase the levels of tipping fees to cover such costs. Squared values for temperature (JULY TEMP

2 Weather data were not available for a small number of counties, and those landfills are not included in the analysis.








SQ, JAN TEMP SQ) and temperature-precipitation interactions (JULY TEMP*PRECIP, JAN TEMP*PRECIP) are also calculated to capture non-linear effects of temperature. As the temperature gets warmer during the winter, costs should decrease but will increase once again when the soil becomes warmer, facilitating increased leachate filtration. Increasing temperature in the hot part of the year should increase leachate costs with the sludge carrying more particulates. But additional increases may result in extreme evaporation, resulting in dryer top soil thus hindering topwater from entering bottom layers in landfills. The turning points of the curves are unknown beforehand. The mean July temperature is just over 76 degrees, while the mean January temperature is just above freezing. Table 3-1 shows the summary statistics for the temperature and other variables.

Pre-development costs are proxied with land prices, and day-to-day operating costs are proxied with retail wages per worker. The U.S. Agricultural Census provides data on the average price of an acre of agricultural land (LAND PRICE in dollars) for counties in 1992 and 1997, which are converted to real values. Other years are interpolated and extrapolated to obtain data for years 1992 to 1999. Many of the landfills are located in rural areas outside of metropolitan areas. This price of land is essentially a shadow price for the cost of landfill expansion. Higher land prices are expected to result in higher tipping fees.

Annual retail wages per worker data (RETAIL WAGES in thousands) are collected from the U.S. Bureau of Economic Analysis CA05 and CA25 reports on a yearly basis for each county. Hired personnel at a typical landfill might include feecollectors, scalemen, foremen, machine operators and drivers, laborers, bookkeepers, and








secreterial support. Depending on the size of the landfill, various tasks may be performed by one individual, such as one person providing bookkeeping and secreterial service. Rather than using a separate wage measure for each position in each county, I use RETAIL WAGES as a general measure of local, limited-skill labor costs. More specific measures such as construction wages might have seasonal variation and may be based on fewer workers and thus less reliable. As such a proxy, increases in retail wages per worker should result in higher tipping fees set by the landfills, but the effects are expected to be small (since leachate and groundwater monitoring costs are far greater to the landfill than payroll costs).

Economies of scale in the landfill industry are investigated using data on the average daily intake for a landfill in each year from 1992 to 1999. This intake figure (AVG INTAKE in tons) and its square (AVG INTAKE SQ in tons) are included to find a minimum efficient sale of operation. Clayton and Huie (1973) found declining long run average cost curves using the synthesized data for landfills sizes 25 tons to 1700 tons of daily intake in Indiana. Table 3-1 shows the mean daily intake of 464 tons for the entire sample. The low, yearly-mean value was in 1995 at 420 tons, and the high, yearly-mean value was 527 tons in 1999.

Regulations costs in addition to the minimum criteria put forth by the EPA impose various burdens on landfills. The most direct policy tool at the state level to reduce waste being deposited in landfills is to implement landfill material bans. A growing concern over landfill expansion has led state policy makers to implement legislation to ban specific materials from being tipped in landfills. The most common bans include motor oil, vehicle batteries, white goods, tires, and yard waste. Oil and








batteries contain materials which can easily contaminate groundwater sources in areas close to water sources and residential populations. White goods, such as stoves and refrigerators, are bulky and take up large amounts of space in the landfills. Also, many contain electrical components containing polychlorinated biphenyls, which can contribute to toxic sludge material in the landfill. Tires also take up a significant amount of landfill space and create uneven settling in the landfill. In fact, tires can rise in the landfill after closure and break covers. Additionally, large piles of tires create ideal habitat for mosquitoes and rodents. Finally, tires fires create harmful fumes that are difficult to extinguish compared to fires on organic materials such as wood products.

Keeping particular items out of landfills may actually decrease the costs of

landfill operation. However, workers at the landfills need to monitor waste to prevent violations of landfill material bans. The sign of the net effect from the material ban depends on the magnitudes of the cost saving effect and the monitoring effect. Banning oil and batteries3 should result in the greatest savings in regards to leachate collection and treatment. In addition, banning tires is expected to result in lower costs associated with mosquito and rodent control, fires, and uneven settling in landfills. Yard waste contributes to excess gas and leachate generation compared to other materials, but composted yard waste is commonly used as an intermediate landfill layer. White goods are the easiest to monitor in incoming loads to the landfill.

State landfill material bans on automotive batteries, motor oil, tires, white goods, and yard waste are captured by the following indicator variables: BATTERY BAN, OIL BAN, TIRE BAN, WHITE GOODS BAN, and YARD WASTE BAN. Battery Ins


3Automotive batteries remain the largest source of lead in landfills.








affected the greatest number of landfills over the time period. Conversely, white good bans affected the least number of landfills. The variable TOTAL BANS is a summation of the five material bans and captures the cummulative effect of the various landfill bans. In addition, landfills located on the border of a particular state also are affected by landfill material bans of surrounding states. Municipal solid waste markets extend over state borders, and state policies affect market behavior. I calculate the difference between the average number of landfill bans of the states bordering the particular county and the number of landfill bans facing a landfill located in a state-border county (BAN DIFF). So a higher value for BAN DIFF indicates landfills in border counties face greater landfill bans from bordering states than their home states. Counties not on the border have a value of zero for BAN DIFF.

Landfills must meet EPA guidelines when locating in wetland and floodplain

areas, but the landfills are also likely to encounter landfill specific regulations and other regulations at the local level near open water. Some counties have implemented 'setbacks' to keep industrial activity from water sources. Many landfills face greater operating restrictions in coastal counties that are likely to increase costs. Also, hydrogeologic factors in coastal areas are likely to increase the costs of collecting and treating leachate compared to arid areas with a more predictable hydrology system and favorable soil characteristics. It is expected that landfills in coastal counties, as indicated by the variable COASTAL, will have higher tipping fees to reflect to these increased costs. It can be seen in Table 3-1 that almost one-fifth of the landfills are located in coastal counties.








In addition to operating and regulatory costs, competition is likely to affect

tipping fees. Both private and municipal landfills exist across the United States. In the report by Sheets and Repa (1990 landfill tipping fee survey, National Solid Wastes Management Association), the authors report 76%, 11%, and 13% of respondents had private ownership/private operator, public ownership/private operator, and public ownership/public operator structures, respectively. Anecdotal evidence suggests publicly owned landfills set their prices to break even, or to price at average total cost. Privately owned landfills are profit maximizers. When there is a mix of both private and firms serving in the same market, measures of competition are likely to be understated. Data on ownership are not available to differentiate between public and private landfills.

As a method to examine effects of competitive pressures on tipping fees, I

calculate three different measures: 50 MILES (small-size market definition), 100 MILES (medium-size market), and 150 MILES (large-size market). These measures contain the number of competitors in each market. For example, 50 MILES measures the number of competitors within a 50 mile radius for a given landfill. Likewise, this measure is constructed for larger areas of radii 100 and 150 miles. It is expected that increasing the number of competitors in the market would result in lower tipping fees. However, such a result is expected to be understated with the presence of publicly owned landfills.

Results

Regressions using the same types of specifications are presented in three different tables: 3-2, 3-3, and 3-4. In each of the three tables, a different measure of market competition is used (50 MILES, 100 MILES, and 150 MILES in Tables 3-2, 3-3 and 3-4, respectively). Also, regressions 1-a, 2-a, and 3-a differ from regressions 1-b, 2-b, and 3b, where the first set of regressions includes TOTAL BANS, while the latter set of








regressions includes the individual landfill material bans. Regressions 1-a and I-b do not include any weather variables. Precipitation and temperature variables (and square values) are added to this base specification in the middle columns, 2-a and 2-b. Finally, columns 3-a and 3-b add to the base specification by including precipitation, temperature, and temperature-precipitation interactions. In general, the results from the regressions support the hypotheses. The regressions include 1570 landfills and 10793 observations.

The regression results from the temperature and precipitation estimates are mixed. JULY TEMP is only statistically significant in Specification 3-b in Table 3-b. In addition, JULY TEMP SQ is not statistically significant in any tables. However, the results for JAN TEMP and JAN TEMP SQ are more promising. As predicted in specifications 2-a and 2-b in the three tables, JAN TEMP is negative and significant while JAN TEMP SQ is positive and significant. This suggests the graph of tipping fee in terms of mean January temperature is a U-shaped quadratic curve. An increase in the winter temperature measure results in a decrease of tipping fees at low temperatures and an increase in tipping fees at higher temperatures. The minimum values (turning points) range from 41 to 45 degrees Fahrenheit across the tables. As a reference, some metropolitan areas within this range include Wichita Falls (TX), Greenville-Spartanburg

(NC), Seattle-Tacoma (WA), Birmingham (AL), Atlanta (GA), and Dallas (TX).

Declines in extra machinery precautions and maintenance and an increase in the ease of obtaining landfill cover occur as mean January temperatures increase to the low 40s. Once the mean temperatures start to increase into the 50s, extra precautions are no longer necessary and soil temperatures increase facilitating the ease in separating proper cover material for the landfill. Also, as the temperature increases throughout the layers in








the landfill, moisture tends to filter more easily through the landfill thus increasing leachate collection costs.

PRECIP is positive and significant in all three tables, indicating an increase in annual precipitation increases the tipping fees. The variable appears alone in specifications 2-a and 2-b and is interacted with JULY TEMP and JAN TEMP in specifications 3-a and 3-b. In the first two specifications across the three tables, an increase in an annual precipitation by an inch increases the tipping fee by $. 16 to $.22. Furthermore, a one-standard deviation increase in PRECIP increases TIP FEE by $2.27 to $3.12. Interacting PRECIP with temperature variables results in negative, statistically significant coefficients for JULY TEMP*PRECIP in specifications 3-a and 3-b. It is expected leachate collection would be highest in regions of the country with warm and wet climates. The negative sign is an unexpected result.

As expected, the coefficient of LAND PRICE is positive and significant in each specification within each of the three regression tables. This evidence suggests land prices may be a constraint on landfill expansion. Although the agricultural land values have a positive effect on tipping fees, the magnitudes are small. In Column 1-a in Table 3-2, a one-standard deviation change in the average price of an acre of agricultural land in the county of a given landfill results in an increase in the tipping fee by 51 cents.

The estimated coefficients for RETAIL WAGES are positive and significant in all columns in Tables 3-2, 3-3, and 3-4. Lowest values for the coefficients seem to appear in Specification 1-a while the highest values are in Specification 2-b. In Column 3-b in Table 3-2, a $1000 increase in the retail wagers per worker results in an increase in TIP FEE by 81 cents. A one standard-deviation increase in RETAIL WAGES results in a








$2.08 increase in the tipping fee. So an increase in general labor costs is reflected in higher tipping fees on average.

AVG INTAKE is negative and significant in all three tables. Furthermore, AVG INTAKE SQ is positive in all tables and statistically significant in many specifications. If the landfills are setting tipping fees in relation to average total cost, the curve produced by the average daily intake variables is a proxy to an average total cost curve. Minimum values of this curve range from 3550 to 3966 tons, 3519 to 4075 tons, and 3423 to 3947 tons in Tables 3-2, 3-3, and 3-4 respectively. The mean value for AVG INTAKE is 464 tons.

The 25th, 50th, and 75th percentiles are 61, 185, and 500 tons respectively.

About 95% The tipping fee would decrease anywhere from $.92 to $1.35 going from the 25th to 75th percentile. Moving from the 25th percentile to the minimum values of the average intake curves would result in a tipping fee decrease of anywhere from $4.83 to $7.17. These figures indicate a large percentage of landfills could continue to increase intake and capture some scale economies but costs seem to rise eventually. It is possible landfills encounter new operation costs to deal with risk factors beyond a minimum efficient scale. For example, the accumulation of municipal solid waste eventually creates odor problems which were largely absent at smaller levels of operation. Landfills may need to take additional measures beyond what is required by regulation to manage odor and disease vectors.

Results from including the variables for landfill material bans are mixed. The

estimated coefficient for BATTERY BAN is positive and largely significant. Having this ban results in a $2.48 to $3.33 increase in tipping fees where significant. If the ban on








battery bans resulted in a savings from avoided costs due to handling leachate, tipping fees would decrease. Monitoring waste inflows to comply with bans is costly to the landfills and increases costs. As a result, tipping fees would increase. The estimated coefficients for BATTERY BAN suggests monitoring costs by the landfill could be high.

The estimated coefficients for motor oil are negative in all tables but only significant in Specification 1-b in Table 3-2. Like automobile batteries, motor oil contributes to the toxicity of leachate collection and thus increases processing costs. In addition, oil contributes to greater fire hazard risks. The negative coefficient indicates landfills realize cost savings with the bans. The estimated coefficient implies that tipping fees decrease by $1.59 with the OIL BAN.

Some of the strongest results from the material bans come from the coefficients on TIRE BAN. The estimated coefficients range from a low of -5.52 to -3.05 and are all statistically significant. Avoiding tire disposal can save the landfill expenses from mosquito control to fire control to closure issues. Tires are one of the more common materials to be banned from landfills. From a public policy perspective, it appears tire bans are serving a useful role.4

The estimated coefficients for WHITE GOODS BAN are positive and largely

statistically significant. Having the ban in place for white goods raises tipping fees from $2.72 to $3.51 where statistically significant. It does not appear monitoring costs for white goods would be high compared to other more compact materials. Some landfills have started scrapping processes at their locations to deliver steel and other metals to recycling centers for money. Facing a ban on white goods could result in lost revenue

4 Tire reuse and recycling is increasing as whole tires are being used in playground equipment, reef construction, and chopped for use in rubber mats, molded objects, and rubberized paving materials.








from some landfills. So landfills may be increasing the price of one 'output' (accepting solid waste) while decreasing the 'output' of another product (scrapping white goods for sale of metals to other companies).

Finally, the magnitudes of the estimated coefficients on YARD WASTE BAN are the highest. Values range from 1.88 to 6.96, with all estimated coefficients being statistically significant. Large amounts of organic matter can significantly increase leachate collection costs. Hence, avoiding this disposal can be beneficial but yard waste is one of the most difficult wastes to monitor when it is banned. Also, some landfills have implemented composting stations at their sites to process organic materials for other uses. Like the case of white goods, not having bulk shipments of yard waste could result in lost revenue if the landfill has the aforementioned facilities. The positive effects from the estimated coefficients suggest avoided leachate problems may be small.

Besides examining the effect of each ban individually, I examine the collective effect of landfill material bans and the effect of bans from neighboring states on landfills located in border counties. TOTAL BANS is included in specifications 1-a, 2-a, and 3-a in each of the three tables. The net effect suggests the material bans result in an increase in the tipping fee. Values for the coefficients on TOTAL BANS range from .87 to .92 where significant. So, a one standard-deviation increase in the number of landfill material bans results in an increase of TIP FEE anywhere from $1.15 to $2.80. The measure of the effect of bans from neighboring states, BAN DIFF, was not statistically significant in any of the regressions.

In addition to the landfill bans, landfills may face more local regulation in coastal areas. Landfills located in coastal counties adjacent to the Pacific Ocean, Gulf of








Mexico, Atlantic Ocean, and Great Lakes have a value of one for the COASTAL indicator. Estimated values for COASTAL range from 6.09 to 9.84. So being located in a coastal county may result in up to $10 higher tipping fees. With one-fifth of the landfills in the coastal counties, the aggregate financial effects of locating and operating landfills in coastal areas are substantial.

The effects of competition are analyzed in each of the regression tables.

Variables 50 MILES, 100 MILES, and 150 MILES are calculated for each landfill. The total number of competitor landfills within different radii are counted for each landfill. Results from the regressions indicate the estimated coefficients for each measure are positive. Only four of the eighteen coefficients are not statistically significant. In terms of magnitude, the largest effects seem to be with the specifications using 50 MILES. Column 1-a in Table 3-2 indicates adding one more competitor within a 50 mile radius results in a $.37 increase in TIP FEE. The results from these estimated coefficients are contrary to the hypothesis. More competition should result in lower prices. Also, it would be expected that competition would have more of an effect on privately owned landfills than publicly owned landfills. However, even though the publicly owned landfills may price at average costs, they do face some pressure to reduce costs and price lower prices as haulers could choose other disposal options.

Finally, year effects are included in the regressions to see if there was a

considerable increase in later years in the sample period compared due to effects of federal regulations. When looking at the individual coefficients from the year coefficients, there is no evidence of an increase in general tipping fees in the early 1990s when the EPA impleneted its minimum criteria guidelines for landfills.








Conclusion

This paper uses a national dataset on landfills containing yearly data on tipping fees and intake volumes from 1992 to 1999 supplemented with other data to examine the effects of various factors on tipping fees. I haven't found an existing empirical study on determinants of tipping fees. Several important findings with policy implications are found in the analysis.

In terms of cost and scale, there are significant economies of scale in operating

landfills. Only five percent of the landfills could be too big. Precipitation has significant effects on increasing tipping fees. Also, increases in land prices and labor costs increase tipping fees.

Regulatory controls in the form of landfill material bans and local restrictions affect tipping fee levels. Landfill bans may save landfills money by reducing leachate, groundwater monitoring, vermin, mosquito, and other treatment costs. However, monitoring the bans is costly. Or a ban could result in lost revenue for a landfill which uses incoming municipal solid waste as an input to produce and sell outputs. In terms of the size of the landfill ban effects, yard waste bans significantly increase tipping fees relative to other material bans. When the net effects of the bans are considered together, banning another good raises tipping fee levels. But local regulation costs in coastal areas of the U.S. may dwarf the effects of the landfill material bans. Holding the price of land constant, operating landfills in coastal counties results in a tipping fee differential nearly one-third of the mean in the sample.

If policy makers are concerned about the rising costs of landfill operations and associated tipping fees, more government intervention could result in better solid waste management policy. Landfills in coastal areas with high precipitation result in high








tipping fee regions. Instead of operating mega-landfills in these regions, it would likely be better to operate landfills on a smaller scale toward interior counties of states. In addition, land costs and labor costs are likely to be lower, resulting in lower tipping fees. However, locating landfills in the most rural areas may result not enable the landfills to take full advantage of economies of scale due to the high transportation costs associated with moving waste from metropolitan areas to rural areas. Either haulers will be deterred from shipping waste to these remote landfills, or integrated waste collection and landfill companies will experience higher priced waste disposal as they absorb these costs. In the latter case, increased shipping costs would result in a smaller optimal size of landfill.5

Future analysis could be extended in a variety of directions. First, the effects of ownership are likely to have different effects on tipping fees with private firms facing more competitive pressures. Identifying ownership and operating structures is a first step in this direction. Second, the overall fit of the regressions could be improved by including more landfill-specific covariates such as landfill size and the amount of capital equipment at the sites. Finally, it would be useful to have information on vertical integration within the solid waste management industry. Landfills operating under an umbrella operation with a hauling division and recycling division may experience economies of scope and/or cost shifting. Unfortunately, attempts to acquire these data previously discussed were unsuccessful. A deeper analysis would need a richer set of data to investigate these questions.





5 See Kenny (1982) for an investigation of optimal plant size considering the size of geographic areas served and input costs.



















1992 Tipping Fees
(means)
* over $60 (4)
* over $40 to $60 (8) D over $2O to $40 (18) ] $0 to $20 (18)


* over $60 (1)
*-1 over $40 to $60 (12) l over$20to40 (31) LII $0 to $20 (4)
Figure 3-1. Mean tipping fees by state. A) In 1992. B) In 1999.





Minimum Maximum


Table 3-1. Summary statistics for tipping fee regressions Variables Mean Standard Deviation
TIP FEE 31.12 18.03 AVG INTAKE 463.60 881.67 AVG INTAKE SQ 992201.40 6808753.00 LAND PRICE 2001.75 4104.68 RETAIL WAGES 14.34 2.57 TOTAL BANS 3.04 1.32 BATTERY BAN 0.93 0.26 OIL BAN 0.58 0.49 TIRE BAN 0.76 0.43 WHITE GOODS BAN 0.31 0.46 YARD WASTE BAN 0.47 0.50 BAN DIFF -0.01 0.97 COASTAL 0.18 0.38 JULY TEMP 76.39 5.48 JULY TEMP SQ 5865.30 844.00 JULY TEMP*PRECIP 2591.26 1150.39 JAN TEMP 33.82 12.69 JAN TEMP SQ 1304.70 910.58 JAN TEMP*PRECIP 1159.39 743.85 PRECIP 33.73 14.20 50 MILES 8.04 5.32 100 MILES 25.79 12.61 150 MILES 52.02 21.93


0.48 0.12
0.0144 74.31 5.63
0 0
0 0 0 0 -5
0
61.34 3762.95 247.42
4.3 18.49 78.86 2.71
0 0 0


261.32 15000
2.25e+08 162954.80
31.17
5 1 1 1 1 1 5 1
93.70 8779.69
5342.22 71.40 5097.96
4033.42 66.42
31 68 119






51


Table 3-2. Tipping fee regressions using small-market definition Independent 1-a 1-b 2-a 2-b 3-a 3-b Variables


AVG INTAKE -0.0024** -0.0033***


AVG INTAKE SQ LAND PRICE RETAIL WAGES TOTAL BANS BAN DIFF BATTERY BAN OIL BAN TIRE BAN WHITE GOODS YARD WASTE BAN


COASTAL 50 MILES JULY TEMP


(0.0010)
3.03e-07 (2.14e-07)
1.24e-04*** (4.32e-05) 0.3696** (0.1657)
0.9189***
(0.3288)
-0.112 (0.4484)


9.0533*** (1.1924)
0.3716***
(0.0984)


(0.0009)
4.19e-07"*
(2.07e-07)
8.60e-05** (4.08e-05) 0.7752***
(0.1779)


-0.2263 (0.4412)
3.3269**
(1.5374)
-1.5940* (0.8471)
-5.3006*** (0.9030)
2.7190***
(0.9218)
6.9614*** (0.8098)
7.9465*** (1.1691)
0.2962*** (0.0939)


JULY TEMP SQ JULY TEMP*PRECIP JAN TEMP JAN TEMP SQ JAN TEMP*PRECIP


PRECIP

Constant


0.2201***
(0.0321)
16.6354*** 11.9953*** -54.1589 (2.2103) (2.5257) (54.5424)


Adjusted R-squared 0.09 0.13 0.18 0.19 0.20 0.21


I 0-/o level o signiiicance
** 5% level of significance
1% level of significance


Robust standard errors in parentheses. All specifications include year effects.


-0.0025*** (0.0010)
3.50e-07 (2.Ole-07)
9.12e-04"* (-3.96e-05) 1.0989"**
(0.2005)
-0.0249
(0.3422)
-0.5648 (0.4166)












6.4454*** (1.3022)
0.2376***
(0.0924)
-0.3097
(1.4082)
-0.0016
(0.0090)


-0.9393*** (0.1690)
0.0106***
(0.0023)


-0.0028***
(0.0008)
3.78e-07* (1.96e-07) 6.77e-05* (3.60e-05) 0.8077***
(0.1842)


-0.0029***
(0.0009)
3.91e-07* (2.Ole-07)
7.94e-05** (-3.95e-05)
1.1111***
(0.1980)


-0.5379
(0.4164)
2.7035*** (1.5483)
-1.2856 (0.7989)
-3.0504** (0.9829) 1.1717
(0.9556)
2.4846** (1.0762)
6.0933*** (1.2769)
0.2280** (0.0922)
-0.8468 (1.3434)
-0.0021
(0.0086)


-0.7883***
(0.1786)
0.0092*** (0.0023)


0.1873***
(0.0361) 70.0773 (52.0909)


-0.0024*** (0.0008)
3.38e-07* (1.93e-07)
7.62e-05** (3.63e-05) 0.8652*** (0.1883) 0.5279 (0.3265)
-0.2108
(0.3942)












7.5829*** (1.2470) 0.1210
(0.0915) 0.2373 (0.1781)


-0.0350***
(0.0086) 0.0452 (0.0967)


-0.0049 (0.0035)
3.1314**
(0.5905)
-16.2527 (12.4424)


-0.2290
(0.3955) 0.6477 (1.4226)
-0.2751
(0.7538)
-3.7873***
(1.0059)
3.3750***
(0.9190)
2.2825**
(1.0320)
6.9093*** (1.1870) 0.1270
(0.0900) 0.2402
(0.1722)


-0.0322***
(0.0081) 0.1702*
(0.1005)


-0.0076**
(0.0035)
3.0209*** (0.5614)
-18.2983 (12.0172)










Table 3-3. Tipping fee regressions using medium-market definition Independent 1-a 1-b 2-a 2-b 3-a 3-b Variables


AVGINTAKE -0.0026*** -0.0033***


AVG INTAKE SQ LAND PRICE RETAIL WAGES TOTAL BANS BAN DIFF BATTERY BAN OIL BAN TIRE BAN WHITE GOODS YARD WASTE BAN COASTAL 100 MILES JULY TEMP JULY TEMP SQ JULY TEMP*PRECIP JAN TEMP JAN TEMP SQ JAN TEMP*PRECIP PRECIP Constant


(0.0009) 3.19e-07 (2.07e-07
1.17e-04** (3.99e-05)
0.2951* (0.1551) .9235*** (0.3188)
-0.0256 (0.4358)


9.1263"** (1.1741)
0.2650*** (0.0340)


(0.0009)
4.15e-07"* (2.04e-07)
8.7le-05** (3.87e-05)
0.6480*** (0.1703)


-0.1480 (0.4335)
3.0967** (1.4432)
-0.9713 (0.8415)
-5.4030***
(0.8604)
3.1501**
(0.9095)
5.9208*** (0.8025)
8.0545***
(1.1555)
0.2104*** (0.0348)


-0.0026*** (0.0009) 3.5 le-07 (2.00e-07)
9.04e-05** (3.82e-05)
0.9770*** (0.1919)
-0.064 (0.3316)
-0.5277
(0.4078)












6.4681*** (1.2905)
0.1847***
(0.0346)
-1.2605
(1.4175) 0.0048 (0.0090)


-1.0248***
(0.1676)
0.0119*** (0.0023)


0.1944*** (0.0321) 13.8069*** 11.0431*** 89.8201 (2.2910) (2.4940) (54.8706)


-0.0029***
(0.0009)
3.87e-07* (2.00e)-07
8.09e-05** (3.8 1e-05) 0.9857*** (0.1906)


-0.5044 (0.4087)
2.4945* (1.4683)
-0.8971 (0.8044)
-3.2470***
(0.9748) 1.3437
(0.9422) 1.8846* (1.0576)
6.1788*** (1.2650)
0.1772*** (0.0354)
-1.7005
(1.3686)
0.008 (0.0087)


-.8775*** (0.1773)
0.0104*** (0.0023)


0.1739*** (0.0359) 102.0465 (53.0049)


-0.0024*** (0.0008)
3.41e-07*
(1.92e-07)
7.65e-05** (3.56e-05) 0.8390*** (0.1829) 0.5764 (0.3256)
-0.1771
(0.3926)












7.6553*** (1.2338)
0.0920** (0.0397) 0.2861 (0.1778)


-0.0361"**
(0.0084)
-0.0139
(0.1048)


-0.0028
(0.0037)
3.1124*** (0.5865)
-18.1664 (12.3335)


-0.0027*** (0.0008)
3.76e-07* (1.94e-07) 6.90e-05* (3.53e-05
0.7943*** (0.1805)


-0.2041 (0.3946) 0.8306 (1.3847)
-0.1751
(0.7629)
-3.7000***
(1.0078)
3.4448***
(0.9137)
2.0376** (1.0170)
7.0414"**
(1.1744)
0.0830** (0.0389) 0.2862*
(0.1735)


-0.0332*** (0.0081) 0.1139
(0.1081)


-0.0058
(0.0037)
3.0144*** (0.5603)
-20.3518* (11.9856)


Adjusted R-squared
* 10% level of significance
** 5% level of significance
*** 1% level of significance


Robust standard errors in parentheses. All specifications include year effects.










Table 3-4. Tipping fee regressions using large-market definition Independent 1-a 1-b 2-a 2-b 3-a 3-b Variables


AVGINTAKE -0.0026*** -0.0033***


AVG INTAKE SQ LAND PRICE RETAIL WAGES TOTAL BANS BAN DIFF BATTERY BAN OIL BAN TIRE BAN WHITE GOODS YARD WASTE BAN COASTAL

150 MILES JULY TEMP JULY TEMP SQ JULY TEMP*PRECIP JAN TEMP JAN TEMP SQ JAN TEMP*PRECIP PRECIP Constant Adjusted R-squared
* 10% level of significance
** 5% level of significance
1% level of significance


(0.0009)
3.32e-07* (2.00e-07)
1.14-e04*** (3.86e-05) 0.4319*** (0.1543)
0.8699*** (0.3187)
-0.0894 (0.4414)


9.8429*** (1.1645)
0.1546*** (0.0186)


(0.0009)
4.18e-07"* (1.99e-07)
8.79e-05**
(3.79e-05) 0.7212*** (0.1689)


-0.1886 (0.4375)
3.0803** (1.4423)
-0.9211
(0.8485)
-5.5228*** (0.8871)
3.5131*** (0.9262)
5.3389*** (0.8319)
8.6381***
(1.1474)
0.1229*** (0.0202)


-0.0026***
(0.0009) 3.54e-07 (1.99e-07)
9.08e-05**
(3.78e-05) 1.0347*** (0.1906)
-0.2047 (0.3275)
-0.6166 (0.4107)












6.8439*** (1.2908)
0. 1135*** (0.0208)
-1.2465 (1.4063)
0.0048 (0.0090)


-1.1434"**
(0.1707)
0.0138*** (0.0023)


0.1772*** (0.0331) 10.6647*** 9.1416*** 89.3160
(2.4458) (2.5420) (54.5133)


Robust standard errors in parentheses. All specifications include year effects.


-0.0029*** (0.0009)
3.88e-07* (1.98e-07)
8.20e-05** (3.78e-05) 1.0376*** (0.1891)


-0.5875
(0.4109) 2.4836*
(1.4686)
-0.9916 (0.8093)
-3.3909*** (0.9780)
1.349
(0.9398) 1.4534 (1.0587)
6.5871*** (1.2663)
0.1091***
(0.0217)
-1.6403 (1.3584) 0.0077 (0.0087)


-0.9971***
(0.1806)
0.0112**
(0.0024)


0.1614*** (0.0362)
99.6971 (52.6796)


-0.0023*** (0.0008)
3.36e-07*
(1.90e-07)
7.78e-05** (3.57e-05) 0.9199*** (0.1835) 0.4784 (0.3260)
-0.2309 (0.3949)












7.7854*** (1.2514) 0.0232
(0.0250) 0.2531 (0.1785)


-0.0351 *** (0.0085)
0.0296 (0.1094)


-0.0043 (0.0039)
3.1147*** (0.5954)
-17.6008 (12.3534)


-0.0027*** (0.0008)
3.73e-07* (1.92e-07)
6.95e-05** (3.54e-05) 0.8768***
(0.1806)


-0.2577 (0.3962) 0.7959
(1.4067)
-0.3507 (0.7732)
-3.7872*** (1.0072)
3.3303***
(0.9217)
2.1513** (1.0115)
7.1134*** (1.1892) 0.0164 (0.0249) 0.2476 (0.1739)


-0.0322*** (0.0081) 0.1643 (0.1123)


-0.0075* (0.0039)
3.0123***
(0.5654)
-19.4956* (11.9873)













CHAPTER 4
DETERMINANTS OF HOUSEHOLD RECYCLING: A MATERIAL-SPECIFIC ANALYSIS OF RECYCLING PROGRAM FEATURES AND UNIT PRICING' Introduction

The past 15 years have been a time of dramatic change for solid waste

management. Beginning in the mid-i 980s, with stricter EPA requirements for landfill construction on the horizon, landfill tipping fees increased dramatically and there was a widespread impression that landfill space was growing scarce and that a landfill "crisis" was inevitable.2 Two clear national trends in solid waste management emerged as a result of local efforts to reduce the quantities of waste being landfilled. The most pervasive was the introduction of residential curbside recycling programs. In 1988, there were approximately 1000 such programs in the U.S.; in 1992, there were almost 5000; by 1999 the number reached just over 9000 (Goldstein and Madtes, 2000). A second, less pervasive but still important, trend during this period was the introduction of volumebased pricing, or unit pricing, of solid waste disposal services wherein households are charged for garbage collection according to the number of containers they set out. Prior to the late 1980s there were perhaps a few dozen such programs. By 1992, there were approximately 2000; and by 1999, just over 4000 (Miranda and Aldy, 1998).



1 Adapted from Journal of Environmental Economics and Management, Vol. 45, Jenkins, R.R., Martinez, S.A., Palmer, K., & Podolsky, M.J., "The Determinants of Household Recycling: A Material-Specific Analysis of Recycling Program Features and Unit Pricing," Pages 294-318, Copyright (2003), with permission from Elsevier.
2 Most of the increase in tipping fees occurred during the middle and late 1980s. In 1985 the national average tipping fee in the U.S. was approximately $11.20 per ton; in 1990, it was approximately $33.75. As of 1997, it remained close to $30.00. (All values are in 1997$.) (U.S. EPA, 1997).








Though the nature of a curbside recycling program is quite different from a unit pricing program, both theoretically provide incentives for a redirection of waste quantities from disposal sites to recycling centers. A curbside program reduces a household's cost of recycling by making recycling more convenient and less time consuming. A unit pricing program increases a household's cost of discarding additional waste relative to its cost of recycling (i.e., not recycling leads to higher fees for waste collection services).3

Each program targets different waste management activities, which might lead to differences in the outcomes of the two programs. For example, unlike a curbside recycling program, unit pricing only gives an indirect incentive to recycle while its direct incentive is to reduce waste quantities. Unit pricing may also create incentives for households to adjust their purchasing habits to generate less solid waste. Thus, the two programs might very well have different effects on household recycling effort.

Economic principles also suggest that the two programs will have different

impacts on recycling and consumption of different recyclable materials (Jenkins, 1993). One suggestion is that volume-based unit pricing will give households an incentive to recycle bulky items that take up lots of garbage container space - such as plastic milk jugs. On the other hand, unit pricing might encourage households to avoid generating bulky wastes in the first place. Households might alter the composition of their consumption bundles so that there is less trash to discard.




3 Without unit pricing, most communities finance waste disposal via general tax revenues or flat fees. From the perspective of households, this places a marginal price of zero on waste disposal. This causes them to dispose of more than the socially efficient amount of waste. A unit pricing program imposes a non-zero marginal price on waste disposal that can potentially correct this problem.








A curbside recycling program also might disproportionately affect certain

materials. As a substitute for drop-off recycling, curbside collection mainly reduces a household's costs of transporting recyclable materials. Compared to a household without any local recycling program, a household with a curbside program will have a much easier time recycling materials that are hard to transport, like glass bottles, which are bulky and can break.

Policy makers would benefit from a better understanding of the impact of the two programs and their features on different recyclable materials. To the municipalities that collect them, different recyclable materials have different costs of recycling as well as different values on the open market. Understanding which program features lead to greater recycling of high valued materials could improve the cost-effectiveness of a community's efforts to promote recycling. In other cases, municipalities sometimes achieve very high recycling effort directed at a few materials. In order to increase their aggregate recycling percentage in an effort to meet state-mandated recycling rate targets, municipalities must sometimes encourage households to recycle additional materials. Understanding how best to promote recycling of a broader range of materials would be beneficial. On the other hand, if the costs of adding a particular material to a curbside program exceed the waste diversion and recycling revenue benefits of doing so, then adding certain materials may not be worthwhile.

This study analyzes a large household-level data set representing 20 metropolitan statistical areas (MSAs) across the country to study the impact of these two popular solid waste programs and their features on the percent recycled of five different materials: glass bottles, plastic bottles, aluminum, newspaper, and yard waste. All communities in








the data set offer curbside recycling of at least one of the five materials; although most offer it only for a subset of the five. However, the data set contains detailed information on the attributes of different recycling options for all five specific materials. For example, the data indicate whether each material is collected at all through a local program and if so whether it is collected curbside or at a local drop-off facility. The data also indicate whether recycling the material is mandatory or voluntary and the age of the recycling program. Finally, the data set contains rich household level socioeconomic information. We augment the household-level data with community-level information on the prices charged for disposal under a unit-pricing program where it is applicable.

The contributions of this paper are more easily understood within the context of the literature that has investigated the determinants of recycling. Thus, we start with a brief review of this literature and adopt from it a simple theoretical model. We then describe our own data and present an empirical model. We present the empirical results, note limitations of the data and in closing, discuss the relevance of our findings to policy.

Prior Research and a Conceptual Framework

This paper makes two contributions to the existing economics literature on

recycling. First it adds to the research on the effectiveness of curbside recycling and unit pricing at encouraging households to recycle. Several papers study various aspects of these programs, sometimes with unit pricing and curbside recycling operating together and sometimes with one program operating in isolation (Kinnaman and Fullerton, 2000; Callan and Thomas, 1997; Fullerton and Kinnaman, 1996; Hong et al. 1993; Van Houtven and Morris, 1999). However, ours is the first that analyzes data from most








major U.S. metropolitan areas and rests on a household-level unit of analysis.4 Household-level is preferred to community-level because households are the decisionmaking units that are the target of recycling policies. Analyzing data from numerous MSAs located in different parts of the country is preferred to an analysis of only one region because it facilitates the identification of policies and demographic variables that are significant across regions.

A second contribution of this paper is to extend previous research by investigating whether and how the impact of these two popular programs differs for different recyclable materials. The few existing material-specific studies have lacked the rich amount of information about both recycling and unit pricing programs contained in our data set (Saltzman et al. 1993; Reschovsky and Stone, 1994). We also examine the effect of household socio-economic characteristics on recycling effort directed at different materials.

Table 4-1 summarizes the existing econometric literature that studies the effects of unit pricing and curbside recycling on household recycling effort. A number of papers have developed conceptual frameworks to study the impact of unit pricing (Fullerton and Kinnaman, 1996; Morris and Holthausen, 1994; Jenkins, 1993). Others, including Podolsky and Spiegel (1998) and Kinnaman and Fullerton (1995), describe the substitution possibilities between waste disposal and recycling as part of household waste management. These papers develop models in which households maximize utility



4 Several econometric studies analyze the impacts on recycling effort of one or both of these two popular programs by examining household-level data; in particular, Nestor and Podolsky (1998), Fullerton and Kinnaman (1996) and Hong et al. (1993). However, the data for all three of these studies are for a single region where curbside recycling and unit pricing co-exist. Several other studies are national in scope but rely on community-level data (Kinnaman and Fullerton, 2000; Miranda et al. (1994); U.S. EPA, 1990). (The latter two use the case study method of analysis.)








subject to a budget constraint that incorporates a unit price for waste collection. The models are the basis for solid waste disposal and recycling demand equations.

On the right hand sides of these equations are three categories of exogenous

variables: characteristics of the goods whose consumption generates waste; descriptions of the local waste management system; and socio-economic factors. The first category includes the price of consumption good i (Pi) and the amount of waste generated per unit of good i (P3) where (i = 1 ... n). The second category consists of the price per unit of waste disposal (Pw) and a vector of recycling program features (RP) including whether the collection occurs at the curb or at a drop-off facility, the length of life of a recycling program, and so on.5 The third category is comprised of socio-economic characteristics

(SE) such as household size, income and education.

Specifically, D and R are the optimal levels of household disposal and recycling,

D = f (1h , P., P ,, en, ew, R-P, SE) (4-1) Ri = Y7 (31, ..., P3n, PI..., Pn, Pw, RP, SE) (4-2) R=Z Rj. (4-3)

Each recycled material, j, has unique characteristics that could affect the relationship between recycling and the exogenous variables. These characteristics include factors such as bulkiness that affect the ease of recycling as well as the availability of substitutes for the material. Thus, each material (R) has a unique recycling demand equation as specified in Eq. 4-2.



5 The price per unit of waste disposal charged to households is usually a volume-based price. For example, households in communities employing a bag/sticker purchase official program bags or stickers, which they affix to garbage bags of the mandated size. Alternatively, households in communities using a subscription can program specify a level of waste disposal per period of time in advance and are charged according to this level.








Consistent with Eq. 4-2, we analyze material-specific recycling behavior for each of five materials. However, since we do not have data on recycling quantities, we actually estimate the effects of the exogenous variables on the intensity of recycling for each material.

Data Description

The primary data source is a recycling survey mailed by Equifax, Inc. in 1992 - a year of increasing popularity for unit pricing and soaring popularity for curbside recycling.6 The survey was mailed to 4600 households residing in 20 U.S. metropolitan areas (please see Table 4-2 for a list of the 20). The survey was targeted toward middle and upper income households in these regions. Sixty-five percent of questionnaires, 2984, were returned. Households responded to questions about recycling participation, recycling program characteristics, household characteristics, and attitudes. Equifax supplemented the survey with its own data on age, income, education and other characteristics for each household.

From the Equifax data set, we selected only households that reported their

communities had an ongoing recycling program (N=1939). Those households who report no recycling program were not asked to report recycling percentages and thus were not eligible for inclusion in our data set. We then appended unit pricing data from three sources. The first is a 1997 report (Miranda and LaPalme, Unit based pricing in the United States: A tally of communities, Nicholas School of the Environment, 1997) that 6 During 1992, the number of curbside recycling programs in the U.S. increased by 10 percent, from just under 4000 to 5404 (Steuteville and Goldstein, 1993).
7 These 4600 households were selected using a stratified sampling method from Equifax's 250,000 member Home Testing Institute Panel. For this panel of homes, Equifax has extensive data on socio-economic household characteristics such as income and education. The 4600 households were selected to provide a mix of ages and household income levels representative of the middle and upper middle class populations in these regions.








identifies which U.S. communities had a unit pricing program for solid waste collection in 1992. The second is an EPA survey (1993), which collected information regarding the actual unit prices charged in 1992 by many of the unit pricing communities that were then in existence. For those communities not included in the EPA survey, we conducted our own telephone survey of community solid waste officials to solicit information on unit prices and other characteristics of the unit pricing program.

Following our telephone survey, we eliminated 123 additional households living in communities with unit pricing from our data set for various reasons. The most common is that we were unable to contact a government representative who could provide information about the unit pricing program. In other cases, the community had multiple trash haulers and solid waste user fees, and we were unable to connect a particular household to a particular fee level. In addition, we deleted several observations due to missing values.

Finally, to reduce the bias associated with avid recyclers being more likely than others to know about drop-off programs, we retained only those households that reported the availability of curbside collection of at least one of the five materials. Stated differently, we excluded from our sample all households living in communities with only drop-off recycling. The reason is that drop-off programs are notorious for being poorly publicized.

Conversely, curbside programs are well promoted and widely recognized, at least in part because of the visibility of curbside containers on collection day. Where drop-off and curbside programs co-exist, drop-off programs are often jointly promoted with curbside recycling. For example, certain occasions such as the introduction or revision of








a curbside program, warrant distribution of instructions for curbside recycling. (Instructions are also distributed to new residents of a neighborhood.) These instructions outline which materials can be placed at the curb and which cannot, and often give instructions for recycling the latter materials at existing drop-off centers.

The extent to which drop-off recycling is promoted alongside curbside recycling varies across communities. Widespread awareness of a curbside program certainly does not guarantee widespread awareness of drop-off centers. To identify communities where residents do have good information about all recycling options, including the less visible drop-off programs, would require data that was unavailable to us, such as community level information on recycling promotion expenditures. In the absence of such data, however, we can reasonably expect that the bias associated with endogeneity of reporting the existence of a drop-off program will be reduced when we eliminate from our sample those communities with only drop-off recycling.8 Our final data set consists of 1049 observations.

To examine the reliability of the policy information reported by respondents, we investigated whether respondents living in the same zip code area, the smallest geographical unit for which we had information, reported the same recycling program characteristics. There were many differences. Phone calls to municipalities as well as anecdotal information suggest that recycling programs differ across neighborhoods even within the same zip code. For instance, curbside recycling is often introduced to a region one neighborhood at a time. Gradual introductions might especially affect data for 1992 when many curbside programs had only recently been initiated. Another possibility is that urban parts of a zip code have curbside recycling while rural parts do not.
8 We discuss the implications of this concern about bias from avid recyclers for our results in Results.








Of the final 1049 observations, 116 are households facing a positive unit price for solid waste collection. Table 4-3 identifies the MSAs with unit pricing programs, the number of communities within each MSA with its own unique unit price, and the number of respondents residing in each MSA. The highest concentration of these respondents is in the Portland MSA, within which 37 respondents reside in the city of Portland, and nine respondents reside in four other Portland MSA communities each of which charges a unique unit price. Another concentration is in the Seattle MSA within which 18 respondents reside in the city of Seattle and 16 reside in six other Seattle area communities, each with its own unit price.

The majority of respondents facing a unit price live in western states. Of the 116 households, 104 live in communities with subscription programs where households subscribe to collection of a pre-specified number of containers. Households can change that number but the waste collection service must be notified (usually by telephone or mail) of the household's desire to change. This feature combined with weekly variations in trash generation probably leads to partially filled containers during some weeks and to storing excess waste until the next collection day during other weeks. The remaining 12 households live in communities with bag/tag/sticker programs where households place their garbage in specially marked plastic bags, or place specially marked tags or stickers on regular garbage containers, and pay a price for the specially marked items that includes the cost of collection. In these communities, households can more readily alter the number of containers discarded.

We define the marginal price of solid waste collection as the price of the second container of waste. The reason is that households virtually always generate some solid








waste, so paying for collection of the first container is difficult to avoid. Not paying for the second container is more likely and can be achieved by increased recycling.9 Figure I shows the distribution of values for the price of the second container across the 116 households with unit pricing. The values range from $0.41 to $3.46. Households in communities with no unit pricing face a zero marginal price for solid waste collection.

Table 4-4 gives the mean values and standard deviations of the independent

variables used in our ordered logit analysis. The first row gives the mean marginal price of solid waste collection (PRICE-SW), $1.91 per 32-gallons, faced by the 116 households in communities with unit pricing programs. Two communities have a different price structure for yard waste and the second row of Table 4-4 gives the mean marginal price of yard waste collection (PRICE-YW). Subsequent rows report information on the characteristics of the recycling programs and the socioeconomic characteristics of the respondents.

In addition to the data reported here, we created a series of dummy variables that indicate the metropolitan statistical area (MSA) where each household is located. This variable is used in the regressions to control for unobserved regional effects such as weather and cultural differences.

Comparing a subset of the socioeconomic data in Table 4-4 with 1990 U.S.

Census information about the characteristics of the general population in the 20 MSAs from which the sample is drawn, illustrates the effect of targeting middle and upper income households. While the sample has approximately the same household size distribution as the larger population, the sample is more highly educated; 44% of


9 Perhaps less easily, households also can avoid paying for the second container by generating less garbage in the first place.








respondents graduated from college while only 22% of the larger population did. In addition, the sample under-represents the lower income segments of the population and over-represents households with incomes between $50,000 and $75,000. However, the median income of the sample is roughly $40,000 which is only $5,000 above the median income in 1990 for the group of 20 MSAs. The sample also has a higher proportion of detached home dwellers than the population at large. These comparisons make explicit the fact that our results should be generalized only to middle and upper income segments of the population

We construct the dependent variable in our analysis using survey responses about recycling participation. Respondents were asked what proportion of the following materials they recycled through all available recycling programs: steel sided cans, glass bottles, plastic bottles, newspaper, magazines, aluminum, other plastics, yard waste and other. As noted already, we chose to study five of these materials and constructed a dependent variable for each of the five. The survey asked whether recycling percentages fell into one of seven possible categories: 0 to 10%; 11 to 25%; 26 to 50%; 51 to 74%; 75 to 84%; 85 to 95% or over 95%. We aggregate the data into three categories of "proportion of the material recycled" - 0 to 10%, 11 to 95%, and over 95%. Table 4-5 gives the percent of respondents falling into the three categories for each of the five materials. Except for yard waste, the majority of respondents recycled over 95 percent of each material. Table V also gives the number of respondents falling into each category and the number of missing observations for each of the five ordered logit equations.

Model Specification

The model that we estimate seeks to identify which policy and socioeconomic factors influence the level of recycling effort households expend on each recyclable








material. We use a latent regression model for ordered data as the framework for estimation. As noted above, for each material type, we define three ordered categories: category 0 for 0 to10% recycled, category 1 for 11 to 95% recycled and category 2 for over 95% percent recycled. For each material type, j, we consider the relationship

Y = /J., x1 + ejj (4-4)

where y * is unobserved level of recycling effort (percentage of material j recycled) and i is an index of households. The vector xi contains the marginal price, recycling program attributes, and socio-economic features for each household. P is a vector of coefficients to be estimated by maximum likelihood estimation (MLE) in an ordered logit model.'0 Assuming 1., is distributed standard logistic, the probability that we observe household i in category k, where k-0, 1 or 2, for materialj is given by




Pr(yji = 0)= 1, (4-5)
1 + epjxji

Pr(Yji =1)= I I 1 (4-6)
1 + e-+pjxji 1 + epjxji

Pr(yji = 2)= 1 - 1 (4-7)
1 + e- +jxji



10 We select the ordered logit specification instead of the ordered probit because the binomial logit is more amenable to incorporating fixed effects than the binomial probit. (Hsaio, 1986). In the case of a binomial logit or probit, traditional maximum likelihood estimators for the 3's will be inconsistent when fixed effects are included in the model. However, the conditional logit model (McFadden, 1974) can be used to find consistent parameter estimates for a logit when fixed effects are included. Unfortunately, the consistent estimator of 03 in a model with fixed effects that is well defined for an binomial logit is not well defined for an ordered logit. Therefore we simply estimate a regular ordered logit with regional metropolitan statistical area dummy variables included. We have also estimated the same model using an ordered probit specification and we find that the results (in terms of which variables are significant and the signs of the effects) are virtually the same.









Results

The intensity of household recycling activities by material is modeled as a

function of the socioeconomic variables and policy variables that are described earlier. We used the same set of independent variables for each material, except that the values for the curbside and drop-off indicator variables varied across materials depending on the type of collection available for the specific material. In addition, the marginal disposal price was different for yard waste.

The results of the econometric estimation of the ordered logit regression for each material are presented in Table 4-6. These results indicate the significance and direction of each variable's effect on the propensity to recycle different materials.1 Because of the non-linear estimation procedure employed, the regression results in Table 4-6 do not provide a good indicator of the magnitude of the effect. To determine magnitudes, we use the estimated logit model coefficients to calculate the marginal effects of different independent variables on the probability that a typical household will fall into each of the three levels of recycling intensity: 0 to 10% of the material recycled, 11 to 95% recycled or over 95 percent recycled. 2 For the significant policy variables, these marginal effects



To examine the sensitivity of the results to our three-way partition of the dependent variable, we also estimated equations with the dependent variable separated into only two partitions - households who recycle between 0 and 10% of a material and those who recycle greater than 10%. For the aluminum, plastic bottle and yard waste equations, the significant policy variables remained so. However, for the newspaper equation, the indicator variable for drop-off collection and the variables representing the number of materials picked up curbside and the length of the recycling program became insignificant (although of the same sign) under the binomial specification. For the glass bottles equation, the indicator variables for both curbside collection and drop-off collection changed to insignificant (although of the same sign) under the binomial specification. Some of the socioeconomic variables that were significant under the multiple category specification became insignificant under the binomial specification. Overall, the binomial specification gave similar, but somewhat weaker results for the bulk of the materials. (For this sensitivity analysis, we use a binomial logit model with MSA dummy variables instead of a conditional binomial logit model with fixed effects. We do this in order to provide the most straightforward comparison to our ordered logit model with three categories.
12 The equation that predicts the probability that an observation will fall into each of the three categories is non-linear in the independent variables. Therefore, the equation that defines the marginal effects of each








are reported in Table 4-7. The table also converts the marginal effect into a percentage change from the actual probability a respondent will fall into a category and reports these percentage changes in parentheses.

The diverse nature of the communities and households represented in our data set led us to question the appropriateness of the standard assumption that all of the disturbance terms in the underlying model have a common variance. In particular, we suspected that the variance of the disturbance terms surrounding the propensity to recycle could be a function of the presence of curbside recycling and the length of time that the recycling program had been in existence. We hypothesize that the variance of the regression disturbance terms are likely to be different across households that have curbside recycling for the relevant material and those that do not. By eliminating the need to transport recyclables to drop-off points at varying distances from the household, curbside recycling tends to even out the time required to recycle across households resulting in less variation in errors. Likewise, we expect households with greater potential experience with recycling to have disturbance terms with a lower variance than those with less experience with recycling. Greater experience with recycling allows households to develop a recycling habit, which will lead to less variation in the error terms.

Using these variables as determinants in a multiplicative model of

heteroskedasticity of the form eji = exp(n) where the z vector includes the three potential contributors to heteroskedasticity, we tested the ordered logit model for each



independent variable on that probability is a function of all of the independent variables. We calculate marginal effects by using the average value for all of the independent variables except where noted in Table 4-7.








material for the presence of heteroskedasticity.13 We found that for two materials, glass and plastic bottles, we could reject the null hypothesis of homoskedasticity. Thus, we apply Harvey's multiplicative heteroskedasticity correction to the models for those two materials (Harvey, 1976).

In the next three subsections, we discuss the results for three categories of independent variables: recycling program features, unit pricing policies and socioeconomic characteristics. A final section describes potential problems presented by our data and their solutions.

Recycling Program Features

This analysis identifies several features of recycling programs that have a

significant effect on intensity of household recycling effort. Two features that are always significant are availability of local drop-off recycling and existence of curbside recycling. Increasing the number of total materials included in the curbside recycling program has a positive effect on recycling effort for newspaper only. Length of program life is also an important determinant of the intensity of recycling effort for newspaper and yard waste. The effects of individual program features are discussed in greater detail in the following paragraphs. 14


13 There are three potential contributors because the amount of time a recycling program has been in place is represented by two categorical indicator variables. 14 One popular program to encourage recycling of beverage containers is a deposit refund program. During the time period of our data, deposit refund programs existed in 10 states, five of which (New York, Massachusetts, Connecticut, California and Oregon) are sampled by our data set. However, the questionnaire directed respondents to report the percentage of materials recycled but to exclude containers returned for a deposit. Assuming that beverage containers are easy to recycle, a possibility is that excluding these containers from consideration might reduce the percentage of the waste stream that is easily recyclable. Thus, states with bottle bills might be less responsive to recycling incentives. Unfortunately, we are unable to test for this directly in our model without excluding the MSA dummy variables, which would create potential endogeneity problems. However, we did look at the coefficients on the indicator variables for those MSAs that have deposit-refund programs to see if they were systematically different in some way from those for the other regions. For glass bottles we saw no discernable difference. For aluminum cans, most of the coefficients for the MSAs with bottle bills were insignificant.








The two most commonly significant recycling program policy variables, the dropoff and curbside program indicators, serve as proxy measures of the convenience of recycling. Introducing a local drop-off program for recycling of a particular material decreases the time and storage costs associated with other modes of recycling such as accumulating materials to haul to more distant recycling centers or participating in infrequent recycling drives for charity. Instituting a curbside recycling program makes recycling even more convenient, thus its effect on recycling effort should be bigger than the effect of a drop-off program. Curbside collection lowers the time and out-of-pocket costs of recycling by completely eliminating the need to transport recyclables to collection points or to store them for long periods of time. The results reported in Tables 4-6 and 4-7 conform to these expectations.

The econometric results reported in Table 4-6 show that for all materials,

instituting a local drop-off program has a positive and significant impact on intensity of recycling effort. The marginal effects reported in Table 4-7 show that the magnitude of the effect of the drop-off program variable varies dramatically across materials. Introducing a local drop-off program increases the probability that over 95% of all glass bottles used in the household are recycled by 42 percentage points; for plastic bottles the marginal effect is 33 percentage points and for aluminum and newspapers it is 19. These results suggest that introducing a local recycling option has a smaller impact on materials for which there were recycling options even before the local drop-off program.15 Charity drives, for example, have traditionally focused on collecting newspapers and/or aluminum. Newspaper carried to (or even purchased at) work may be recycled at work 15 This effect might be exaggerated because of a possible over-representation of avid recyclers reporting drop-off programs. Avid recyclers might be more likely to seek out recycling alternatives in the absence of a local program.








and beverage cans used outside the home may be recycled at the place of use. Adding a local drop-off program is likely to have little impact on this type of recycling behavior.

The different magnitudes also suggest that introducing a local drop-off program has a greater impact on materials for which transportation and storage would be most difficult for households. Without a local program, for materials not collected by special drives or recycled away from home, the household must travel to a distant recycling center. Relative to glass and plastic bottles, newspapers and aluminum (after it has been crushed) are compact and clean and thus more likely to be accumulated and transported long distances. In contrast, storing and transporting glass and plastic bottles is more burdensome to households. Adding a drop-off recycling program reduces households' transportation costs by improving the proximity of recycling centers. Improved proximity might also increase the frequency of drop-offs that would reduce households' storage costs. Thus, it is not surprising that introducing a drop-off program has a bigger impact on glass and plastic bottles than on newspapers and aluminum.

Introducing a local drop-off option for yard waste increases the probability that over 95% of it will be recycled by 19 percentage points. While the magnitude of this effect is similar to newspaper and aluminum, it represents a percentage increase above baseline recycling levels similar to that experienced for glass and plastic bottles approximately 60. (Table 4-7 presents these percentage increases or semi-elasticities in brackets.) This finding suggests that drop-off recycling has a larger effect on yard waste (a material with high transportation and storage costs) than appears at first glance.

As expected, the presence of curbside recycling has a positive and significant

effect on intensity of recycling activity for all five materials. The magnitude of this varies








substantially across materials, just as the magnitude of the effect of the drop-off option did. Table 4-7 shows that introducing a curbside recycling program increases the probability that the average household recycles over 95% of glass and plastic bottles by more than 50 percentage points; aluminum by more than 40 percentage points; and yard waste and newspaper by around 25 percentage points. The interpretation of the differences across materials is similar to that offered for the drop-off program variable. Bulky and potentially messy materials such as glass and plastic bottles are difficult to transport and thus more responsive to the introduction of curbside than are other materials. Also, the small percentage point response of yard waste to curbside recycling actually represents a fairly substantial percentage increase over baseline recycling levels.

Table 4-7 also shows the marginal effects of replacing an existing drop-off

recycling program with a curbside recycling program. The size of the difference is fairly similar for glass bottles, aluminum and plastic bottles. Replacing a drop-off program with a curbside program leads to roughly a 20% increase in the probability of recycling over 95% of these materials. For newspaper and yard waste, replacing a drop-off program with a curbside program increases the probability of recycling over 95% by about 5%. This is a small percentage change for newspaper (9%) and a slightly larger percentage change for yard waste (14%).

Experience with a recycling program has a positive effect on recycling effort for newspaper; for yard waste, experience is significant only once the recycling program has lasted at least 2 years. Table 4-7 reports the marginal effect of having a program in place for more than two years versus having it in place between one and two years. The magnitudes are quite small. For yard waste, greater experience with recycling programs








increases the probability that over 95% of it is recycled by less than 5%. In the case of newspaper, while program length has a positive effect on recycling effort, the coefficient on the indicator variable for a program of over two years in length is smaller than the coefficient on the indicator variable for a program of between one and two years in length. This means that the marginal effect of going from a program of 1 to 2 years in length to a program of over 2 years in length is actually negative, but only slightly so. The finding that recycling effort increases with experience is consistent with Reschovsky and Stone (1994) which finds that the probability of participating in recycling rises for newspaper, glass, plastic, cardboard, metal and composting when households feel knowledgeable about the recycling program.

Our findings on the effects of other features of curbside recycling programs are mixed. The total number of materials collected curbside has a small, significant, positive effect on the intensity of newspaper recycling. Increasing the number of materials collected curbside by 1 leads to a 2.5% increase in the probability that a typical household will recycle over 95% of its newspaper waste. Making a curbside recycling program mandatory has no statistically discemable effect on intensity of recycling effort for any of the materials.16 This finding is congruent with Kinnaman and Fullerton's




16 This finding could be attributable to a lack of enforcement of a mandatory recycling rule or law. If people perceive that the rule will not be enforced, then they have no incentive to comply. Unfortunately, we were unable to obtain systematic information on the enforcement of mandatory recycling requirements. The information we did locate was unclear about whether enforcement was for a mandatory recycling requirement or for curbside separation rules. Folz (Recycling policy and performance: Trends in participation, diversion, and costs, working paper, 2001) reports that the use of enforcement tactics increased from 37 percent of the cities in his sample in 1989 to 55 percent in 1996. The enforcement tactics used are described as refusing to pick up trash, tagging bins with instructions about proper recycling practice or issuing written warnings about improper separation of recyclables from other solid wastes. We conclude that there is evidence that mandatory requirements are sometimes enforced, however, we do not have a clear sense of how often they are enforced








(2000) result that communities in states with mandated recycling do not recycle significantly greater quantities.

Unit Pricing Policy Variables17

The econometric results reported in Table 4-6 indicate that the price of disposal is not a significant determinant of intensity of household recycling effort for any of the materials. This finding suggests that increasing the price of disposal does not increase the intensity of recycling effort. There are several possible explanations why the data reveal no effect.

First, the average price of disposal for the unit-pricing communities in our sample simply may be too low to create a response from our relatively high-income households. The sample's median household income is approximately $40,000 per year, which equates to an hourly wage of roughly $20. At that wage level, if the amount of time associated with recycling 32 gallons of trash is more than 5.75 minutes then the household will have time costs of recycling that exceed the avoided $1.91 average disposal charge. Thus, as an incentive to recycle, unit pricing is ineffective.

Second, a disposal price provides only an indirect signal to increase recycling, whereas it provides a direct signal to reduce trash. When faced with the prospect of paying a unit price for trash disposal, households may respond by changing their purchasing habits or making other changes in behavior that have a more direct impact on waste disposal. We return to this point in the conclusion of the paper.




" Initially we set out to identify the effects of the level of the disposal price as well as other unit pricing program characteristics such as program type (bag/tag/sticker, subscription can) on the propensity to recycle. However, due to the small number of observations for bag/tag/sticker communities, we were unable to identify the effects of the type of unit-pricing program on the intensity of household recycling efforts.








Finally, most of the unit pricing programs included in our sample are subscription can programs which provide a discontinuous signal to reduce disposal and therefore may provide only a weak incentive to households to recycle instead of disposing of solid waste (Nestor and Podolsky, 1998). Communities with bag/tag/sticker unit pricing programs may be more responsive to a given price level.

Our finding of no effect of disposal price on recycling efforts is consistent with the findings of earlier studies by Kinnaman and Fullerton (2000), Fullerton and Kinnaman (1996) and Reschovsky and Stone (1994). All of these earlier studies find that unit pricing does not significantly affect the level of recycling or the probability of participation in recycling programs. However, our findings differ from those of Hong (1999), Callan and Thomas (1997) and Hong et al. (1993). For a sample of Korean households, Hong (1999) finds that a unit price has a significant positive effect on the recycling rate and that the elasticity of recycling with respect to price is approximately

0.5. Callan and Thomas (1997) finds that the presence of unit pricing increases the recycling rate by approximately 6.5 percent. Hong et al. (1993) indicates that unit pricing increases the probability that households will participate more often in recycling. Van Houtven and Morris (1999) finds that the presence of unit pricing positively affects the probability that a household will participate in recycling but has no effect on the quantity of recyclables set out for collection.

Socioeconomic Factors

The econometric models also include a number of socioeconomic variables

describing various characteristics of the households. The statistical significance and size of the effects of these variables on intensity of recycling effort vary substantially across materials. Below we discuss those variables that have a statistically significant effect.








Household income has a significant and positive effect on intensity of recycling effort for newspaper only.18 We can calculate the "marginal" effects of moving from one income category to the next highest income category. For example, we find that for a typical household, moving from the "income between $35,000 and $49,999" category to the "income between $50,000 and $74,999" category leads to a 3.6% increase in the probability of recycling over 95% of all newspaper waste generated.

The level of education attained by the most highly educated person in the household has a significant but small effect on intensity of recycling effort for all materials except plastic bottles and yard waste.19 The marginal effects for a typical household of moving from the "high school graduate" category to the "college graduate" category is to increase the probability of recycling over 95% of aluminum and newspaper by 0.1% and 1.5% respectively. Curiously, for glass bottles, the level of education has a small negative effect on intensity of recycling effort.

A number of other socioeconomic variables also influence the intensity of yard waste recycling efforts. Increasing population density by 1000 persons per square mile leads to a 1.3% increase in the probability that a typical household recycles 10% or less of its yard waste. A likely reason is a growing scarcity of appropriate outdoor storage space as population becomes denser. Residents of single-family dwellings are substantially more likely to recycle larger quantities of their yard waste than residents of multi-family dwellings. Again, the reason might be a lack of outdoor or indoor storage IS This finding is in harmony with the theoretical results in Saltzman et al. (1993). They suggest that as long as newspaper is a normal good, because its recyclable content cannot be altered by the household, the impact of income on recycled newspaper should always be positive. The impact of income on other recyclable materials will be determined by whether the goods that are the source of the material are normal as well as to what extent the household can reduce the amount of a material associated with a good. For example, the glass content of beverage products can be reduced by switching to plastic or cardboard so that when income increases, glass recycling might decrease. 19 Recall that the sample is more highly educated than one would expect of a randomly selected sample.








space. Household size has a significant and positive effect on recycling efforts for glass bottles and yard waste. Increasing the number of occupants of the average household by

1 person leads to a 3% increase in the probability that the household will recycle over 95% of its glass bottle waste and a 2% increase in the probability of recycling over 95% of yard waste. This finding may be due to the fact that glass bottle and yard waste recycling are time intensive -- bottles must be cleaned, yard waste must be bagged. As the number of occupants rises, the amount of time required from each occupant decreases thereby reducing the implicit cost on any one individual. Finally, age has a positive, but small, impact on intensity of recycling for all materials except glass bottles. Data Limitations

A perennial problem with survey research is the nonresponse problem: are there systematic biases introduced into the data by the exclusion of those who failed to respond to the survey? A concern regarding our own data is that bias exists because respondents who are avid recyclers may have been more inclined to mail back the questionnaire. These individuals may recycle on their own initiative and thus, compared to the general population, be less responsive to recycling incentives. The methods for correcting such bias require information about the characteristics of the questionnaire recipients who did not respond, or, at a minimum, street addresses for those who did not respond.20 Unfortunately our efforts to obtain that information were futile.

There is another reason why our data might mis-state the response to recycling incentives that would be expected from the general population. Avid recyclers may be better informed than others about the existence of recycling options. This problem is less 20 Cameron et al. (1999) discuss econometric methods for dealing with non-response bias in mail surveys. Their method uses zip-code-level information for non-respondents combined with data collected from the survey to estimate a pooled-data probit model for response probability.








of a concern for curbside recycling options that are quite visible to the least enthusiastic recyclers. The concern is larger for drop-off recycling options. If avid recyclers are more knowledgeable than other community members about all recycling options, their natural enthusiasm for recycling might make them appear less responsive to a drop-off program. In the absence of a drop-off option, avid recyclers are more likely than others to have already located recycling opportunities outside their own community. Thus, our estimate of the response to drop-off recycling incentives might be biased downward.

However, it might also be biased upward. The general population might be less responsive than avid recyclers to drop-off programs simply because they are less likely to be aware of a program's existence. The net result of these two opposite sources of bias is unclear. Future research could clarify this uncertainty by analyzing all households rather than only households who report awareness of program existence.

Our data set lacks information on community characteristics that might influence the intensity of household recycling effort for all members of a community. Communitylevel variables such as measures of recycling promotion activities or the general attitude toward environmental issues should be included in our equations because omitting these variables can cause a problem of endogeneity. Of particular concern is that the excluded community level variables might be correlated with the dependent variable, which would result in biased coefficient estimates for the included independent variables. To address these concerns, we test for the significance of regional indicator variables which are included to capture unobserved community-level heterogeneity. 21 The results of F-tests


21 To examine the sensitivity of the results to the use of an ordered logit-model specification with fixed effects, we estimated a binomial logit model with fixed effects using a conditional maximum-likelihood estimator. For this model we partitioned the dependent variable into two groups- respondents recycling between 0% and 10% of a material and those recycling greater than 10%. The sign and significance of the








suggest that MSA level indicator variables as a group are significant determinants of recycling intensity for each material.

Conclusion and Policy Implications

This study uses a unique household-level data set, representing primarily middle and upper income households in 20 MSAs across the country, to examine the effect of two popular solid waste programs, curbside recycling and unit pricing, on the percent recycled of five different materials found in the municipal solid waste stream: glass bottles, plastic bottles, aluminum, newspaper, and yard waste. The study also assesses the impact of other attributes of recycling programs (e.g., mandatory or voluntary) along with socioeconomic characteristics of households on recycling activity. The results presented here provide new insights that could help policy makers to improve the costeffectiveness of a community's recycling program and to design a program to achieve mandated recycling rate goals. Consistent with expectations, a curbside recycling program increases households' intensity of recycling and the results differ across recyclable materials. The effect of a unit pricing program, on the other hand, is less clear.

The analysis indicates that drop-off and curbside recycling programs increase

households' intensity of recycling for the five materials. The magnitude of the effect of these programs varies dramatically across materials with the largest impacts on glass and plastic bottles. The size of the impact on yard waste recycling effort is also large relative to the average intensity of recycling effort observed in the sample. We conclude that policy variables were very similar for all the equations except the one for newspaper. For it, many of the significant policy variables became insignificant under the binomial specification except the curbside indicator variable which remained significant. There were more differences in the sign and significance of the socioeconomic independent variables. In many cases, a variable that was significant under the ordered logit specification was insignificant under the binomial logit specification. The sign changes were only for insignificant variables. Overall, we chose to focus on the ordered logit results because the advantages from the third partition seemed to outweigh the disadvantages of modeling fixed effects with MSA indicator variables.








introducing a local recycling option has a smaller impact on materials, such as newspaper and aluminum, for which there were recycling options such as charity drives or workplace or other away-from-home recycling stations even before the local drop-off program.

We further conclude that introducing a local drop-off program has a greater impact on materials such as glass and plastic bottles whose transportation and storage would be most difficult for households. Local governments should take this finding into consideration when selecting which materials to include in a recycling program.

Curbside recycling programs have a bigger effect on behavior than drop-off

programs. For three of the materials, a curbside program increases the probability that the average household recycles over 95% by approximately 20% more than the increase generated by a drop-off program. Nonetheless, drop-off programs also are effective at increasing recycling. A budget-constrained community with no recycling program at all could see measurable waste diversion with the introduction of a less expensive drop-off alternative. Local governments considering implementing curbside recycling could compare the benefits of the expected increase in recycling activity to the incremental costs of implementing curbside as opposed to drop-off recycling.

The impact of unit pricing on the intensity of recycling effort for specific

materials is less clear. Unit pricing gives a direct incentive to decrease waste quantities. In response to such a program, households might adjust their consumption towards goods that generate easy-to-recycle wastes, likely those wastes eligible for collection by a local recycling program.22 These easy-to-recycle wastes increase in quantity; however, the 22 Such an adjustment is suggested by Hong (1999) which finds that the price elasticity of total waste quantities is positive but the price elasticity of non-recyclable waste quantities is negative.








percentage of that quantity that is recycled might not change. If unit pricing does increase recycling quantities by shifting consumption toward materials that are collected by a community's recycling program, its impact on recycling will not be detected by examining the percent of a material a household recycles.

Our findings indicate that the added convenience created by a recycling program creates a stronger incentive to recycle than having to pay at the margin for trash disposal. Of course, the levels of unit prices charged are important to the impact of the program. The mean fee for our sample was $1.91 per 32-gallon container. At these price levels, collecting more materials at curbside will produce greater waste diversion than will implementing unit-pricing. However, if the costs of adding a particular material to a curbside program exceed the waste diversion and recycling revenue benefits of doing so, then adding materials may not be worthwhile.

Recycling programs appear to become more effective over time. Greater experience with a recycling program leads to increased recycling effort directed at newspapers and yard waste. However, the magnitudes of these effects are quite small. Of interest to policy-makers perhaps is that this effect is not negative; that is, households do not appear to become less enthusiastic over time about participating in recycling.

Of course, which materials to include in a recycling program also depends on the market prices of recyclable materials and on collection and processing costs. For example, our findings suggest that introducing curbside recycling has a big effect on the recycling of plastic bottles, one of the highest valued materials of those we studied.3 However, collection and transportation costs are also high for plastic bottles due to their 23 The following are average prices recyclers were paying for materials in late January or early February, 2000 in 8 urban centers across the country: Aluminum cans -$750 per ton; Natural HDPE (a type of plastic container) -- $300 per ton; Newspaper number 8 -- $70 per ton; Amber glass -- $27 per ton (Truini, 2000).








low density. Policy makers can combine the insights from this study with information on the material composition of their local waste stream, local collection and transportation costs and current market prices for recyclable plastic to decide if curbside recycling is a cost-effective means of managing plastic waste.

The study suggests several issues for future research. First, due to a lack of

variation in our data, we were unable to analyze the differences in responses to the two main approaches to implementing unit pricing for solid waste disposal services: bag/tag/sticker versus subscription can. Van Houtven and Morris (1999) and Nestor and Podolsky (1998) analyze data from Marietta, Georgia and find that there are differences and that a bag program causes larger reductions in waste quantities than a subscription can program. Future research could explore if the different program types affect recycling of different materials in different ways.

Second, the nature of our data set has limited us to focusing on recycling intensity (percentage of each material type generated by the household that is recycled). However, policy makers and solid waste planners ultimately need more information about how recycling program characteristics and unit prices affect material-specific quantities of both recycling and waste disposal by households. Providing such information requires national household-level data on quantities of materials recycled and discarded.

Third, research into the costs of implementing curbside recycling programs with different scopes compared to the costs of implementing a drop-off program as well as a unit pricing program would be useful to policy makers seeking to design effective and efficient waste management strategies.








Finally, our data sample is focused on middle and upper income households in

urban and suburban areas of the U.S. Therefore, our conclusions are applicable to these types of households. Future research is needed to identify the implications of solid waste and recycling policies for the recycling behavior of lower income households, of households in more rural areas in the U.S., and of households in other countries.24


24 Hong (1999) has analyzed data for Korea, Sterner and Bartelings (1999) for Sweden.






84


Table 4-1. Previous studies on effects of unit pricing and recycling programs on effort
Author(s) (year) Dependent Independent Policy Variables Data Variable
Aggregate or Unit Price Recycling National or Household Level Material Specific Program Regional Recycling Attributes Quantities
Van Houtven and Aggregate No, but dummy No Regional - Yes


Morris (1999)



Hong (1999)
Hong and Adams (1999)
Sterner and Bartelings (1999)


Kinnaman and Fullerton (2000) Nestor and Podolsky (1998) Callan and Thomas (1997)


Fullerton and Kinnaman (1996) Rechovsky and Stone (1994)


Hong et al. (1993)



Saltzman et al. (1993)


(quantity is weight, not volume)


Aggregate Aggregate

By Material (community proportion, not quantity recycled) Aggregate

Aggregate

Aggregate (percent of total waste stream recycled) Aggregate

By Material (proportion, not quantity, recycled) Aggregate (recycling participation yes/no) By Material


for presence of each of two types of unit pricing program Yes
Yes


Marietta, Georgia



National - Korea Regional Portland, Oregon Regional Southwest Sweden

National

Regional Marietta, Georgia Regional Massa-chusetts


Regional Charlottes-ville Regional upstate NY


No, but dummy for presence of unit pricing program Yes

No, but dummy for presence of unit pricing program Yes



No


Regional Portland


Regional - PA and NJ










Table 4-2. Metropolitan Statistical Areas sampled Boston/Hartford Corridor Detroit New York Metro (New Jersey side) Philadelphia Minneapolis/St. Paul Atlanta San Francisco Phoenix Houston Tampa New York City Metro (New York and Connecticut) Portland Camden, New Jersey Chicago Seattle St. Louis Los Angeles Dallas-Fort Worth Miami Denver Boston/Hartford Corridor










Table 4-3. Unit pricing programs
MSA Number of Communities Number of Observations Program Type with Unique Unit Price
Los Angeles 1 1 Subscription San Francisco 8 20 Subscription Chicago 7 10 Bag/Tag/Sticker Detroit I I Bag/Tag/Sticker Minneapolis/St. Paul 2 2 Subscription Portland 5 46* Subscription Philadelphia 1 2 Bag/Tag/Sticker Seattle 7 34** Subscription
*Of the 46 households living in the Portland MSA, 37 are located in the city of Portland.
**Of the 34 households living in the Seattle MSA, 18 are located in the city of Seattle.










40


35 4


30


0 .)
0
�20 S15 10







0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8
Unit Price ($ per 32 gallons) Figure 4-1. Distribution of unit prices






88


Table 4-4. Summary statistics for independent variables in recycling logit regressions Variable Mean Standard Deviation

PRICE-SW ($ per 32-gal.) $1.91 $0.86 PRICE-YW ($ per 32-gal) $1.90 $0.86 NEWSPAPER CURB 0.916 0.278 GLASS BOTTLES CURB 0.886 0.318 ALUM CURB 0.853 0.355 PLASTIC BOTTLES CURB 0.775 0.417 YARD WASTE CURB 0.528 0.500 TOT MATERIALS CURB 3.900 1.200 NEWSPAPTER DROP 0.056 0.229 GLASS BOTTLES DROP 0.071 0.256 ALUM DROP 0.104 0.305 PLASTIC BOTTLES DROP 0.127 0.333 YARD WASTE DROP 0.057 0.232 MAND RECY PGM 0.528 0.500 RECY PGM 1 TO 2 3.900 1.200 RECY PGM > 2 0.056 0.229 POP DENS 5.820 5.923 INC 10000 TO 14999 0.068 0.251 INC 15000 TO 24999 0.133 0.339 INC 25000 TO 34999 0.135 0.342 INC 35000 TO 49999 0.208 0.406 INC 50000 TO 74999 0.258 0.438 INC > 75000) 0.140 0.347 HOUSEHOLD SIZE 2.700 1.400 HOUSEHOLD HEAD AGE 47.900 15.900 DETACHED HOUSE) 0.726 0.446 HOME OWNERSHIP 0.793 0.405 EDUC HS GRAD 0.511 0.500 EDUC COLL GRAD 0.247 0.431 EDUC BEYOND COLL 0.195 0.397 Note: In the logit equations, this PRICE-SW and PRICE-YW are in dollars per gallon. Also, the price variables are only for the 116 households living in programs with unit pricing programs. The mean values for the material-specific variables use the numer of observations in the relevant logit regressions.









Table 4-5. Proportions of materials recycled Material Percentage and Percentage and Percentage and Total Number
Number of Number of Number of Missing
Respondents Respondents Respondents
Recycling 0 to Recycling 11 to Recycling Over
10% 95% 95%
Newspaper 8.8% 16.6% 74.6% 100% 92 173 777 1042 7 Glass Bottles 11.3% 22.2% 66.5% 100% 117 229 687 1033 16 Aluminum 15.0% 21.8% 63.2% 100% 152 221 639 1012 37 Plastic Bottles 17.8% 28.0% 54.2% 100% 180 284 550 1014 35 Yard Waste 43.3% 22.8% 33.9% 100% 417 220 326 963 86









Table 4-6. Recycling participation ordered logits Independent Newspaper Glass Bottles Aluminum Variables


DROP

CURB

MAND RECY PGM*CURB TOT MATERIALS CURB RECY PGM 1 TO 2 RECY PGM > 2 PRICE

POP DENS INC 10000 TO 14999 INC 15000 TO 24999 INC 25000 TO 34999 INC 35000 TO 49999 INC 50000 TO 74999 INC > 75000 HOUSEHOLD SIZE HOUSEHOLD HEAD AGE DETACHED HOUSE HOME OWNERSHIP EDUC HS GRAD EDUC COLL GRAD EDUC BEYOND COLL Constant


Number of observations Log likelihood Chi squared statistic Heteroskedasticity corrected
* 10% level of significance
** 5% level of significance
*** 1% level of significance


0.7886* (0.4594)
1.1423*** (0.4164) 0.2467 (0.2030) 0.1434* (0.0772)
0.6111*** (0.1907)
0.5614*** (0.2091)
-11.2608 (7.8743) 0.0056 (0.0179)
0.9985** (0.4170) 0.4590 (0.3506) 0.6408* (0.3637)
0.8246** (0.3575)
1.0418**
(0.3762)
1.2203"**
(0.4277) 0.0437 (0.0646)
0.0175*** (0.0059)
-0.0065 (0.2068) 0.2651 (0.2118)
1.1600*** (0.3265)
1.2460*** (0.3641)
1.2520*** (0.3802)
-3.4140*** (0.7327)


1042
-693.146 137.609*** Yes


1033
-781.65 196.779*** Yes


1012
-849.992 136.450*** Yes


1014
-867.607 282.917**
Yes


Standard errors in parentheses.


Yard Waste


1.5184*** (0.4137)
2.1026*** (0.4733)
-0.0379 (0.1070)
-0.0917 (0.0612) 0.1668 (0.1312) 0.1737 (0.1518)
0.1227 (3.3013) 0.0000 (0.0106) 0.4387 (0.2675)
-0.0654 (0.1999) 0.0380 (0.2007) 0.0075 (0.1972) 0.1268 (0.2112) 0.0689 (0.2302)
0.0901** (0.0406) 0.0049 (0.0034) 0.0710 (0.1114)
0.3517** (0.1433)
0.4879** (0.2095)
0.4711** (0.2251)
0.5269** (0.2388)
-2.1129*** (0.6027)


0.8617** (0.3642)
1.7808***
(0.3577) 0.2041 (0.1797)
-0.1172 (0.0914) 0.1395 (0.1712) 0.2146 (0.1896)
-4.7538 (6.1760)
-0.0083 (0.0151)
0.9363** (0.4147)
-0.1214 (0.3403)
-0.1043 (0.3483) 0.0772 (0.3309) 0.0312 (0.3457) 0.0915 (0.3838) 0.0070 (0.0543) 0.0092* (0.0053) 0.1708 (0.1817) 0.3690* (0.1964)
1.0557*** (0.3112)
1.0612*** (0.3443)
0.9415**
(0.3531)
-1.7563*** (0.6690)


Plastic Bottles
2.1350*** (0.3984)
2.9083*** (0.4603) 0.2226 (0.1621)
-0.0831 (0.0790) 0.3272* (0.1670) 0.1300 (0.1745) 1.6401 (5.0294)
-0.0044 (0.0140)
1.2334*** (0.3784) 0.1881 (0.2809)
0.6419** (0.2941) 0.4555 (0.2807) 0.4673 (0.2977) 0.5157 (0.3260) 0.0717 (0.0491)
0.0097** (0.0048)
0.4021** (0.1694) 0.2911
(0.1841) 0.0862 (0.2662) 0.0533 (0.2879) 0.2480 (0.3037)
-2.8715*** (0.6885)


1. 1074*** (0.1896)
1.3111 ***
(0.1896) 0.1435 (0.2037) 0.0677 (0.0659) 0.2684 (0.1821)
0.4520** (0.2017)
-4.6836 (5.1007)
-0.051 1*** (0.0188)
-0.3413 (0.4128)
-0.1285 (0.3545)
-0.3247 (0.3634)
-0.4338 (0.3568)
-0.1164 (0.3651)
-0.2746 (0.3986) 0.1010* (0.0557)
0.0134** (0.0055)
0.7399*** (0.1880) 0.3779* (0.2165)
-0.3421 (0.3466)
-0.4060 (0.3828)
-0.2598 (0.3894)
-2.6362*** (0.6427)

963
-848.263 357.350*** Yes


Standard errors in parentheses.


Yes









Table 4-7. Marginal effects of significant policy variables from logits Policy Variable Newspaper Glass Bottles Aluminum Plastic Yard Waste Bottles
Total Materials Cubside
Recycle 0 - 10% -0.0090
[-0.1020]
Recycle 11 - 95% -0.0162
[-0.0974]
Recycle over 95% 0.0252
[0.0338]
Drop-off
Recycle 0 - 10% -0.0932 -0.5467 -0.1900 -0.5291 -0.2696
[-1.0559] [-4.8268] [-1.2647] [-2.9804] [-0.6226] Recycle 11 - 95% -0.0923 0.1284 -0.0027 0.1980 0.0743 [-0.5562] [0.5791] [-0.0123] [0.7069] [0.3252] Recycle over 95% 0.1856 0.4183 0.1926 0.3311 0.1953 [0.2489] [0.6290] [0.3051] [0.6104] [0.5770] Curbside (not mandatory)
Recycle 0 - 10% -0.1198 -0.6427 -0.3210 -0.6404 -0.3143
[-1.3567] [-5.6743] [-2.1375] [-3.6075] [-0.7258] Recycle 11 - 95% -0.1347 -0.0028 -0.0941 0.0987 0.0718 [-0.8112] [-0.0126] [-0.4308] [0.3525] [0.3141] Recycle over 95% 0.2545 0.6455 0.4151 0.5417 0.2425 [0.3413] [0.9706] [0.6574] [0.9986] [0.7164] Drop-off to curbside
Recycle 0 - 10% -0.0266 -0.0960 -0.1311 -0.1113 -0.0446
[-0.3008] [-0.8474] [-0.8728] [-0.6270] [-0.1031] Recycle 11 - 95% -0.0423 -0.1312 -0.0914 -0.0993 -0.0025
[-0.2550] [-0.5917] [-0.4185] [-0.3545] [-0.0111] Recycle over 95% 0.0689 0.2271 0.2225 0.2106 0.0472 [0.0924] [0.3415] [0.3524] [0.3882] [0.1394] Program length over 2 years
Recycle 0 - 10% 0.0027 -0.0440 [0.0310] [-.1016] Recycle 11 - 95% 0.0053 0.0059 [0.0318] [0.0258]
Recycle over 95% -0.0080 0.0381 Note: Numbers in brackets convert the marginal effect into a percentage change from the average intensity of recycling effort observed in the sample. For total materials curbside, the marginal effect is calculated assuming a one-unit increase in the total number of materials recycled curbside. For binary indicator variables, marginal effects are calculated by solving the model once with the significant indicator variable of interest set at one and all other variables set at their mean value, solving again with the indicator variable of interest set at zero and all other variables set at their means, and then taking the difference. The marginal effect for drop-off (curbside) is calculated with the curbside (drop-off) dummy variable set at zero. The "drop-off to curbside" marginal effect is defined as the difference between the marginal effect of curbside and the marginal effect of drop-off. For program length, the marginal effect gives the difference between having a program in place between one and two years and having a program for more than two years. The sum of marginal effects may not equal zero due to rounding.




Full Text

PAGE 1

ESSAYS IN THE ECONOMICS OF SOLID WASTE MANAGEMENT AND RECYCLING By SALVADOR A. MARTINEZ A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA EM PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2004

PAGE 2

This dissertation is dedicated to my parents, Manuel and Laurell, and my siblings, Manjo and Felice, for their encouragement, support, and inspiration.

PAGE 3

ACKNOWLEDGMENTS I thank my undergraduate economics professors at Weber State University for stimulating my interest in economics early in my studies. Also, I thank the economics faculty at the University of Florida for imparting knowledge of tools and methods for undertaking economic research. I would like to thank researchers at the Public Utility Research Center (UF), Bureau of Economic and Business Research (UF), Resources for the Future, and the University of Florida for providing opportunities for undertaking economic research and policy analysis. I thank my supervisory committee members Steven Slutsky and David Figlio for their keen insights into public economics and guidance in research. I am grateful to have Donna Lee as an outside member on my committee. I give special thanks to my dissertation chair, Larry Kenny, for his insights into economic research, his financial support toward of my dissertation expenses, and his patience and understanding.

PAGE 4

LIST OF TABLES Table Page 21 . Summary statistics for recycling policy regressions 26 2-2. States adopting requirement for local government recycling programs 26 2-3. Total landfill material bans OLS and fixed effects regressions 27 2-4. Recycling grants and loans probit regressions 28 25. Recycling tax incentives probit regressions 29 31 . Summary statistics for tipping fee regressions 50 3-2. Tipping fee regressions using small-market definition 5 1 3-3. Tipping fee regressions using medium-market definition 52 34. Tipping fee regressions using large-market definition 53 41 . Previous studies on effects of unit pricing and recycling programs on effort 84 4-2. Metropolitan Statistical Areas sampled 85 4-3. Unit pricing programs 86 4-4. Summary statistics for independent variables in recycling logit regressions 88 4-5. Proportions of materials recycled 89 4-6. Recycling participation ordered logits 90 4-7. Marginal effects of significant policy variables from logits 91 VI

PAGE 5

LIST OF FIGURES Figure 21 . T otal number of landfill material bans 2-2. States with recycling grants or loans ... 23. States with recycling tax incentives 34. Mean tipping fees by state 41 . Distribution of unit prices Page ....23 ....24 ....25 ....49 ....87 Vll

PAGE 6

TABLE OF CONTENTS Page ACKNOWLEDGMENTS iii LIST OF TABLES vi LIST OF FIGURES vii ABSTRACT viii CHAPTER 1 INTRODUCTION 1 2 ADOPTION OF STATE SOLIDWASTE AND RECYCLING POLICIES 4 Introduction 4 Solid Waste and Recycling Policies 7 Methodology and Explanatory Variables 10 Results 17 Conclusion 21 3 DETERMINANTS OF LANDFILL TIPPING FEES Introduction 31 Landfill Operations and Costs under Regulation 33 Methodology and Estimation 34 Results 40 Conclusion 47 4 DETERMINANTS OF HOUSEHOLD RECYCLING: A MATERIAL-SPECIFIC ANALYSIS OF RECYCLING PROGRAM FEATURES AND UNIT PRICING Introduction 54 Prior Research and a Conceptual Framework 57 Data Description 60 Model Specification 65 Results 67 Recycling Program Features 69 IV

PAGE 7

Unit Pricing Policy Variables 74 Socioeconomic Factors 75 Data Limitations 77 Conclusion and Policy Implications 79 5 CONCLUSION 92 REFERENCE LIST 96 BIOGRAPHICAL SKETCH 101 V

PAGE 8

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ESSAYS IN THE ECONOMICS OF SOLID WASTE MANAGEMENT AND RECYCLING By Salvador A. Martinez August 2004 Chair: Lawrence Kenny Major Department: Economics My study examined the determinants of the implementation of state solid waste and recycling policies (landfill material bans, “market development” initiatives, recycling program availability) and how solid waste management and recycling policies and other factors influence the prices set by landfills to accept solid waste and the percentage of various materials recycled by households. I examined state adoption of landfill material bans, recycling grants and loans, recycling tax incentives, and requirements for local government to offer recycling using state panel data from 1988 to 1999 when the U.S. Environmental Protection Agency implemented stricter guidelines on landfills. Less Republican control of state government, greater environmental organization membership, and a greater percentage of the state population in metro areas positively affected the adoption of these pro-environmental policies. A national panel dataset on landfills is used (supplemented with other data) to examine the effects of government regulations, cost factors, and competition on tipping viii

PAGE 9

fees while investigating economies-of-scale effects. Implementing landfill material bans for automobile tires and motor oil reduces tipping fees; automobile battery bans, white good bans, and yard waste bans raise tipping fees. Hence, the cost savings from avoided treatment of the different materials and monitoring costs due to implementing the bans are different. Landfills in coastal counties and areas with higher leachate costs experience higher tipping fees. Surprisingly, increased competition does not result in lower tipping fees. Finally, a large percentage of the landfills could continue to accept more solid waste per year and decrease costs. My study then uses a household-level dataset (covering 20 MSAs in the U.S.) to examine the impact of recycling and unit pricing program availability on the percentage recycled of five different materials: glass bottles, aluminum, plastic bottles, newspaper, and yard waste. The availability of curbside recycling has a stronger effect than drop-off recycling across all materials. Effects from mandatory recycling on the percentage recycled is inconclusive. Also, there is no evidence that unit pricing (price of household waste disposal varying with amount discarded) increases the percentage of materials recycled by households. IX

PAGE 10

CHAPTER 1 INTRODUCTION Chapter 2 uses a variety of econometric models to examine the determinants of the following state solid waste management policies: landfill material bans, state provision of recycling grants and loans, state provision of recycling tax incentives, and state requirements for local governments to implement recycling programs. Using statelevel data from 1988 to 1999, 1 use a set of covariates to explain the adoption of these various pro-environmental policies in a median voter framework. Results indicate that Republican control of state government results in less-strict pro-environmental standards being implemented. There is evidence that states with more pro-environmental preferences (proxied by the percentage of the state population belonging to the National Audubon society) implement stricter solid waste and recycling policy. In addition, states with older populations provide more landfill bans, and have a higher probability of having recycling grants or loans. As a consistently strong voting group, retirees could be influential in shifting the position of the median voter. Finally, higher concentrations of the state population in metropolitan areas increase the probability of states implementing requirements for local governments to offer recycling programs. Transaction costs are lower for grassroots environmental lobbying in more densely populated areas, facilitating support for local recycling policies such as local recycling availability. Chapter 3 examines the impacts of three factors on tipping fees (the prices set by U.S. solid waste landfills to accept waste) while looking at scale effects of operation. Regulatory controls, operating costs, and competition affect tipping fees in different 1

PAGE 11

2 ways. In terms of regulatory controls, landfill bans put in place by state government restrict certain materials from being deposited in landfills. Having automobile tire bans and motor oil bans in effect reduces tipping fees, while automobile battery bans, white good bans, and yard waste bans result in higher tipping fees. This suggests the tire and motor-oil bans result in net cost savings for the landfill where the labor costs of monitoring the bans are lower than the costs of handling toxic materials and dealing with increased risks (such as tire fires and vermin problems). Landfills located in areas with greater precipitation must use more resources in controlling leachate and protecting groundwater sources. A major portion of long-run operating costs, including leachate collection (the handling and disposal of sludge-type material from the landfill) are proxied by weather variables. Increasing the amount of precipitation increases the tipping fees set by landfills. Other operating costs (such as labor wages) increase tipping fees. Also, an increase in the shadow cost of expanding landfills, agricultural land prices, results in higher tipping fees. Regarding scale effects, there is evidence that a substantial percentage of landfills could accept greater amounts of waste and experience economies of scale. Chapter 4 examines the percent recycled of five materials (glass bottles, plastic bottles, aluminum, newspaper, and yard waste) by households using a dataset covering 20 metropolitan statistical areas. The households are located in communities with various recycling program regimes and unit pricing programs. Access to curbside recycling has a significant and positive effect on the percentage recycled of all five materials. The length of a recycling programÂ’s existence has a significant effect on two of the recyelable materials. There is no significant evidence that mandatory recycling increases the

PAGE 12

3 percentage recycled for any of the materials. Finally, the level of the marginal cost of disposal facing the households (the unit price) is insignificant in the regressions.

PAGE 13

CHAPTER 2 ADOPTION OF STATE SOLID WASTE AND RECYCLING POLICIES Introduction Many state governments are implementing solid-waste management and recycling policies (to increase recycling activity, and reduce the amount of municipal solid waste entering landfills) to address one of AmericaÂ’s continuing enviroiunental challenges, municipal solidwaste disposal.' The concerns surrounding solidwaste disposal have escalated as several landfills are reaching capacity and many older landfills have leaked hazardous waste into groundwater (Menell, 1990). Building new landfills is not a popular option among various stakeholders, especially those citizens residing in areas near proposed landfills.^ A variety of factors may influence the state adoption of solid-waste management and recycling policies. Citizens who value recycling^ may be represented by environmental organizations that generate political action with lobbying, grassroots pressure, public advocacy, and research. State legislators are concerned about their state becoming a regional dumping ground for solid waste imported from other states. The aim of my study is to analyze the factors influencing state governments in adopting solid-waste management and recycling policies when the Environmental Protection Agency (EPA) implemented stricter landfill standards. Landfill material bans, ' Kinnaman and Fullerton (The economics of residential solid waste management, NBER working paper, 1999) review literature on residential solidwaste management. ^ Nelson et al. (1992). ^ Aadland and Caplan (1999) and Jakus et al. (1996) discuss residential valuation of curbside and drop-off recycling, respectively. 4

PAGE 14

5 state grants and recycling loans, and state requirements for local government to develop recycling are the solid-waste management and recycling policies investigated in the paper using econometric analysis with data from 1988 to 1999. I use various explanatory variables to capture political and socioeconomic forces within the states that affect the adoption of pro-environmental solid-waste management and recycling policies. The Resource Conservation and Recovery Act (RCRA) has largely left solid-waste management and recycling policy to be determined by state governments. Surprisingly, the economics literature investigating the adoption of solid-waste and recycling policies at the state level is non-existent. To my best knowledge, the only existing empirical research on solid-waste management and recycling policy adoption is at the community level. The stateÂ’s reasons for implementing solid-waste and recycling policies are quite different than those at the local level. Determining state policies can be considered more important because states set the rules, and local governments operate within those guidelines. Unlike local governments developing solid-waste and recycling policy at the local level without state guidelines, local governments in states with solid-waste and recycling policies are constrained in their decision making. Any losses or gains resulting from operating recycling programs are absorbed by the local governments. However, the programs can easily become more costly than anticipated, as there is no cost sharing between local governments and state governments."^ Also, states are able to solve externality problems among communities by implementing solid-waste and recycling policies. '' Kinnaman (2000) looks at recycling revenues and costs within a single community. He found the local economy paid an average of $102.45 per ton to recycle the material, an amount just over $55 per ton more than the cost of disposing in the landfill.

PAGE 15

6 Feiock and West (1993), Tawil (On the politieal economy of municipal curbside programs: Evidence from Massachusetts, working paper, 1995), Callan and Thomas (1999), and Mrozek (2000) examine the adoption of solidwaste management and recycling policies at the local level. Except for Feiock and West (1993), who use a crosssectional dataset, the studies use data from communities within a particular state. In addition, only Mrozek (2000) uses data from more than one time period. Unlike theirs, my study uses a national dataset from 1988-1999, thus considering variation across states and over time. For example, the panel data set allows me to estimate state responses to stricter landfill operating guidelines imposed by the EPA. Previous authors examine only one type of solid-waste or recycling policy (such as adoption of some type of recycling program, curbside recycling, and unit-pricing for residential waste disposal). I examine state landfill material bans, state provision of recycling grants or loans, state provision of recycling tax incentives, and requirements for local governments to implement recycling programs; and thus cover two of the four major categories of the EPAÂ’s agenda of integrated solid-waste management (source reduction, landfilling, combustion, and recycling). Finally, my analysis includes a richer set of covariates to explain the adoption of solid-waste and recycling policies. The impact of regional spillover effects from solid waste are examined. With states unable to prevent interstate shipments of solid waste, some state governments may be implementing pro-environmental solid-waste and recycling policies to limit the growth and size of their home landfills and deter imports of solid waste. I also am able to include proxies for the size and strength of environmental groups (League of Conservation Voter scores averaged for each state, and membership in

PAGE 16

7 the National Audubon Society); and for political party control of state government (including governor office and state legislature). Results from including a measure of political party control suggest that Republican control of state government results in less-strict standards, which is consistent with Democrats supporting greater environmental protection then Republicans. Results from including a measure of National Audubon membership confirm that states with a larger environmental movement implement stricter solid-waste and recycling policies. Other results indicate that states with older populations provide more landfill bans, and are more likely to have recycling grants or loans. This is consistent with retirees having a greater desire to leave a legacy, in terms of environmental protection. Higher concentrations of the state population in metropolitan areas increase the probability of states implementing requirements for local governments to offer recycling programs, thus supporting the hypothesis that metropolitan areas are conducive to pro-environmental lobbying. Solid Waste and Recycling Policies The Resource Conservation and Recovery Act of 1976 was amended by the Hazardous and Solid Waste Amendments of 1984 (HSWA), directing the Environmental Protection Agency to develop a set of minimum criteria for solid-waste management facilities that receive either household hazardous waste or small quantities of exempt hazardous waste. The criteria were not put forth until October 1991. Both existing and new municipal solid-waste landfills were affected by the rules, which became effective in October 1993. The criteria included minimum standards regarding location restrictions, operations, design, groundwater monitoring and corrective action, closure and postclosure care, and financial assurance. In addition, the EPA required states to develop

PAGE 17

8 and implement permit programs to ensure compliance no later than April 1993. This increase in federal intervention in solid-waste management policy brought attention to the choice to implement other policies that make it easier to comply with federal regulations. The first solid-waste management policy to consider is the landfill material ban. A growing concern over landfill expansion and the desire to increase recycling opportunities (coupled with the new changes to RCRA) has supposedly led state governments to pass legislation that bans particular materials from the landfill. The landfill material bans I use in the econometric analysis are motor oil, vehicle batteries, white goods, tires, and yard waste. Environmental concern regarding vehicle batteries and motor oil is directly related to the contaminants from these materials escaping from the landfill area to the surrounding areas, especially to those areas with water sources and residential populations. White goods and tires are bulky items that take up larger amounts of space in the landfill, and can be “remanufactured” into other materials. White goods, a class of appliances including stoves, refrigerators, and clothes dryers, may contain components in their electrical wiring that can be harmful if leaked into water supplies. Finally, I consider landfill material bans on yard waste. Yard waste is estimated to be the second largest contributor to national municipal solid-waste generation, behind paper and paperboard (Kreith, 1994). Since the landfill material bans collectively aim at extending the capacity of landfills and lowering health risks, the landfill material bans are measured as the number of these five materials that are banned (TOTAL BANS). In 1988, about 96% of the states had passed landfill bans for two or fewer of the five mentioned materials. This figure has slowly decreased to 37% in 1999. The mean number of TOTAL BANS each year for all

PAGE 18

9 states has steadily increased over time but only nine states had passed landfill material bans for all of the materials by 1999. Vehicle batteries are the most popular material to be banned, with over 80% of the states having the material banned from 1993 to 1999. Fewer states passed bans for white goods (only 3 1% of the states had a white good landfill ban in 1999). The greatest push for landfill bans seems to be concentrated in the Midwest and on the western and eastern coasts. Some of the states in the Rocky Mountains have not aggressively implemented landfill material bans. Figure 2-1 maps state values for TOTAL BANS in 1992 and 1999. Next, I consider the adoption of state recycling grants, loans, and tax incentives. Several states provide grants and loans to both local governments and businesses, to implement recycling programs and stimulate demand for recycled products. Some grants are given on a competitive basis, while others are “block” grants. States like New Jersey and Pennsylvania give grants to local jurisdictions, as a reward tied directly to the amount of a material recovered from the solid-waste stream. Tax incentives include such items as tax breaks or exemptions on equipment or materials used in manufacturing final products using post-consumer recycled content or on machinery used in recycling facilities. Collectively, recycling grants, loans, and tax incentives are referred to as recycling “market development” initiatives in the recycling industry. The dependent variables constructed from information on these initiatives include GRANTLOAN and TAX fNCNTV, which are used only for 1992 and 1999, because of data availability. The first variable takes on the value 0 or 1; a positive value indicates that the state offered recycling grants or loans in the particular year. Likewise, the variable TAX INCNTV takes on the value 0 or 1, reflecting whether a state offered

PAGE 19

10 recycling tax incentives in the particular year. Half the states provided no types of grants, loans, or tax incentives in 1992. This figure decreased to six states in 1999. Some states, that provided market development initiatives in 1992, did not do so in 1999. Figures 2-2 and 2-3 show states with positive values for GRANTLOAN and TAX INCNTV in 1992 and 1999. The final state recycling initiative 1 consider is the adoption of state legislation requiring local government units to develop recycling programs (LOCAL RECYCLING). The local government units may be counties, cities and counties, cities, or solid-waste management districts. Adoption of these policies is concentrated in the western states, and a major part of the east coast. Unlike landfill bans and market development initiatives, states are not reversing the legislation on the development of these local recycling programs. So I estimate hazard regressions, to examine the determinants affecting adoption of these requirements, using the same covariates as used in other regressions. Data for TOTAL BANS, GRANTLOAN, TAX INCNTV, and the adoption of state legislation requiring local government units to develop recycling programs come from Glenn (1992a, 1992b, 1998a, 1998b, 1999), Goldstein (1997a, 1997b, 2000a), Goldstein and Madtes (2000), Kreith (1994), Raymond Communieations (State recyeling laws update, 1999), Steuteville (1994a, 1994b, 1995a, 1995b, 1996a, 1996b), Steuteville and Goldstein (1993), and Steuteville et al. (1993). Methodology and Explanatory Variables The decision to adopt state solid-waste and recycling policies can be examined in a median voter framework. State solid-waste and recycling policies determined by the median voter, where the number of voters preferring a stricter amount of policy (P) is

PAGE 20

11 equal to the number of voters preferring a weaker policy. The policy P* is preferred by the median voter to any other amount of solid-waste and recycling policy. With state solid-waste and recycling policies being implemented by elected officials, households vote for the candidate yielding higher utility levels. Citizens vote for state legislators within local voting districts across the state, and over time, in elections. Thus, voters select candidates (in various elections) whose choice of P is equal to their own preferred level of P. The amount of P implemented by state policy makers is expected to represent the amount of P desired by the median of the various district median voters. To explain the choice of solid-waste and recycling policies, a variety of socio-economic and political variables are used in the econometric analysis for the three classes of solid-waste and recycling policies. Since landfills pose more of a health risk in densely populated areas, landfill regulation should be more likely in densely populated areas. The population density variable POP DENSITY is the number of persons per square land mile (in thousands) in the state. The variable is constructed from data on state land and water area size in 1987 defined by the U.S. Geological Survey and U.S. Census Bureau data on state population from 1988-1999. States with higher POP DENSITY are expected to be more likely to implement pro-recycling initiatives.^ Landfills in areas with less expensive land can expand landfill space at a relatively low cost. To counteract the low land prices, policy makers may be more inclined to pass landfill bans and other pro-recycling policies to counteract potential landfill growth. Using data from the U.S. Agricultural Censuses in 1992 and 1997, AG Â’ Mrozek (2000) expects higher urban density to explain economies of scale in the local government decision to implement curbside recycling. His results are mixed.

PAGE 21

12 LAND equals the market value of land and buildings per acre. The coefficient for AG LAND is expected to be negative; that is, the higher price of land should result in fewer pro-recycling policies. Previous economic research investigated whether environmental goods are normal. Evidence supports the notion that individuals with higher socioeconomic status are more likely to support pro-environmental activities. Using per capita income data from the Bureau of Economic Analysis and the personal consumption price expenditures implicit price deflator, I calculate real personal income (in thousands) per capita (INCOME) from 1988 to 1999 to determine whether solid-waste and recycling policies are “normal.” The expected sign for the INCOME coefficient is positive, but Mrozek (2000) finds no effect, Callan and Thomas (1999) find a negative effect, and Feiock and West (1993) find a positive effect. Previous studies investigating household-level recycling behavior use age as a characteristic to explain recycling quantities. Using annual estimates from the U.S. Census Bureau, AGE65 is defined as the percentage of the state population that is age 65 or over for years 1988-1999. Illinois, Florida, and Pennsylvania had the highest percentage of residents age 65 and older in 1999. Small (Preserving family lands, land and people. Trust for Public Land, San Francisco, CA, 1996) notes that several elderly are holding onto land, as compared to selling, to prevent development. Older individuals could be supporting pro-environmental protection in recycling to leave a legacy.^ With older populations likely to be more concerned regarding environmental protection, the ^ Kahn (2000) investigates demographic change and the demand for environmental regulation using demographic covariates in California environmental ballots, local government environmental expenditures across the U.S., and Congressional voting on the environmental law regressions. Results on the variable percentage of the population 65 years or older are mixed.

PAGE 22

13 coefficient for AGE65 is expected to be positive/ Also, retirees are attracted to amenity rich areas with scenic beauty. Thus, AGE65 is also a measure of the stock of natural amenities. On the national, level policy makers have debated whether states should be able to restrict interstate shipments of solid waste. State politicians are concerned about their states becoming the dumping ground in their region. Thus legislators may be passing pro-recycling initiatives to reduce the growth of waste products generated in the state, and also may be legislating to deter imports of solid waste. The greatest amount of interstate waste shipments occur in the Midwest and East. For example, Feliciano and Worth (1997 summary of Indiana solid waste facility data, IDEM Office of Solid and Hazardous Waste Management) report that 27% of annual waste disposal in Indiana came from out of state. Most of the out-of-state waste Indiana receives is from Illinois. Data on municipal solid-waste generation at the state level from BioCycle ‘s o annual survey and state population data from the U.S. Census Bureau are used to construct the explanatory variable WASTE, which is the weighted average waste generation per capita of the states contiguous to the particular state for years 1988-1999. The more waste generated by neighbors, the more likely the home state will implement the solid-waste management and recycling initiatives. To capture the size and strength of environmental groups in the state, I include three different measures. First, using membership data from the Audubon Society and ’ Callan and Thomas (1999) use a community’s median age and its square as covariates explaining the adoption of unit-pricing, a program where households pay for each unit of trash disposal. They suggest that waste levels are low at the ends of the life cycle and the need for unit-pricing follows waste generation over the life cycle. * Glenn (1992a, 1992b, 1998a, 1998b, 1999), Goldstein (1997a, 1997b, 2000a), Goldstein and Madtes (2000), Steuteville (1994a, 1994b, 1995a, 1995b, 1996a, 1996b), Steuteville and Goldstein (1993), and Steuteville et al. (1993).

PAGE 23

14 U.S. Census Bureau state population estimates, I calculate the percentage of the state population belonging to the National Audubon Society from 1988-1999. I was not able to obtain complete membership data from the Sierra Club or National Wildlife Federation. In states with a higher percentage of members in environmental groups, it is expected the median voter would prefer a stricter standard. In addition, legislators are expected to represent the preferences of a median voter in a median voter model. The League of Conservation VotersÂ’ (LCV) average rating for each stateÂ’s U.S. Senators and Representatives for 1988-1999 is another measure of the median voter preferences. The LCV rating ranges from 0 to 100, and measures support for environmental initiatives (100 implies full environmental support). Higher Audubon membership percentages and LCV scores should result in more states implementing pro-recycling initiatives and landfill material bans. Transaction costs of organizing and promoting pro-environmental legislation should be reduced when individuals and groups are in close proximity. With reduced transaction costs and pro-environmental members finding greater ease in disseminating information, the percentage of the populace with pro-environmental ideology would be expected to increase. Hence, the position of the median voter would change, resulting in a stricter P*. I use the percentage of the state population living in metropolitan areas (METRO POP) to capture this occurrence.^ The U.S. Statistical Abstract contains values available in even years from 1988 to 1998; interpolation is used for the odd years. A Â’ Callan and Thomas (1999) use a rural/non-rural indicator in their regression explaining the adoption of unit-pricing and find a negative and statistically significant effect using an economies of scale hypothesis for its inclusion.

PAGE 24

15 higher percentage of the state population living in metropolitan areas is expected to result in more pro-environmental solid-waste management and recycling legislation. Finally, I define a political variable REPUBLICAN CONTROL from the Council of State Governments (1989, 1991, 1993, 1995, 1997, 1999, 2001), which takes on the value 0 if both houses of the state legislature and the governorÂ’s position are controlled by Democrats, the value 2 if Republicans are in control of both houses and the governorship, and the value 1 if the political power in the state is shared by both parties for 1988-1999. The hypothesis is that state government controlled by the Republicans should be less likely to implement environmental protection (or solid-waste management and recycling initiatives) given the ideological positions of the party. Their ideology is anti-regulation and thus anti-environment. However, interpretation can be complex, as Republicans represent higher-income individuals who favor stricter standards. To capture the effects of the changes in RCRA, dummy indicators are included for two time periods; 1991-1993 and 1994-1999. The default time frame is 1988-1990. These three breakdowns capture 1) the period when the EPA was formulating its requirements for landfills; 2) the transition period when states had to develop and implement permit programs, with landfills complying with EPA regulations; and 3) the period after the transition. The variables are collected for all 50 states and the District of Columbia with some exceptions. AG LAND is not available for DC. With Alaska and Hawaii not part of the continental U.S., the variable WASTE is not calculated for those states. REPUBLICAN CONTROL is not available for DC or Nebraska. So a maximum of 49 observations is possible using WASTE or using REPUBLICAN CONTROL. Table 2-1

PAGE 25

16 gives summary statistics for these variables. Also, Table 2-2 list the states adopting local recycling requirements. The econometric analysis uses a variety of models to examine the robustness of the effects of the independent variables in explaining the policies. First, Table 2-3 shows results from regressing TOTAL BANS on the previously described explanatory variables using Ordinary Least Squares (OLS) and Fixed Effects (state fixed effects) regressions for the period 1988 to 1999 with robust standard errors at the state level. Specification 1 is the base specification. Specification 2 adds the political variables REP CONTROL, AUDUBON, and LCV to the base specification, and drops WASTE. Finally, Specification 3 adds the WASTE variable to Specification 2. Two different tables show market-development initiatives. Table 2-4 gives probit regression results, showing whether the state provided recycling loans or grants for two time periods: 1992, and 1992 & 1999. The variable AG LAND is included in the 1992 regressions, using the same specification framework as for the landfill material bans. In a similar fashion. Table 2-5 presents the probit regressions for the dependent variable TAX INCNTV for years 1992 and 1992 & 1999. Finally, Table 2-6 shows results from Weibull proportional hazard regressions (investigating the adoption of a law requiring local government to offer recycling with the same specification framework used for TOTAL BANS). Overall, the fit of the regressions for the various solid-waste and recycling policies is reasonable, with many hypotheses supported. Low R-square values in the OLS regressions indicate that more variation of TOTAL BANS can be explained by variables not in the regression. Relative to the landfill material ban regressions, probits for GRANTLOAN and TAX INCNTV do not perform as well. The overall fit is good in

PAGE 26

17 two of the three Weibull regressions explaining the adoption of a requirement for local government to implement recycling. Results AGE65 has a significantly positive impact on the number of landfill material bans, and on provision of recycling grants; but is unrelated to use of tax incentives, and requiring local governments to implement recycling. The positive effect is consistent with the hypotheses of older populations considering environmental protection as a legacy, and areas with a greater percentage of retirees being rich in natural amenities and thereby warranting greater environmental protection. An increasing percentage of individuals over 65 (coupled with the fact that a large percentage of this population is voting), older individuals could be instrumental in shifting the position of the median voter. A one standard-deviation rise in AGE65 results in a .12 to .50 increase in the total amount of landfill material bans, where significant. The estimated coefficients for AGE65 are relatively the same in Table 2-4 under various specifications. In Column 3-b, the marginal effect of AGE65 implies that increasing the percentage of the state population age 65 and over by 1% results in a 4% increase in the probability of providing recycling grants or loans. AGE65 is not statistically significant in Table 2-5 or 2-6. Age variables covering the other age groups were tested in the regressions. F-tests and Likelihood Ratio tests did not indicate that the group of other age classifications should be included. The estimated coefficients on METRO POP have the hypothesized sign in all but two regressions, suggesting that metropolitan areas enable grassroots environmental groups to easily organize and lobby for pro-recycling policies. Also, the cost of recycling may be lower in states with more people living in metro areas. The coefficients are

PAGE 27

18 significant in just one column in Table 2-4 and three columns in Table 2-5. The marginal effect for METRO POP in Column 2-b in Table 2-4 is .01 indicating that a 1% increase in the percentage of the state population living in metropolitan areas results in a .01 increase in the probability of providing recycling tax incentives. METRO POP is significantly positive in the first two specifications in Table 2-6. These specifications imply that in each year, a state with a higher percentage of the population living in metropolitan areas has a higher probability of adopting a requirement for local governments to offer recycling. Where the coeffieients for POP DENSITY are significant in the regressions for landfill material bans, grants and loans, and tax incentives, the coefficients are negative and donÂ’t provide support for the initial hypothesis that poliey makers in regions with higher population density would implement these pro-environmental policies to combat the growth of landfills. The result for POP DENSITY is only positive and significant in the first specification in Table 2-6, for the recycling regressions. The coefficients for INCOME are positive and significant in the fixed effects regressions in Table 2-3 (for landfill material bans and Table 2-4 for a eouple of specifications in the short model); these results are consistent with the hypothesis on environmental protection being a normal good. A one standard-deviation rise in per capita income results in an increase of .38 to .43 landfill material bans. It could be that fNCOME is picking up different effects in the various dependent variables. There is a time cost with recycling; so as income increases, the opportunity costs of devoting time to recycling increase. The coeffieients on INCOME are negative but not significant in

PAGE 28

19 the recycling regressions. This suggests that the income and time cost effects roughly cancel each other out. Waste generation in neighboring states is expected to result in stricter solid-waste and recycling policies. Coefficients on WASTE are significant and positive only in the OLS regressions in Table 2-3, with a one standard-deviation increase in WASTE resulting in .24 and .23 increases in landfill material bans, for columns 1-a and 3-a, respectively. It could be that the state fixed effects are collinear with WASTE. AG LAND was only available in years 1992 and 1997. The variable was included in regressions for TOTAL BANS using the two years of data, but AG LAND was not statistically significant in a variety of specifications, which are not reported. When including the variable in the probit regressions, the estimated coefficient is negative in all three specifications in Tables 2-4 and 2-5, and is significant in two of the three specifications in Table 2-6. The negative coefficient is consistent with the hypothesis that policy makers are less likely to implement the recycling grants and loans as the price of agricultural land increases, because it is more costly for landfills to expand. The incentive to stimulate recycling among local governments and firms is decreased as the shadow price on waste disposal increases. With the Republican party favoritism toward business and opposition to big government, there should exist significant differences between Republicans and Democrats regarding environmental concerns. The first political variable, REP CONTROL, is a measure of both partisanship and party ideology. Recall it takes the values of 0, 1, and 2 for Democratic control of both state legislative bodies and the governorship, mixed control, and Republican control of both state legislative bodies and

PAGE 29

20 the governorship, respectively. The estimated coefficient in negative as predicted, and is and significant in all but one table. The coefficient is negative in all specifications in Table 2-3, meaning more Republican control of state government results in less landfill material bans being passed. Changing political control of the governorship and the legislative bodies from Democratic to Republican control results in a decrease of from .48 to .5 1 landfill material bans. The sign results for REP CONTROL are reversed between Tables 2-4 and 2-5, positive for grants or loans and negative for tax incentives. The former result is unexpected, but the grants and loans may be interpreted as subsidies for business and hence pro-Republican. Next, AUDUBON is the percentage of the state population having membership in the National Audubon Society. The coefficients are not significant for the TOTAL BANS regressions or the recycling requirement hazards. However, the sign is significant in one of four specifications in the recycling grants and loans probits and is significant in three out of the four tax incentives probits in Table 2-5. Part of this difference between Table 2-3 and Tables 2-4 and 2-5 may be due to there no being enough year to year variation in the data. So there is some slight evidence that environmental activism may result in states adopting pro-environmental solid-waste and recycling policies. As a result of electoral pressures, the environmental sentiment of the stateÂ’s Congressional delegation is expected to mirror that of the electorate. Thus, the average League of Conservation Voters (LCV) score for the stateÂ’s Representatives and Senators is used as another measure of environmental sentiment in the state. The hypothesis that states with more pro-environmental sentiment are more likely to adopt pro-environmental

PAGE 30

21 policies is supported in three of the four tables. Coefficients are positive and significant in Tables 2-3, 2-4, and 2-6, but are negative and significant in Table 2-5. Environmentalists may be more likely to support recycling grant and loan programs that provide local benefits than to support tax incentives which result in more “free financing” opportunities for businesses. There is evidence more solid-waste and recycling policies are passed as the EPA implemented its landfill regulations. The dummy indicators for 1991-1993 and 19941999 are positive and statistically significant in all specifications in Table 2-3. Compared to 1988-1990, there were 1.31-1.46 more landfill material bans in 1991-1993 and 1.612.1 more in 1994-1999. When looking the at the 1994-1999 period indicator in the tax incentives and recycling grant and loan regressions, the probability of adoption was higher in 1999 than in the transition period. Conclusion This study is an investigation into the determinants of adoption of solid-waste management and recycling policies by states. It extends the literature by using state panel data covering several solid-waste management and recycling policies accounting for half of the EPA’s platform for integrated solid-waste management, using a richer set of covariates (such as more political variables and a measure of interstate solid waste spillovers) and explaining state policies instead of local policies. I find that states with higher percentages of the population 65 or older tend to adopt more landfill bans and have a higher probability of having recycling grants or loans. Also, there is evidence that states with greater concentrations of population in metropolitan areas provide recycling tax incentives, and are more likely to adopt a requirement for the development of recycling programs at the local level probably indicating lower costs of implementing

PAGE 31

22 recycling. When the waste generation around a home state is high, there is some evidence a state will implement more landfill material bans. Finally, the results indicate that Republican control results in less strict standards. The party effects are greater than those for any other variable. State membership in environmental groups or proenvironmental records of the stateÂ’s Congressional delegation reflect the environmental preferences of the stateÂ’s populace. I find that states with a more pro-environmental populace are more likely to adopt pro-environmental policies. I initially stated the optimal amount of state solid-waste and recycling policies, P*, is determined by the median voter and have included covariates to try and explain the selected pro-environmental policies. When examining the results from the different state policies, it seems the results of the landfill material bans regressions and the Weibull regressions on the requirement for local recycling have the most significant results consistent with initial hypotheses. One possible explanation for the more significant results in these regressions compared to other regressions in the tables is that the benefits from reduced disposal of potentially harmful materials and opportunities to recycle are likely to directly affect citizens more than the provision of recycling grants, loans, and tax incentives. A more comprehensive data set containing additional years of data for GRANTLOAN and TAX INCNTV, including actual amounts of grants and loans provided, would allow estimation techniques that might provide more insight into the determinants behind the adoption of state solid-waste management and recycling policies.

PAGE 32

23 1992 Landfill Material Bans (12 1999 Landfill Material Bans ( 10 ) (13 ( 10 ) 2-1. Total number of landfill material bans A) In 1992. B) In 1999.

PAGE 33

24 1 992 Recycling Grants or Loans I Recycling Grants or Loans (13) Q No Recycling Grants or Loans (38) A Figure 2-2. States with recycling grants or loans. A) In 1992. B) In 1999.

PAGE 34

25 Figure 2-3. States with recycling tax incentives. A) In 1992. B) In 1999.

PAGE 35

26 Table 2-1 . Summary statistics for recycling policy regressions Variables States Mean Standard Deviation Minimum Maximum Time Period TOTAL BANS 51 2.16 1.70 0 5 1988-99 GRANTLOAN 51 0.49 0.50 0 1 1992, 1999 TAX INCNTV 51 0.46 0.50 0 1 1992, 1999 AGE65 51 12.86 2.59 3.69 25.98 1988-99 METRO POP 51 66.92 21.56 20.04 100 1988-99 POP DENSITY 51 320.61 1146.19 0.92 9136 1988-99 INCOME 51 23.12 3.96 14.91 37.76 1988-99 WASTE 49 1.16 0.27 0.24 2.70 1988-99 AG LAND 50 1615.94 1543.64 184.02 7558.80 1992, 1997 REP CONTROL 49 0.91 0.63 0 2 1988-99 AUDUBON 51 0.19 0.10 0.07 1.05 1988-99 LCV 50 46.80 21.90 0 94.42 1988-99 Table 2-2. States adopting requirement for local government recycling programs 1988 1989 1990 1991 1992 1993-1999 MD, NY, PA MN, NC, VA, WA, AL, CA, DC AZ, CT AR, NV, OR, SC, WV NE, SD none Note: RI adopted in 1986 while NJ and VT adopted in 1987.

PAGE 36

27 Table 2-3. Total landfill material bans OLS and fixed effects regressions Independent Variables OLS Fixed Effects OLS Fixed Effects OLS Fixed Effects 1-a Lb 2^ 2ib La Lb AGE65 0.1885*** 0.0358 0.1920*** 0.0587* 0.1777*** 0.0453* (0.0483) (0.0255) (0.0563) (0.0322) (0.0530) (0.0239) METRO POP 0.0024 0.0306 0.0090 0.0288 0.0069 0.0339 (0.0089) (0.0307) (0.0083) (0.0301) (0.0082) (0.0299) POP DENSITY -0.0003*** 0.0015*** -0.0024** 0.0050 -0.0023** 0.0001 (0.0001) (0.0005) (0.0010) (0.0150) (0.0011) (0.0136) INCOME 0.0076 0.0964*** 0.0193 0.0977* 0.0223 0.1096** (0.0539) (0.0455) (0.0551) (0.0513) (0.0652) (0.0558) WASTE 0.9014** 0.1380 0.8527* 0.0441 (0.4657) (0.3859) (0.4456) (0.3553) REP CONTROL -0.2088 -0.2535** -0.1500 -0.2422** (0.1313) (0.1189) (0.1324) (0.1200) AUDUBON -0.7905 0.6816 -1.0409 0.4215 (1.5711) (0.6985) (1.5754) (0.6750) LCV 0.0180** 0.0021 0.0196** -0.0016 (0.0074) (0.0063) (0.0080) (0.0058) 1991-1993 1.3224*** 1.3788*** 1.4689*** 1.3090*** 1.4580*** 1.3180*** (0.1950) (0.1977) (0.2124) (0.2124) (0.2189) (0.2199) 1994-1999 1.6741*** 1.6390*** 2.0117*** 1.6623*** 1.8221*** 1.6089*** (0.2433) (0.2225) (0.2364) (0.2325) (0.1931) (0.2267) Constant -2.7087** -4.3704** -2.8062** -4.7070* -3.5279*** -4.1270* (1.0533) (1.9545) (1.2184) (2.4073) (1.2071) (2.2123) Number of states 49 49 49 49 47 47 Number of 588 588 588 588 564 564 observations Adjusted R^ 0.3728 0.7602 0.4081 0.7641 0.2531 0.7891 Root MSE ^r\o/ 1.3358 1.4674 1.4674 0.8095 1.4259 0.7978 * 10% level of significance Robust standard errors are in parentheses. ** 5% level of significance *** 1% level of significance

PAGE 37

28 Table 2-4. Recycling grants and loans probit regressions Independent Variables 1992 1992 & 1999 1992 1992 & 1999 1992 1992 & 1999 1-a Ub 2-a 2^ hA 3ib AGE65 0.0853 0.1130* 0.1521* 0.1382** 0.1240 0.1068* (0.0664) (0.0601) (0.0907) (0.0553) (0.0871) (0.0594) METRO POP -0.0135 -0.0079 0.0389* 0.0136 0.0320 0.0062 (0.0175) (0.0106) (0.0227) (0.0100) (0.0261) (0.0108) POP DENSITY 0.0040 -0.0002* 0.0080 -0.0011 0.0085 -0.0011 (0.0040) (0.0001) (0.0054) (0.0012) (0.0061) (0.0012) INCOME 0.2907*** 0.1237* -0.0322 -0.0153 -0.0038 0.0236 (0.1031) (0.0674) (0.1282) (0.0826) (0.1435) (0.0940) WASTE 0.5401 0.7912 0.9775 0.9714 (1.3157) (0.6727) (1.5060) (0.7088) AG LAND -0.0006 -0.0013* -0.0014* (0.0005) (0.0007) (0.0008) REP CONTROL 0.9869* 0.0016 1.0064 0.0077 (0.5498) (0.2215) (0.6254) (0.2395) AUDUBON 7.3819** 3.9903 6.5406 4.0359 (3.7593) (2.4915) (4.4308) (2.8178) LCV 0.0564 0.0118 0.0615** 0.0134 (0.0241) (0.0100) (0.0257) (0.0117) 1994-1999 0.6621 1.4847*** 1.0989** (0.4273) (0.4767) (0.5554) Constant -7.6193*** -5.1799** -9.1998*** -4.2417*** -10.1064*** -5.3253*** (2.4491) (1.5332) (2.6783) (1.5803) (2.8575) (1.9389) Number of states 48 49 49 49 47 47 Number of obs. 48 98 49 98 47 94 Prob > X ^ 0.0125 0.0000 0.0059 0.0001 0.0032 0.0004 Pseudo R^ 0.2216 0.2318 0.5116 0.2747 0.5065 0.2753 * 10% level of significance Robust standard errors in parentheses. ** 5% level of significance *** 1% level of significance

PAGE 38

29 Table 2-5. Recycling tax incentives probit regressions Independent Variables 1992 1992 & 1999 1992 1992 & 1999 1992 1992 & 1999 1-a Ub 2ia 2^ 3-a 3ib AGE65 0.0187 -0.0154 0.1200 0.0574 0.0694 0.0307 (0.0788) (0.0665) (0.0921) (0.0785) (0.0782) (0.0748) METRO POP 0.0080 0.0227** 0.0209 0.0295*** 0.0090 0.0279** (0.0136) (0.0094) (0.0153) (0.0107) (0.0204) (0.0117) POP DENSITY 0.0021 0.0003 0.0052 0.0011 0.0064 0.0012 (0.0034) (0.0006) (0.0040) (0.0012) (0.0047) (0.0013) INCOME 0.0200 -0.1035* -0.0226 -0.1136 0.1618 -0.0921 (0.0907) (0.0599) (0.1228) (0.0802) (0.1369) (0.0835) WASTE -0.8501 -0.4092 -1.2322 -0.7918 (1.0904) (0.6449) (1.3096) (0.6490) AG LAND -0.0004 -0.0004 -0.0005 (0.0004) (0.0005) (0.0005) REP CONTROL -0.7609** -0.1749 -1.0453** -0.2662 (0.3875) (0.2375) (0.4806) (0.2387) AUDUBON 6.8812* 3.9170 11.2264** 4.5331* (3.7461) (2.4272) (4.5388) (2.7504) LCV -0.0622*** -0.0314*** -0.1018*** -0.0379*** (0.0164) (0.0090) (0.0226) (0.0106) 1994-1999 1.0057*** 1.0921*** 1.2756*** (0.3858) (0.4009) (0.4581) Constant -0.3055 1.1687 -1.1758 -0.0377 -1.3204 1.0469 (2.2509) (1.5194) (-2.3544) (1.8343) (2.4902) (1.8110) Number of states 48 49 49 49 47 47 Number of obs. 48 98 49 98 47 94 Prob > X ^ 0.7662 0.0777 0.0120 0.0003 0.0061 0.0008 Pseudo R^ 0.0399 0.0925 0.2743 0.1647 0.4008 0.1798 * 10% level of significance Robust standard errors in parentheses. ** 5% level of significance *** 1% level of significance

PAGE 39

30 Table 2-6. Local recycycling requirement Weibull proportional hazard regressions Independent Variables I 2 3 AGE65 -0.0024 -0.0445 -0.0395 (0.0745) (0.0914) (0.0820) METRO POP 0.0258* 0.0403* 0.0335 (0.0156) (0.0222) (0.0239) POP DENSITY 0.0002** -0.0009 -0.0022 (0.0001) (0.0022) (0.0020) INCOME -0.0368 -0.1894 -0.0749 (0.1022) (0.1363) (0.1556) WASTE -1.1914 -1.2014 (1.0175) (1.1562) REP CONTROL -0.5085 -0.7972* (0.4264) (0.4322) AUDUBON 1.2071 -1.9413 (3.2458) (4.1530) LCV 0.0327* 0.0442** (0.0168) (0.0185) Constant -2.4755 -2.2665 -2.8530 (1.8772) (2.2512) (2.2992) / In p -0.0807 -0.057 0.0588 (0.1388) (0.1444) (0.1579) Number of states 46 46 44 Number of adoptions 19 17 17 Number of obs. 377 394 370 Prob > X ^ 0 0.1249 0.0184 * 10% level of significance ** 5% level of significance *** 1% level of significance

PAGE 40

CHAPTER 3 DETERMINANTS OF LANDFILL TIPPING FEES Introduction Landfills continue to be the most widely used method of managing solid waste in the United States and are necessary in most any type of solid waste management system. The quantity of waste generated continues to increase while existing capacity of existing landfills continues to decrease. Siting new landfills is becoming more difficult because of increased environmental concerns from citizens and government related to location and operation of landfills. In addition, a variety of federal and state requirements for municipal solid waste landfills have been implemented in the past decade. Tipping fees, the prices set by landfills to accept solid waste, have increased steadily over the past decade. However, tipping fees are not the same across and within states. This analysis examines the determinants of tipping fees at municipal solid waste landfills covering years from 1992 to 1999 using appropriate controls for regulatory policies, competition, and cost factors. To the best of the my knowledge, there is no existing published, empirical research examining the determinants of tipping fees through econometric analysis. Understanding the factors influencing waste disposal costs are important to policy makers implementing integrated solid waste management plans. Existing research on the theory of optimal pricing of municipal solid waste disposal in landfills is limited. Ready and Ready (1995) develop a theoretical model for pricing a depleteable asset that can be replaced at some cost. For example, the landfill space essentially remains fixed but is depleted as more municipal solid waste is tipped in 31

PAGE 41

32 the landfill. Space in the landfill is replaced with expansion of the landfill or construction of a new landfill. The authors find a component of the optimal fee grows at a real interest rate as space in the landfill is depleted and drops when a new landfill is built. So the shadow value of the resource increases as it becomes more scarce in the authorsÂ’ model. Also, limited survey and empirical research exists in examining tipping fees and related costs. Sheets and Repa (1990 landfill tipping fee survey, National Solid Wastes Management Association) surveyed over 200 municipal solid waste landfills in 1990 for information on tipping fees, size, intake amounts, and other characteristics. Based on prior monitoring of landfills in relation to the 1990 survey, the authors note the tipping fees have increased in different rates between various regions in the United States. The northeastern states, whose landfills have the least average remaining capacity, tend to have the highest tipping fees. Furthermore, Sheets and Repa suggest environmental standards on landfills are possibly raising the average tipping fee prices. Clayton and Huie (1973) use synthesized cost data to estimate annual total cost functions for various sizes of landfills using 1970 data, but the effects of individual factors on cost are not determined. This chapter examines the effects of various forces broadly classified as operational, regulatory, and competitive factors on tipping fees set by solid waste landfills. I identify the effects of several cost factors and find there may be some benefits from locating moderate-sized landfills in drier areas away from coastal counties to the interior of states. Effects from operation costs and regulation factors appear to override competition effects in terms of magnitude.

PAGE 42

33 Landflll Operations and Costs under Regulation Typical costs involved in landflll operations include pre-development, construction, operating, and closure costs. Major pre-development costs include site selection and design. Groundwater sources must be protected based on understanding the surface and subsurface geology of the area. Areas with higher precipitation have greater groundwater monitoring concerns. Locating landfills in remote areas usually results in lower prices paid for land space, but transportation costs are usually higher in the construetion process. Construction costs include such items as excavation, leachate colleetion and treatment, groundwater monitoring, drainage controls for surface water, fencing, structures, and scales. The various costs related to each type of construction cost will vary depending on the geography and climate of the landfill. Operating costs include such items as labor, equipment and maintenance, administrative costs, and fuel. Closing a landfill requires one-time closure costs plus post-closure costs that can last over two decades. Gas control systems and cover are the major one-time closure costs; but the landfills need to provide leachate control, groundwater monitoring, land surface eare, and other items far into the future.' The Environmental Protection Agency was instructed to implement a set of minimum criteria for solid waste management facilities as part of the Hazardous and Solid Waste Amendments of 1984. After much consideration, the Agency put forth these criteria in October 1991. The eriteria cover location restrictions, operations, design, groundwater monitoring and corrective aetion, closure and post-closure care, and financial assurance. Compliance from existing landfills was required by no later than ' Kreith (1994) provides more in-depth discussionon landflll operations.

PAGE 43

34 April 1993 as required by state permitting programs. The large increases in national average tipping fees in the 1990s as compared to the 1980s are said to be due to the new federal regulations. Methodology and Estimation This paper analyzes the effects of three broad categories on the tipping fees set by landfills: landfill operation costs, regulatory controls, and competition. A variety of variables and proxies are used to control for the effects of the various categories on tipping fees using yearly data on U.S. landfills from 1992 to 1999 from Chartwell Information. As can be seen from the Figure 3-1, tipping fees vary greatly across the United States. Tipping fee values are in real values. When examining the mean levels of tipping fees in 1992 and 1999, waste disposal fees appear to be highest in the Northeast, West Coast, and upper Midwest. In 1992 the mean for landfill tipping fees was $28.30 per ton. Tipping fees generally increased from 1992 to 1999. The largest increases in mean tipping fee levels seemed to occur in states with the lower tipping fee levels in 1992. Thirty-seven of the forty-eight continental states had higher mean tipping fees in 1999 compared to 1992. With landfills likely pricing to at least cover long-run average costs, landfill operation costs are expected to account for a significant portion of tipping fee variation across the U.S. and over time. Regardless of where landfills are located, managers need to account for leachate collection and groundwater monitoring. Leachate refers to the liquids going toward the bottom of the landfill carrying dissolved and suspended contaminates. Precipitation and moisture in the landfill waste contribute to the amount of liquids in the landfill. Once the leachate is collected, it must be shipped to be treated further or additional facilities must be constructed on-site to treat it. Together, leachate

PAGE 44

35 and groundwater monitoring costs are in the construction, operating, and closure costs of operating a landfill. Since data on the costs of various leachate collection and groundwater monitoring systems are not available, these costs are proxied with weather data. The National Oceanographic and Atmospheric Administration collects weather data from numerous weather stations across the United States. In the county where a particular landfill exists, the mean July temperature (JULY TEMP in degrees Fahrenheit) and mean January temperature (JAN TEMP in degrees Fahrenheit) are calculated as an average for the weather stations in the county and for the years between 1970 to 1999 using the TD3220 report. The countyÂ’s average annual precipitation (PRECIP in inches) is calculated.^ The averages are calculated over a long time period to capture the long-term effects of climate on long-run costs, not year to year changes in costs due to random weather. So tipping fees reflect the long-run costs of leachate collection. It is important to explain the effects of differences in long-run weather between different locations on tipping fees. Landfills leachate costs are based on long-run weather patterns, not todayÂ’s weather. Leachate collection costs are expected to be greatest in warmer areas with high precipitation, as contaminants can more easily migrate to the landfill bottoms. Hence, tipping fees set by the landfills should be higher, reflecting these increased costs. On the other hand, extreme cold weather makes daily landfill operations more difficult with extra machinery precautions and maintenance, labor force productivity losses, and higher costs to cover material. Decreases in winter temperatures are expected to increase the levels of tipping fees to cover such costs. Squared values for temperature (JULY TEMP ^ Weather data were not available for a small number of counties, and those landfills are not included in the analysis.

PAGE 45

36 SQ, JAN TEMP SQ) and temperature-precipitation interactions (JULY TEMP*PRECIP, JAN TEMP*PRECIP) are also calculated to capture non-linear effects of temperature. As the temperature gets warmer during the winter, costs should decrease but will increase once again when the soil becomes warmer, facilitating increased leachate filtration. Increasing temperature in the hot part of the year should increase leachate costs with the sludge carrying more particulates. But additional increases may result in extreme evaporation, resulting in dryer top soil thus hindering topwater from entering bottom layers in landfills. The turning points of the curves are unknown beforehand. The mean July temperature is just over 76 degrees, while the mean January temperature is just above freezing. Table 3-1 shows the summary statistics for the temperature and other variables. Pre-development costs are proxied with land prices, and day-to-day operating costs are proxied with retail wages per worker. The U.S. Agricultural Census provides data on the average price of an acre of agricultural land (LAND PRICE in dollars) for counties in 1992 and 1997, which are converted to real values. Other years are interpolated and extrapolated to obtain data for years 1992 to 1999. Many of the landfills are located in rural areas outside of metropolitan areas. This price of land is essentially a shadow price for the cost of landfill expansion. Higher land prices are expected to result in higher tipping fees. Annual retail wages per worker data (RETAIL WAGES in thousands) are collected from the U.S. Bureau of Economic Analysis CA05 and CA25 reports on a yearly basis for each county. Hired personnel at a typical landfill might include feecollectors, scalemen, foremen, machine operators and drivers, laborers, bookkeepers, and

PAGE 46

37 secreterial support. Depending on the size of the landfill, various tasks may be performed by one individual, such as one person providing bookkeeping and secreterial service. Rather than using a separate wage measure for each position in each county, I use RETAIL WAGES as a general measure of local, limited-skill labor costs. More specific measures such as construction wages might have seasonal variation and may be based on fewer workers and thus less reliable. As such a proxy, increases in retail wages per worker should result in higher tipping fees set by the landfills, but the effects are expected to be small (since leachate and groundwater monitoring costs are far greater to the landfill than payroll costs). Economies of scale in the landfill industry are investigated using data on the average daily intake for a landfill in each year from 1992 to 1999. This intake figure (AVG INTAKE in tons) and its square (AVG INTAKE SQ in tons) are included to find a minimum efficient sale of operation. Clayton and Huie (1973) found declining long run average cost curves using the synthesized data for landfills sizes 25 tons to 1700 tons of daily intake in Indiana. Table 3-1 shows the mean daily intake of 464 tons for the entire sample. The low, yearly-mean value was in 1995 at 420 tons, and the high, yearly-mean value was 527 tons in 1999. Regulations costs in addition to the minimum criteria put forth by the EPA impose various burdens on landfills. The most direct policy tool at the state level to reduce waste being deposited in landfills is to implement landfill material bans. A growing concern over landfill expansion has led state policy makers to implement legislation to ban specific materials from being tipped in landfills. The most common bans include motor oil, vehicle batteries, white goods, tires, and yard waste. Oil and

PAGE 47

38 batteries contain materials which can easily contaminate groundwater sources in areas close to water sources and residential populations. White goods, such as stoves and refrigerators, are bulky and take up large amounts of space in the landfills. Also, many contain electrical components containing polychlorinated biphenyls, which can contribute to toxic sludge material in the landfill. Tires also take up a significant amount of landfill space and create uneven settling in the landfill. In fact, tires can rise in the landfill after closure and break covers. Additionally, large piles of tires create ideal habitat for mosquitoes and rodents. Finally, tires fires create harmful fumes that are difficult to extinguish compared to fires on organic materials such as wood products. Keeping particular items out of landfills may actually decrease the costs of landfill operation. However, workers at the landfills need to monitor waste to prevent violations of landfill material bans. The sign of the net effect from the material ban depends on the magnitudes of the cost saving effect and the monitoring effect. Banning oil and batteries^ should result in the greatest savings in regards to leachate collection and treatment. In addition, banning tires is expected to result in lower costs associated with mosquito and rodent control, fires, and uneven settling in landfills. Yard waste contributes to excess gas and leachate generation compared to other materials, but composted yard waste is commonly used as an intermediate landfill layer. White goods are the easiest to monitor in incoming loads to the landfill. State landfill material bans on automotive batteries, motor oil, tires, white goods, and yard waste are captured by the following indicator variables: BATTERY BAN, OIL BAN, TIRE BAN, WHITE GOODS BAN, and YARD WASTE BAN. Battery ^ns 3 Automotive batteries remain the largest source of lead in landfills.

PAGE 48

39 affected the greatest number of landfills over the time period. Conversely, white good bans affected the least number of landfills. The variable TOTAL BANS is a summation of the five material bans and captures the cummulative effect of the various landfill bans. In addition, landfills located on the border of a particular state also are affected by landfill material bans of surrounding states. Municipal solid waste markets extend over state borders, and state policies affect market behavior. I calculate the difference between the average number of landfill bans of the states bordering the particular county and the number of landfill bans facing a landfill located in a state-border county (BAN DIFF). So a higher value for BAN DIFF indicates landfills in border counties face greater landfill bans from bordering states than their home states. Counties not on the border have a value of zero for BAN DIFF. Landfills must meet EPA guidelines when locating in wetland and floodplain areas, but the landfills are also likely to encounter landfill specific regulations and other regulations at the local level near open water. Some counties have implemented ‘setbacks’ to keep industrial activity from water sources. Many landfills face greater operating restrictions in coastal counties that are likely to increase costs. Also, hydrogeologic factors in coastal areas are likely to increase the costs of collecting and treating leachate compared to arid areas with a more predictable hydrology system and favorable soil characteristics. It is expected that landfills in coastal counties, as indicated by the variable COASTAL, will have higher tipping fees to reflect to these increased costs. It can be seen in Table 3-1 that almost one-fifth of the landfills are located in coastal counties.

PAGE 49

40 In addition to operating and regulatory costs, competition is likely to affect tipping fees. Both private and municipal landfills exist across the United States. In the report by Sheets and Repa (1990 landfill tipping fee survey. National Solid Wastes Management Association), the authors report 76%, 1 1%, and 13% of respondents had private ownership/private operator, public ownership/private operator, and public ownership/public operator structures, respectively. Anecdotal evidence suggests publicly owned landfills set their prices to break even, or to price at average total cost. Privately owned landfills are profit maximizers. When there is a mix of both private and firms serving in the same market, measures of competition are likely to be understated. Data on ownership are not available to differentiate between public and private landfills. As a method to examine effects of competitive pressures on tipping fees, I calculate three different measures: 50 MILES (small-size market definition), 100 MILES (medium-size market), and 150 MILES (large-size market). These measures contain the number of competitors in each market. For example, 50 MILES measures the number of competitors within a 50 mile radius for a given landfill. Likewise, this measure is constructed for larger areas of radii 100 and 150 miles. It is expected that increasing the number of competitors in the market would result in lower tipping fees. However, such a result is expected to be understated with the presence of publicly owned landfills. Results Regressions using the same types of specifications are presented in three different tables: 3-2, 3-3, and 3-4. In each of the three tables, a different measure of market competition is used (50 MILES, 100 MILES, and 150 MILES in Tables 3-2, 3-3 and 3-4, respectively). Also, regressions 1-a, 2-a, and 3 -a differ from regressions 1-b, 2-b, and 3b, where the first set of regressions includes TOTAL BANS, while the latter set of

PAGE 50

41 regressions includes the individual landfill material bans. Regressions 1-a and 1-b do not include any weather variables. Precipitation and temperature variables (and square values) are added to this base specification in the middle columns, 2-a and 2-b. Finally, columns 3-a and 3-b add to the base specification by including precipitation, temperature, and temperature-precipitation interactions. In general, the results from the regressions support the hypotheses. The regressions include 1570 landfills and 10793 observations. The regression results from the temperature and precipitation estimates are mixed. JULY TEMP is only statistically significant in Specification 3-b in Table 3-b. In addition, JULY TEMP SQ is not statistically significant in any tables. However, the results for JAN TEMP and JAN TEMP SQ are more promising. As predicted in specifications 2-a and 2-b in the three tables, JAN TEMP is negative and significant while JAN TEMP SQ is positive and significant. This suggests the graph of tipping fee in terms of mean January temperature is a U-shaped quadratic curve. An increase in the winter temperature measure results in a decrease of tipping fees at low temperatures and an increase in tipping fees at higher temperatures. The minimum values (turning points) range from 41 to 45 degrees Fahrenheit across the tables. As a reference, some metropolitan areas within this range include Wichita Falls (TX), Greenville-Spartanburg (NC), Seattle-Tacoma (WA), Birmingham (AL), Atlanta (GA), and Dallas (TX). Declines in extra machinery precautions and maintenance and an increase in the ease of obtaining landfill cover occur as mean January temperatures increase to the low 40s. Once the mean temperatures start to increase into the 50s, extra precautions are no longer necessary and soil temperatures increase facilitating the ease in separating proper cover material for the landfill. Also, as the temperature increases throughout the layers in

PAGE 51

42 the landfill, moisture tends to filter more easily through the landfill thus increasing leachate collection costs. PRECIP is positive and significant in all three tables, indicating an increase in annual precipitation increases the tipping fees. The variable appears alone in specifications 2-a and 2-b and is interacted with JULY TEMP and JAN TEMP in specifications 3-a and 3-b. In the first two specifications across the three tables, an increase in an annual precipitation by an inch increases the tipping fee by $.16 to $.22. Furthermore, a one-standard deviation increase in PRECIP increases TIP FEE by $2.27 to $3.12. Interacting PRECIP with temperature variables results in negative, statistically significant coefficients for JULY TEMP*PRECIP in specifications 3-a and 3-b. It is expected leachate collection would be highest in regions of the country with warm and wet climates. The negative sign is an unexpected result. As expected, the coefficient of LAND PRICE is positive and significant in each specification within each of the three regression tables. This evidence suggests land prices may be a constraint on landfill expansion. Although the agricultural land values have a positive effect on tipping fees, the magnitudes are small. In Column 1-a in Table 3-2, a one-standard deviation change in the average price of an acre of agricultural land in the county of a given landfill results in an increase in the tipping fee by 5 1 cents. The estimated coefficients for RETAIL WAGES are positive and significant in all columns in Tables 3-2, 3-3, and 3-4. Lowest values for the coefficients seem to appear in Specification 1-a while the highest values are in Specification 2-b. In Column 3-b in Table 3-2, a $1000 increase in the retail wagers per worker results in an increase in TIP FEE by 81 cents. A one standard-deviation increase in RETAIL WAGES results in a

PAGE 52

43 $2.08 increase in the tipping fee. So an increase in general labor costs is reflected in higher tipping fees on average. AVG INTAKE is negative and significant in all three tables. Furthermore, AVG INTAKE SQ is positive in all tables and statistically significant in many specifications. If the landfills are setting tipping fees in relation to average total cost, the curve produced by the average daily intake variables is a proxy to an average total cost curve. Minimum values of this curve range from 3550 to 3966 tons, 35 19 to 4075 tons, and 3423 to 3947 tons in Tables 3-2, 3-3, and 3-4 respectively. The mean value for AVG INTAKE is 464 tons. The 25th, 50th, and 75th percentiles are 61, 185, and 500 tons respectively. About 95% The tipping fee would decrease anywhere from $.92 to $1.35 going from the 25th to 75th percentile. Moving from the 25th percentile to the minimum values of the average intake curves would result in a tipping fee decrease of anywhere from $4.83 to / $7.17. These figures indicate a large percentage of landfills could continue to increase intake and capture some scale economies but costs seem to rise eventually. It is possible landfills encounter new operation costs to deal with risk factors beyond a minimum efficient scale. For example, the accumulation of municipal solid waste eventually creates odor problems which were largely absent at smaller levels of operation. Landfills may need to take additional measures beyond what is required by regulation to manage odor and disease vectors. Results from including the variables for landfill material bans are mixed. The estimated coefficient for BATTERY BAN is positive and largely significant. Having this ban results in a $2.48 to $3.33 increase in tipping fees where significant. If the ban on

PAGE 53

44 battery bans resulted in a savings from avoided costs due to handling leachate, tipping fees would decrease. Monitoring waste inflows to comply with bans is costly to the landfills and increases costs. As a result, tipping fees would increase. The estimated coefficients for BATTERY BAN suggests monitoring costs by the landfill could be high. The estimated coefficients for motor oil are negative in all tables but only significant in Specification 1-b in Table 3-2. Like automobile batteries, motor oil contributes to the toxicity of leachate collection and thus increases processing costs. In addition, oil contributes to greater fire hazard risks. The negative coefficient indicates landfills realize cost savings with the bans. The estimated coefficient implies that tipping fees decrease by $1.59 with the OIL BAN. Some of the strongest results from the material bans come from the coefficients on TIRE BAN. The estimated coefficients range from a low of -5.52 to -3.05 and are all statistically significant. Avoiding tire disposal can save the landfill expenses from mosquito control to fire control to closure issues. Tires are one of the more common materials to be banned from landfills. From a public policy perspective, it appears tire bans are serving a useful role.'* The estimated coefficients for WHITE GOODS BAN are positive and largely statistically significant. Having the ban in place for white goods raises tipping fees from $2.72 to $3.5 1 where statistically significant. It does not appear monitoring costs for white goods would be high compared to other more compact materials. Some landfills have started scrapping processes at their locations to deliver steel and other metals to recycling centers for money. Facing a ban on white goods could result in lost revenue Tire reuse and recycling is increasing as whole tires are being used in playground equipment, reef construction, and chopped for use in rubber mats, molded objects, and rubberized paving materials.

PAGE 54

45 from some landfills. So landfills may be increasing the price of one ‘output’ (accepting solid waste) while decreasing the ‘output’ of another product (scrapping white goods for sale of metals to other companies). Finally, the magnitudes of the estimated coefficients on YARD WASTE BAN are the highest. Values range from 1.88 to 6.96, with all estimated coefficients being statistically significant. Large amounts of organic matter can significantly increase leachate collection costs. Hence, avoiding this disposal can be beneficial but yard waste is one of the most difficult wastes to monitor when it is banned. Also, some landfills have implemented composting stations at their sites to process organic materials for other uses. Like the case of white goods, not having bulk shipments of yard waste could result in lost revenue if the landfill has the aforementioned facilities. The positive effects from the estimated coefficients suggest avoided leachate problems may be small. Besides examining the effect of each ban individually, I examine the collective effect of landfill material bans and the effect of bans from neighboring states on landfills located in border counties. TOTAL BANS is included in speeifications 1-a, 2-a, and 3-a in each of the three tables. The net effect suggests the material bans result in an increase in the tipping fee. Values for the coefficients on TOTAL BANS range from .87 to .92 where significant. So, a one standard-deviation increase in the number of landfill material bans results in an increase of TIP FEE anywhere from $1.15 to $2.80. The measure of the effect of bans from neighboring states, BAN DIFF, was not statistically significant in any of the regressions. In addition to the landfill bans, landfills may face more local regulation in coastal areas. Landfills located in coastal counties adjacent to the Pacific Ocean, Gulf of

PAGE 55

46 Mexico, Atlantic Ocean, and Great Lakes have a value of one for the COASTAL indicator. Estimated values for COASTAL range from 6.09 to 9.84. So being located in a coastal county may result in up to $10 higher tipping fees. With one-fifth of the landfills in the coastal counties, the aggregate financial effects of locating and operating landfills in coastal areas are substantial. The effects of competition are analyzed in each of the regression tables. Variables 50 MILES, 100 MILES, and 150 MILES are calculated for each landfill. The total number of competitor landfills within different radii are counted for each landfill. Results from the regressions indicate the estimated coefficients for each measure are positive. Only four of the eighteen coefficients are not statistically significant. In terms of magnitude, the largest effects seem to be with the specifications using 50 MILES. Column 1-a in Table 3-2 indicates adding one more competitor within a 50 mile radius results in a $.37 increase in TIP FEE. The results from these estimated coefficients are contrary to the hypothesis. More competition should result in lower prices. Also, it would be expected that competition would have more of an effect on privately owned landfills than publicly owned landfills. However, even though the publicly owned landfills may price at average costs, they do face some pressure to reduce costs and price lower prices as haulers could choose other disposal options. Finally, year effects are included in the regressions to see if there was a considerable increase in later years in the sample period compared due to effects of federal regulations. When looking at the individual coefficients from the year coefficients, there is no evidence of an increase in general tipping fees in the early 1990s when the EPA impleneted its minimum criteria guidelines for landfills.

PAGE 56

47 Conclusion This paper uses a national dataset on landfills containing yearly data on tipping fees and intake volumes from 1992 to 1999 supplemented with other data to examine the effects of various factors on tipping fees. I havenÂ’t found an existing empirical study on determinants of tipping fees. Several important findings with policy implications are found in the analysis. In terms of cost and scale, there are significant economies of scale in operating landfills. Only five percent of the landfills could be too big. Precipitation has significant effects on increasing tipping fees. Also, increases in land prices and labor costs increase tipping fees. Regulatory controls in the form of landfill material bans and local restrictions affect tipping fee levels. Landfill bans may save landfills money by reducing leachate, groundwater monitoring, vermin, mosquito, and other treatment costs. However, monitoring the bans is costly. Or a ban could result in lost revenue for a landfill which uses incoming municipal solid waste as an input to produce and sell outputs. In terms of the size of the landfill ban effects, yard waste bans significantly increase tipping fees relative to other material bans. When the net effects of the bans are considered together, banning another good raises tipping fee levels. But local regulation costs in coastal areas of the U.S. may dwarf the effects of the landfill material bans. Holding the price of land constant, operating landfills in coastal counties results in a tipping fee differential nearly one-third of the mean in the sample. If policy makers are concerned about the rising costs of landfill operations and associated tipping fees, more government intervention could result in better solid waste management policy. Landfills in coastal areas with high precipitation result in high

PAGE 57

48 tipping fee regions. Instead of operating mega-landfills in these regions, it would likely be better to operate landfills on a smaller scale toward interior counties of states. In addition, land costs and labor costs are likely to be lower, resulting in lower tipping fees. However, locating landfills in the most rural areas may result not enable the landfills to take full advantage of economies of scale due to the high transportation costs associated with moving waste from metropolitan areas to rural areas. Either haulers will be deterred from shipping waste to these remote landfills, or integrated waste collection and landfill companies will experience higher priced waste disposal as they absorb these costs. In the latter case, increased shipping costs would result in a smaller optimal size of landfill.^ Future analysis could be extended in a variety of directions. First, the effects of ownership are likely to have different effects on tipping fees with private firms facing more competitive pressures. Identifying ownership and operating structures is a first step in this direction. Second, the overall fit of the regressions could be improved by including more landfill-specific covariates such as landfill size and the amount of capital equipment at the sites. Finally, it would be useful to have information on vertical integration within the solid waste management industry. Landfills operating under an umbrella operation with a hauling division and recycling division may experience economies of scope and/or cost shifting. Unfortunately, attempts to acquire these data previously discussed were unsuccessful. A deeper analysis would need a richer set of data to investigate these questions. ^ See Kenny (1982) for an investigation of optimal plant size considering the size of geographic areas served and input costs.

PAGE 58

49 1992 Tipping Fees (means) 1 over $60 (4) FI over $40 to $60 (8) n over $20 to $40 (18) $0 to $20 (18) A Figure 3-1. Mean tipping fees by state. A) In 1992. B) In 1999.

PAGE 59

50 Table 3-1. Summary statistics for tipping fee regressions Variables Mean Standard Deviation Minimum Maximum TIP FEE 31.12 18.03 0.48 261.32 AVG INTAKE 463.60 881.67 0.12 15000 AVG INTAKE SQ 992201.40 6808753.00 0.0144 2.25e+08 LAND PRICE 2001.75 4104.68 74.31 162954.80 RETAIL WAGES 14.34 2.57 5.63 31.17 TOTAL BANS 3.04 1.32 0 5 BATTERY BAN 0.93 0.26 0 1 OIL BAN 0.58 0.49 0 1 TIRE BAN 0.76 0.43 0 1 WHITE GOODS BAN 0.31 0.46 0 1 YARD WASTE BAN 0.47 0.50 0 1 BAN DIFF -0.01 0.97 -5 5 COASTAL 0.18 0.38 0 1 JULY TEMP 76.39 5.48 61.34 93.70 JULY TEMP SQ 5865.30 844.00 3762.95 8779.69 JULY TEMP*PRECIP 2591.26 1150.39 247.42 5342.22 JAN TEMP 33.82 12.69 4.3 71.40 JAN TEMP SQ 1304.70 910.58 18.49 5097.96 JAN TEMP*PRECIP 1159.39 743.85 78.86 4033.42 PRECIP 33.73 14.20 2.71 66.42 50 MILES 8.04 5.32 0 31 100 MILES 25.79 12.61 0 68 150 MILES 52.02 21.93 0 119

PAGE 60

51 Table 3-2. Tipping fee regressions using small-market definition Independent Variables 1-a 1-b 2-a 2-b 3-a 3-b AVG INTAKE -0.0024** -0.0033*** -0.0025*** -0.0029*** -0.0024*** -0.0028*** (0.0010) (0.0009) (0.0010) (0.0009) (0.0008) (0.0008) AVG INTAKE SQ 3.03e-07 4.19e-07** 3.50e-07 3.91e-07* 3.38e-07* 3.78e-07* (2.14e-07) (2.07e-07) (2.01e-07) (2.01e-07) (1.93e-07) (1.96e-07) LAND PRICE 1.24e-04*** 8.60e-05** 9.12e-04** 7.94e-05** 7.62e-05** 6.77e-05* (4.32e-05) (4.08e-05) (-3.96e-05) (-3.95C-05) (3.63e-05) (3.60e-05) RETAIL WAGES 0.3696** 0.7752*** 1.0989*** 1.1111*** 0.8652*** 0.8077*** (0.1657) (0.1779) (0.2005) (0.1980) (0.1883) (0.1842) TOTAL BANS 0.9189*** -0.0249 0.5279 (0.3288) (0.3422) (0.3265) BAN DIFF -0.112 -0.2263 -0.5648 -0.5379 -0.2108 -0.2290 (0.4484) (0.4412) (0.4166) (0.4164) (0.3942) (0.3955) BATTERY BAN 3.3269** 2.7035*** 0.6477 (1.5374) (1.5483) (1.4226) OIL BAN -1.5940* -1.2856 -0.2751 (0.8471) (0.7989) (0.7538) TIRE BAN -5.3006*** -3.0504** -3.7873*** (0.9030) (0.9829) (1.0059) WHITE GOODS 2.7190*** 1.1717 3.3750*** (0.9218) (0.9556) (0.9190) YARD WASTE BAN 6.9614*** 2.4846** 2.2825** (0.8098) (1.0762) (1.0320) COASTAL 9.0533*** 7.9465*** 6.4454*** 6.0933*** 7.5829*** 6.9093*** (1.1924) (1.1691) (1.3022) (1.2769) (1.2470) (1.1870) 50 MILES 0.3716*** 0.2962*** 0.2376*** 0.2280** 0.1210 0.1270 (0.0984) (0.0939) (0.0924) (0.0922) (0.0915) (0.0900) JULY TEMP -0.3097 -0.8468 0.2373 0.2402 (1.4082) (1.3434) (0.1781) (0.1722) JULY TEMP SQ -0.0016 -0.0021 (0.0090) (0.0086) JULY TEMP*PRECIP -0.0350*** -0.0322*** (0.0086) (0.0081) JAN TEMP -0.9393*** -0.7883*** 0.0452 0.1702* (0.1690) (0.1786) (0.0967) (0.1005) JAN TEMP SQ 0.0106*** 0.0092*** (0.0023) (0.0023) JAN TEMP*PRECIP -0.0049 -0.0076** (0.0035) (0.0035) PRECIP 0.2201*** 0.1873*** 3.1314*** 3.0209*** (0.0321) (0.0361) (0.5905) (0.5614) Constant 16.6354*** 11.9953*** -54.1589 70.0773 -16.2527 -18.2983 (2.2103) (2.5257) (54.5424) (52.0909) (12.4424) (12.0172) Adjusted R-squared 0.09 0.13 0.18 0.19 0.20 0.21 * 10% level of significance Robust standard errors in parentheses. ** 5% level of significance All specifications include year effects. *** 1% level of significance

PAGE 61

52 Table 3-3. Tipping fee regressions using medium-market definition Independent 1-a 1-b 2-a 2-b 3-a 3-b Variables -0.0026*** -0.0033*** -0.0026*** -0.0029*** -0.0024*** -0.0027*** (0.0008) (0.0008) AVG INTAKE AVG INTAKE SQ LAND PRICE RETAIL WAGES TOTAL BANS BAN DIFF BATTERY BAN OIL BAN TIRE BAN WHITE GOODS YARD WASTE BAN COASTAL 100 MILES JULY TEMP JULY TEMP SQ JULY TEMP*PRECIP JAN TEMP JAN TEMP SQ JAN TEMP*PRECIP PRECIP Constant (0.0009) (0.0009) 3.19e-07 4.15e-07** (2.07e-07 (2.04e-07) 1.17e-04*** 8.71e-05** (3.99e-05) (3.87e-05) 0.2951* 0.6480*** (0.1551) .9235*** (0.3188) (0.1703) -0.0256 -0.1480 (0.4358) (0.4335) 3.0967** (1.4432) -0.9713 (0.8415) -5.4030*** (0.8604) 3.1501*** (0.9095) 5.9208*** (0.8025) 9.1263*** 8.0545*** (1.1741) (1.1555) 0.2650*** 0.2104*** (0.0340) (0.0348) 13.8069*** 11.0431*** (2.2910) (2.4940) (0.0009) (0.0009) 3.51e-07 3.87e-07* (2.00e-07) (2.00e)-07 9.04e-05** 8.09e-05** (3.82e-05) (3.81e-05) 0.9770*** 0.9857*** (0.1919) -0.064 (0.3316) (0.1906) -0.5277 -0.5044 (0.4078) (0.4087) 2.4945* (1.4683) -0.8971 (0.8044) -3.2470*** (0.9748) 1.3437 (0.9422) 1.8846* 6.4681*** (1.0576) 6.1788*** (1.2905) (1.2650) 0.1847*** 0.1772*** (0.0346) (0.0354) -1.2605 -1.7005 (1.4175) (1.3686) 0.0048 0.008 (0.0090) (0.0087) -1.0248*** -.8775*** (0.1676) (0.1773) 0.0119*** 0.0104*** (0.0023) (0.0023) 0.1944*** 0.1739*** (0.0321) (0.0359) 89.8201 102.0465 (54.8706) (53.0049) 3.41e-07* 3.76e-07* (1.92e-07) (1.94e-07) 7.65e-05** 6.90e-05* (3.56e-05) (3.53e-05 0.8390*** 0.7943*** (0.1829) 0.5764 (0.3256) (0.1805) -0.1771 -0.2041 (0.3926) (0.3946) 0.8306 (1.3847) -0.1751 (0.7629) -3.7000*** (1.0078) 3.4448*** (0.9137) 2.0376** (1.0170) 7.6553*** 7.0414*** (1.2338) (1.1744) 0.0920** 0.0830** (0.0397) (0.0389) 0.2861 0.2862* (0.1778) (0.1735) -0.0361*** -0.0332*** (0.0084) (0.0081) -0.0139 0.1139 (0.1048) (0.1081) -0.0028 -0.0058 (0.0037) (0.0037) 3.1124*** 3.0144*** (0.5865) (0.5603) -18.1664 -20.3518* (12.3335) (11.9856) Adjusted R-squared OJJ 0T5 0T9 0.19 0.21 0.22 * 1 0% level of significance Robust standard errors in parentheses. ** 5% level of significance All specifications include year effects. *** 1% level of significance

PAGE 62

53 Table 3-4. Tipping fee regressions using large-market definition Independent Variables 1-a 1-b 2-a 2-b 3-a 3-b AVG INTAKE -0.0026*** -0.0033*** -0.0026*** -0.0029*** -0.0023*** -0.0027*** (0.0009) (0.0009) (0.0009) (0.0009) (0.0008) (0.0008) AVG INTAKE SQ 3.32e-07* 4.18e-07** 3.54e-07 3.88e-07* 3.36e-07* 3.73e-07* (2.00e-07) (1.99e-07) (1.99e-07) (1.98e-07) (1.90e-07) (1.92e-07) LAND PRICE 1.14-e04*** 8.79e-05** 9.08e-05** 8.20e-05** 7.78e-05** 6.95e-05** (3.86e-05) (3.79e-05) (3.78e-05) (3.78e-05) (3.57e-05) (3.54e-05) RETAIL WAGES 0.4319*** 0.7212*** 1.0347*** 1.0376*** 0.9199*** 0.8768*** (0.1543) (0.1689) (0.1906) (0.1891) (0.1835) (0.1806) TOTAL BANS 0.8699*** -0.2047 0.4784 (0.3187) (0.3275) (0.3260) BAN DIFF -0.0894 -0.1886 -0.6166 -0.5875 -0.2309 -0.2577 (0.4414) (0.4375) (0.4107) (0.4109) (0.3949) (0.3962) BATTERY BAN 3.0803** 2.4836* 0.7959 (1.4423) (1.4686) (1.4067) OIL BAN -0.9211 -0.9916 -0.3507 (0.8485) (0.8093) (0.7732) TIRE BAN -5.5228*** -3.3909*** -3.7872*** (0.8871) (0.9780) (1.0072) WHITE GOODS 3.5131*** 1.349 3.3303*** (0.9262) (0.9398) (0.9217) YARD WASTE BAN 5.3389*** 1.4534 2.1513** (0.8319) (1.0587) (1.0115) COASTAL 9.8429*** 8.6381*** 6.8439*** 6.5871*** 7.7854*** 7.1134*** (1.1645) (1.1474) (1.2908) (1.2663) (1.2514) (1.1892) 150 MILES 0.1546*** 0.1229*** 0.1135*** 0.1091*** 0.0232 0.0164 (0.0186) (0.0202) (0.0208) (0.0217) (0.0250) (0.0249) JULY TEMP -1.2465 -1.6403 0.2531 0.2476 (1.4063) (1.3584) (0.1785) (0.1739) JULY TEMP SQ 0.0048 0.0077 (0.0090) (0.0087) JULY TEMP*PRECIP -0.0351*** -0.0322*** (0.0085) (0.0081) JAN TEMP -1.1434*** -0.9971*** 0.0296 0.1643 (0.1707) (0.1806) (0.1094) (0.1123) JAN TEMP SQ 0.0138*** 0.0112*** (0.0023) (0.0024) JAN TEMP*PRECIP -0.0043 -0.0075* (0.0039) (0.0039) PRECIP 0.1772*** 0.1614*** 3.1147*** 3.0123*** (0.0331) (0.0362) (0.5954) (0.5654) Constant 10.6647*** 9.1416*** 89.3160 99.6971 -17.6008 -19.4956* (2.4458) (2.5420) (54.5133) (52.6796) (12.3534) (11.9873) Adjusted R-squared 0.12 0.15 0.19 0.19 0.19 0.21 * 10% level of significance Robust standard errors in parentheses. ** 5% level of significance All specifications include year effects. *** 1% level of significance

PAGE 63

CHAPTER 4 DETERMINANTS OF HOUSEHOLD RECYCLING: A MATERIAL-SPECIFIC ANALYSIS OF RECYCLING PROGRAM FEATURES AND UNIT PRICING' Introduction The past 1 5 years have been a time of dramatic change for solid waste management. Beginning in the mid-1980s, with stricter EPA requirements for landfill construction on the horizon, landfill tipping fees increased dramatically and there was a widespread impression that landfill space was growing scarce and that a landfill “crisis” was inevitable. Two clear national trends in solid waste management emerged as a result of local efforts to reduce the quantities of waste being landfilled. The most pervasive was the introduction of residential curbside recycling programs. In 1988, there were approximately 1000 such programs in the U.S.; in 1992, there were almost 5000; by 1999 the number reached just over 9000 (Goldstein and Madtes, 2000). A second, less pervasive but still important, trend during this period was the introduction of volumebased pricing, or unit pricing, of solid waste disposal services wherein households are charged for garbage collection according to the number of containers they set out. Prior to the late 1980s there were perhaps a few dozen such programs. By 1992, there were approximately 2000; and by 1999, just over 4000 (Miranda and Aldy, 1998). ‘ Adapted from Journal of Environmental Economics and Management, Vol. 45, Jenkins, R.R., Martinez, S.A., Palmer, K., & Podolsky, M.J., “The Determinants of Household Recycling; A Material-Specific Analysis of Recycling Program Features and Unit Pricing,” Pages 294-318, Copyright (2003), with permission from Elsevier. 2 Most of the increase in tipping fees occurred during the middle and late 1980s. In 1985 the national average tipping fee in the U.S. was approximately $ 1 1 .20 per ton; in 1990, it was approximately $33.75. As of 1997, it remained close to $30.00. (All values are in 1997$.) (U.S. EPA, 1997). 54

PAGE 64

55 Though the nature of a curbside recycling program is quite different from a unit pricing program, both theoretically provide incentives for a redirection of waste quantities from disposal sites to recycling centers. A curbside program reduces a householdÂ’s cost of recycling by making recycling more convenient and less time consuming. A unit pricing program increases a householdÂ’s cost of discarding additional waste relative to its cost of recycling (i.e., not recycling leads to higher fees for waste collection services).^ Each program targets different waste management activities, which might lead to differences in the outcomes of the two programs. For example, unlike a curbside recycling program, unit pricing only gives an indirect incentive to recycle while its direct incentive is to reduce waste quantities. Unit pricing may also create incentives for households to adjust their purchasing habits to generate less solid waste. Thus, the two programs might very well have different effects on household recycling effort. Economic principles also suggest that the two programs will have different impacts on recycling and consumption of different recyclable materials (Jenkins, 1993). One suggestion is that volume-based unit pricing will give households an incentive to recycle bulky items that take up lots of garbage container space such as plastic milk jugs. On the other hand, unit pricing might encourage households to avoid generating bulky wastes in the first place. Households might alter the composition of their consumption bundles so that there is less trash to discard. ^ Without unit pricing, most communities finance waste disposal via general tax revenues or flat fees. From the perspective of households, this places a marginal price of zero on waste disposal. This causes them to dispose of more than the socially efficient amount of waste. A unit pricing program imposes a non-zero marginal price on waste disposal that can potentially correct this problem.

PAGE 65

56 A curbside recycling program also might disproportionately affect certain materials. As a substitute for drop-off recycling, curbside collection mainly reduces a householdÂ’s costs of transporting recyclable materials. Compared to a household without any local recycling program, a household with a eurbside program will have a much easier time recycling materials that are hard to transport, like glass bottles, which are bulky and can break. Policy makers would benefit from a better understanding of the impact of the two programs and their features on different recyelable materials. To the municipalities that collect them, different recyclable materials have different costs of recycling as well as different values on the open market. Understanding whieh program features lead to greater recycling of high valued materials could improve the cost-effectiveness of a communityÂ’s efforts to promote reeycling. In other eases, municipalities sometimes achieve very high recycling effort directed at a few materials. In order to increase their aggregate reeycling percentage in an effort to meet state-mandated recycling rate targets, municipalities must sometimes encourage households to recycle additional materials. Understanding how best to promote recycling of a broader range of materials would be beneficial. On the other hand, if the costs of adding a particular material to a curbside program exceed the waste diversion and recycling revenue benefits of doing so, then adding certain materials may not be worthwhile. This study analyzes a large household-level data set representing 20 metropolitan statistical areas (MSAs) across the country to study the impact of these two popular solid waste programs and their features on the percent recycled of five different materials: glass bottles, plastic bottles, aluminum, newspaper, and yard waste. All eommunities in

PAGE 66

57 the data set offer curbside recycling of at least one of the five materials; although most offer it only for a subset of the five. However, the data set contains detailed information on the attributes of different recycling options for all five specific materials. For example, the data indicate whether each material is collected at all through a local program and if so whether it is collected curbside or at a local drop-off facility. The data also indicate whether recycling the material is mandatory or voluntary and the age of the recycling program. Finally, the data set contains rich household level socioeconomic information. We augment the household-level data with community-level information on the prices charged for disposal under a unit-pricing program where it is applicable. The contributions of this paper are more easily understood within the context of the literature that has investigated the determinants of recycling. Thus, we start with a brief review of this literature and adopt from it a simple theoretical model. We then describe our own data and present an empirical model. We present the empirical results, note limitations of the data and in closing, discuss the relevance of our findings to policy. Prior Research and a Conceptual Framework This paper makes two contributions to the existing economics literature on recycling. First it adds to the research on the effectiveness of curbside recycling and unit pricing at encouraging households to recycle. Several papers study various aspects of these programs, sometimes with unit pricing and curbside recycling operating together and sometimes with one program operating in isolation (Kinnaman and Fullerton, 2000; Callan and Thomas, 1997; Fullerton and Kinnaman, 1996; Hong et al. 1993; Van Houtven and Morris, 1999). However, ours is the first that analyzes data from most

PAGE 67

58 major U.S. metropolitan areas and rests on a household-level unit of analysis/ Household-level is preferred to community-level because households are the decisionmaking units that are the target of recycling policies. Analyzing data from numerous MSAs located in different parts of the country is preferred to an analysis of only one region because it facilitates the identification of policies and demographic variables that are significant across regions. A second contribution of this paper is to extend previous research by investigating whether and how the impact of these two popular programs differs for different recyclable materials. The few existing material-specific studies have lacked the rich amount of information about both recycling and unit pricing programs contained in our data set (Saltzman et al. 1993; Reschovsky and Stone, 1994). We also examine the effect of household socio-economic characteristics on recycling effort directed at different materials. Table 4-1 summarizes the existing econometric literature that studies the effects of unit pricing and curbside recycling on household recycling effort. A number of papers have developed conceptual frameworks to study the impact of unit pricing (Fullerton and Kinnaman, 1996; Morris and Holthausen, 1994; Jenkins, 1993). Others, including Podolsky and Spiegel (1998) and Kinnaman and Fullerton (1995), describe the substitution possibilities between waste disposal and recycling as part of household waste management. These papers develop models in which households maximize utility '* Several econometric studies analyze the impacts on recycling effort of one or both of these two popular programs by examining household-level data; in particular, Nestor and Podolsky (1998), Fullerton and Kinnaman (1996) and Hong et al. (1993). However, the data for all three of these studies are for a single region where curbside recycling and unit pricing co-exist. Several other studies are national in scope but rely on community-level data (Kinnaman and Fullerton, 2000; Miranda et al. (1994); U.S. EPA, 1990). (The latter two use the case study method of analysis.)

PAGE 68

59 subject to a budget constraint that incorporates a unit price for waste collection. The models are the basis for solid waste disposal and recycling demand equations. On the right hand sides of these equations are three categories of exogenous variables: characteristics of the goods whose consumption generates waste; descriptions of the local waste management system; and socio-economic factors. The first category includes the price of consumption good i (Pj) and the amount of waste generated per unit of good i (Pi) where (/ = 1 . . . n). The second category consists of the price per unit of waste disposal (Pw) and a vector of recycling program features (RP) including whether the collection occurs at the curb or at a drop-off facility, the length of life of a recycling program, and so on.^ The third category is comprised of socio-economic characteristics (SE) such as household size, income and education. Specifically, D and R are the optimal levels of household disposal and recycling. Each recycled material, j, has unique characteristics that could affect the relationship between recycling and the exogenous variables. These characteristics include factors such as bulkiness that affect the ease of recycling as well as the availability of substitutes for the material. Thus, each material (Rj) has a unique recycling demand equation as specified in Eq. 4-2. ’ The price per unit of waste disposal charged to households is usually a volume-based price. For example, households in communities employing a bag/sticker purchase official program bags or stickers, which they affix to garbage bags of the mandated size. Alternatively, households in communities using a subscription can program specify a level of waste disposal per period of time in advance and are charged according to this level. (4-1) Yj (Pb •••> pn, Pl> •••, Pn, Pw, RP, SE) (4-2) R = S Rj (4-3)

PAGE 69

60 Consistent with Eq. 4-2, we analyze material-specific recycling behavior for each of five materials. However, since we do not have data on recycling quantities, we actually estimate the effects of the exogenous variables on the intensity of recycling for each material. Data Description The primary data source is a recycling survey mailed by Equifax, Inc. in 1992 a year of increasing popularity for unit pricing and soaring popularity for curbside recycling.^ The survey was mailed to 4600 households residing in 20 U.S. metropolitan areas (please see Table 4-2 for a list of the 20).^ The survey was targeted toward middle and upper income households in these regions. Sixty-five percent of questionnaires, 2984, were returned. Households responded to questions about recycling participation, recycling program characteristics, household characteristics, and attitudes. Equifax supplemented the survey with its own data on age, income, education and other characteristics for each household. From the Equifax data set, we selected only households that reported their communities had an ongoing recycling program (N=1939). Those households who report no recycling program were not asked to report recycling percentages and thus were not eligible for inclusion in our data set. We then appended unit pricing data from three sources. The first is a 1997 report (Miranda and LaPalme, Unit based pricing in the United States: A tally of communities, Nicholas School of the Environment, 1997) that ® During 1992, the number of curbside recycling programs in the U.S. increased by 10 percent, from just under 4000 to 5404 (Steuteville and Goldstein, 1993). These 4600 households were selected using a stratified sampling method from Equifax’s 250,000 member Home Testing Institute Panel. For this panel of homes, Equifax has extensive data on socio-economic household characteristics such as income and education. The 4600 households were selected to provide a mix of ages and household income levels representative of the middle and upper middle class populations in these regions.

PAGE 70

61 identifies which U.S. communities had a unit pricing program for solid waste collection in 1992. The second is an EPA survey (1993), which collected information regarding the actual unit prices charged in 1992 by many of the unit pricing communities that were then in existence. For those communities not included in the EPA survey, we conducted our own telephone survey of community solid waste officials to solicit information on unit prices and other characteristics of the unit pricing program. Following our telephone survey, we eliminated 123 additional households living in communities with unit pricing from our data set for various reasons. The most common is that we were unable to contact a government representative who could provide information about the unit pricing program. In other cases, the community had multiple trash haulers and solid waste user fees, and we were unable to connect a particular household to a particular fee level. In addition, we deleted several observations due to missing values. Finally, to reduce the bias associated with avid recyclers being more likely than others to know about drop-off programs, we retained only those households that reported the availability of curbside collection of at least one of the five materials. Stated differently, we excluded from our sample all households living in communities with only drop-off recycling. The reason is that drop-off programs are notorious for being poorly publicized. Conversely, curbside programs are well promoted and widely recognized, at least in part because of the visibility of curbside containers on collection day. Where drop-off and curbside programs co-exist, drop-off programs are often jointly promoted with curbside recycling. For example, certain occasions such as the introduction or revision of

PAGE 71

62 a curbside program, warrant distribution of instructions for curbside recycling. (Instructions are also distributed to new residents of a neighborhood.) These instructions outline which materials can be placed at the curb and which cannot, and often give instructions for recycling the latter materials at existing drop-off centers. The extent to which drop-off recycling is promoted alongside curbside recycling varies across communities. Widespread awareness of a curbside program certainly does not guarantee widespread awareness of drop-off centers. To identify communities where residents do have good information about all recycling options, including the less visible drop-off programs, would require data that was unavailable to us, such as community level information on recycling promotion expenditures. In the absence of such data, however, we can reasonably expect that the bias associated with endogeneity of reporting the existence of a drop-off program will be reduced when we eliminate from our sample those communities with only drop-off recycling. Our final data set consists of 1049 observations. To examine the reliability of the policy information reported by respondents, we investigated whether respondents living in the same zip code area, the smallest geographical unit for which we had information, reported the same recycling program characteristics. There were many differences. Phone calls to municipalities as well as anecdotal information suggest that recycling programs differ across neighborhoods even within the same zip code. For instance, curbside recycling is often introduced to a region one neighborhood at a time. Gradual introductions might especially affect data for 1992 when many curbside programs had only recently been initiated. Another possibility is that urban parts of a zip code have curbside recycling while rural parts do not. * We discuss the implications of this concern about bias from avid recyclers for our results in Results.

PAGE 72

63 Of the final 1049 observations, 1 16 are households facing a positive unit price for solid waste collection. Table 4-3 identifies the MS As with unit pricing programs, the number of communities within each MSA with its own unique unit price, and the number of respondents residing in each MSA. The highest concentration of these respondents is in the Portland MSA, within which 37 respondents reside in the city of Portland, and nine respondents reside in four other Portland MSA communities each of which charges a unique unit price. Another concentration is in the Seattle MSA within which 18 respondents reside in the city of Seattle and 16 reside in six other Seattle area communities, each with its own unit price. The majority of respondents facing a unit price live in western states. Of the 116 households, 104 live in communities with subscription programs where households subscribe to collection of a pre-specified number of containers. Households can change that number but the waste collection service must be notified (usually by telephone or mail) of the householdÂ’s desire to change. This feature combined with weekly variations in trash generation probably leads to partially filled containers during some weeks and to storing excess waste until the next collection day during other weeks. The remaining 12 households live in communities with bag/tag/sticker programs where households place their garbage in specially marked plastic bags, or place specially marked tags or stickers on regular garbage containers, and pay a price for the specially marked items that includes the cost of collection. In these communities, households can more readily alter the number of containers discarded. We define the marginal price of solid waste collection as the price of the second container of waste. The reason is that households virtually always generate some solid

PAGE 73

64 waste, so paying for collection of the first container is difficult to avoid. Not paying for the second container is more likely and can be achieved by increased recycling.Â’ Figure I shows the distribution of values for the price of the second container across the 116 households with unit pricing. The values range from $0.41 to $3.46. Households in communities with no unit pricing face a zero marginal price for solid waste collection. Table 4-4 gives the mean values and standard deviations of the independent variables used in our ordered logit analysis. The first row gives the mean marginal price of solid waste collection (PRJCE-SW), $1.91 per 32-gallons, faced by the 116 households in communities with unit pricing programs. Two communities have a different price structure for yard waste and the second row of Table 4-4 gives the mean marginal price of yard waste collection (PRICEYW). Subsequent rows report information on the characteristics of the recycling programs and the socioeconomic characteristics of the respondents. In addition to the data reported here, we created a series of dummy variables that indicate the metropolitan statistical area (MSA) where each household is located. This variable is used in the regressions to control for unobserved regional effects such as weather and cultural differences. Comparing a subset of the socioeconomic data in Table 4-4 with 1990 U.S. Census information about the characteristics of the general population in the 20 MSAs from which the sample is drawn, illustrates the effect of targeting middle and upper income households. While the sample has approximately the same household size distribution as the larger population, the sample is more highly educated; 44% of Â’ Perhaps less easily, households also can avoid paying for the second container by generating less garbage in the first place.

PAGE 74

65 respondents graduated from college while only 22% of the larger population did. In addition, the sample under-represents the lower income segments of the population and over-represents households with incomes between $50,000 and $75,000. However, the median income of the sample is roughly $40,000 which is only $5,000 above the median income in 1990 for the group of 20 MSAs. The sample also has a higher proportion of detached home dwellers than the population at large. These comparisons make explicit the fact that our results should be generalized only to middle and upper income segments of the population We construct the dependent variable in our analysis using survey responses about recycling participation. Respondents were asked what proportion of the following materials they recycled through all available recycling programs: steel sided cans, glass bottles, plastic bottles, newspaper, magazines, aluminum, other plastics, yard waste and other. As noted already, we chose to study five of these materials and constructed a dependent variable for each of the five. The survey asked whether recycling percentages fell into one of seven possible categories: 0 to 10%; 1 1 to 25%; 26 to 50%; 51 to 74%; 75 to 84%; 85 to 95% or over 95%. We aggregate the data into three categories of “proportion of the material recycled” 0 to 10%, 1 1 to 95%, and over 95%. Table 4-5 gives the percent of respondents falling into the three categories for each of the five materials. Except for yard waste, the majority of respondents recycled over 95 percent of each material. Table V also gives the number of respondents falling into each category and the number of missing observations for each of the five ordered logit equations. Model Specification The model that we estimate seeks to identify which policy and socioeconomic factors influence the level of recycling effort households expend on each recyclable

PAGE 75

66 material. We use a latent regression model for ordered data as the framework for estimation. As noted above, for each material type, we define three ordered categories: category 0 for 0 to 10% recycled, category 1 for 1 1 to 95% recycled and category 2 for over 95% percent recycled. For each material type, j, we consider the relationship ( 4 4 ) where y* is unobserved level of recycling effort (percentage of material j recycled) and i is an index of households. The vectors, contains the marginal price, recycling program attributes, and socio-economic features for each household, p is a vector of coefficients to be estimated by maximum likelihood estimation (MLE) in an ordered logit model. Assuming Sj^ is distributed standard logistic, the probability that we observe household i in category k, where ^0,1 or 2, for material j is given by 1 + e PjXji , -ll+B:Xi: , PjXji 1 + e ^ J 1 + e J J (4-5) (4-6) (4-7) We select the ordered logit specification instead of the ordered probit because the binomial logit is more amenable to incorporating fixed effects than the binomial probit. (Hsaio, 1986). In the case of a binomial logit or probit, traditional maximum likelihood estimators for the PÂ’s will be inconsistent when fixed effects are included in the model. However, the conditional logit model (McFadden, 1974) can be used to find consistent parameter estimates for a logit when fixed effects are included. Unfortunately, the consistent estimator of P in a model with fixed effects that is well defined for an binomial logit is not well defined for an ordered logit. Therefore we simply estimate a regular ordered logit with regional metropolitan statistical area dummy variables included. We have also estimated the same model using an ordered probit specification and we find that the results (in terms of which variables are significant and the signs of the effects) are virtually the same.

PAGE 76

67 Results The intensity of household recycling activities by material is modeled as a function of the socioeconomic variables and policy variables that are described earlier. We used the same set of independent variables for each material, except that the values for the curbside and drop-off indicator variables varied across materials depending on the type of collection available for the specific material. In addition, the marginal disposal price was different for yard waste. The results of the econometric estimation of the ordered logit regression for each material are presented in Table 4-6. These results indicate the significance and direction of each variable’s effect on the propensity to recycle different materials.'' Because of the non-linear estimation procedure employed, the regression results in Table 4-6 do not provide a good indicator of the magnitude of the effect. To determine magnitudes, we use the estimated logit model coefficients to calculate the marginal effects of different independent variables on the probability that a typical household will fall into each of the three levels of recycling intensity: 0 to 10% of the material recycled, 1 1 to 95% recycled or over 95 percent recycled. For the significant policy variables, these marginal effects “To examine the sensitivity of the results to our three-way partition of the dependent variable, we also estimated equations with the dependent variable separated into only two partitions households who recycle between 0 and 10% of a material and those who recycle greater than 10%. For the aluminum, plastic bottle and yard waste equations, the significant policy variables remained so. However, for the newspaper equation, the indicator variable for drop-off collection and the variables representing the number of materials picked up curbside and the length of the recycling program became insignificant (although of the same sign) under the binomial specification. For the glass bottles equation, the indicator variables for both curbside collection and drop-off collection changed to insignificant (although of the same sign) under the binomial specification. Some of the socioeconomic variables that were significant under the multiple category specification became insignificant under the binomial specification. Overall, the binomial specification gave similar, but somewhat weaker results for the bulk of the materials. (For this sensitivity analysis, we use a binomial logit model with MSA dummy variables instead of a conditional binomial logit model with fixed effects. We do this in order to provide the most straightforward comparison to our ordered logit model with three categories. The equation that predicts the probability that an observation will fall into each of the three categories is non-linear in the independent variables. Therefore, the equation that defines the marginal effects of each

PAGE 77

68 are reported in Table 4-7. The table also converts the marginal effect into a percentage change from the actual probability a respondent will fall into a category and reports these percentage changes in parentheses. The diverse nature of the communities and households represented in our data set led us to question the appropriateness of the standard assumption that all of the disturbance terms in the underlying model have a common variance. In particular, we suspected that the variance of the disturbance terms surrounding the propensity to recycle could be a function of the presence of curbside recycling and the length of time that the recycling program had been in existence. We hypothesize that the variance of the regression disturbance terms are likely to be different across households that have curbside recycling for the relevant material and those that do not. By eliminating the need to transport recyclables to drop-off points at varying distances from the household, curbside recycling tends to even out the time required to recycle across households resulting in less variation in errors. Likewise, we expect households with greater potential experience with recycling to have disturbance terms with a lower variance than those with less experience with recycling. Greater experience with recycling allows households to develop a recycling habit, which will lead to less variation in the error terms. Using these variables as determinants in a multiplicative model of heteroskedasticity of the form Sj^ exp(jz ^ ) where the z vector includes the three potential contributors to heteroskedasticity, we tested the ordered logit model for each independent variable on that probability is a function of all of the independent variables. We calculate marginal effects by using the average value for all of the independent variables except where noted in Table 4 7 .

PAGE 78

69 1 material for the presence of heteroskedasticity. We found that for two matenals, glass and plastic bottles, we could reject the null hypothesis of homoskedasticity. Thus, we apply HarveyÂ’s multiplicative heteroskedasticity correction to the models for those two materials (Harvey, 1976). In the next three subsections, we discuss the results for three categories of independent variables: recycling program features, unit pricing policies and socioeconomic characteristics. A final section describes potential problems presented by our data and their solutions. Recycling Program Features This analysis identifies several features of recycling programs that have a significant effect on intensity of household recycling effort. Two features that are always significant are availability of local drop-off recycling and existence of curbside recycling. Increasing the number of total materials included in the curbside recycling program has a positive effect on recycling effort for newspaper only. Length of program life is also an important determinant of the intensity of recycling effort for newspaper and yard waste. The effects of individual program features are discussed in greater detail in the following paragraphs.*'* There are three potential contributors because the amount of time a recycling program has been in place is represented by two categorical indicator variables. One popular program to encourage recycling of beverage containers is a deposit refund program. During the time period of our data, deposit refund programs existed in 10 states, five of which (New York, Massachusetts, Connecticut, California and Oregon) are sampled by our data set. However, the questionnaire directed respondents to report the percentage of materials recycled but to exclude containers returned for a deposit. Assuming that beverage containers are easy to recycle, a possibility is that excluding these containers from consideration might reduce the percentage of the waste stream that is easily recyclable. Thus, states with bottle bills might be less responsive to recycling incentives. Unfortunately, we are unable to test for this directly in our model without excluding the MSA dummy variables, which would create potential endogeneity problems. However, we did look at the coefficients on the indicator variables for those MSAs that have deposit-refund programs to see if they were systematically different in some way from those for the other regions. For glass bottles we saw no discemable difference. For aluminum cans, most of the coefficients for the MSAs with bottle bills were insignificant.

PAGE 79

70 The two most commonly significant recycling program policy variables, the dropoff and curbside program indicators, serve as proxy measures of the convenience of recycling. Introducing a local drop-off program for recycling of a particular material decreases the time and storage costs associated with other modes of recycling such as accumulating materials to haul to more distant recycling centers or participating in infrequent recycling drives for charity. Instituting a curbside recycling program makes recycling even more convenient, thus its effect on recycling effort should be bigger than the effect of a drop-off program. Curbside collection lowers the time and out-of-pocket costs of recycling by completely eliminating the need to transport recyclables to collection points or to store them for long periods of time. The results reported in Tables 4-6 and 4-7 conform to these expectations. The econometric results reported in Table 4-6 show that for all materials, instituting a local drop-off program has a positive and significant impact on intensity of recycling effort. The marginal effects reported in Table 4-7 show that the magnitude of the effeet of the drop-off program variable varies dramatically across materials. Introducing a local drop-off program increases the probability that over 95% of all glass bottles used in the household are recycled by 42 percentage points; for plastic bottles the marginal effect is 33 percentage points and for aluminum and newspapers it is 19. These results suggest that introdueing a local recycling option has a smaller impact on materials for which there were recycling options even before the local drop-off program.'^ Charity drives, for example, have traditionally focused on collecting newspapers and/or aluminum. Newspaper carried to (or even purchased at) work may be recycled at work This effect might be exaggerated because of a possible over-representation of avid recyclers reporting drop-off programs. Avid recyclers might be more likely to seek out recycling alternatives in the absence of a local program.

PAGE 80

71 and beverage cans used outside the home may be recycled at the place of use. Adding a local drop-off program is likely to have little impact on this type of recycling behavior. The different magnitudes also suggest that introducing a local drop-off program has a greater impact on materials for which transportation and storage would be most difficult for households. Without a local program, for materials not collected by special drives or recycled away from home, the household must travel to a distant recycling center. Relative to glass and plastic bottles, newspapers and aluminum (after it has been crushed) are compact and clean and thus more likely to be accumulated and transported long distances. In contrast, storing and transporting glass and plastic bottles is more burdensome to households. Adding a drop-off recycling program reduces householdsÂ’ transportation costs by improving the proximity of recycling centers. Improved proximity might also increase the frequency of drop-offs that would reduce householdsÂ’ storage costs. Thus, it is not surprising that introducing a drop-off program has a bigger impact on glass and plastic bottles than on newspapers and aluminum. Introducing a local drop-off option for yard waste increases the probability that over 95% of it will be recycled by 19 percentage points. While the magnitude of this effect is similar to newspaper and aluminum, it represents a percentage increase above baseline recycling levels similar to that experienced for glass and plastic bottles approximately 60. (Table 4-7 presents these percentage increases or semi-elasticities in brackets.) This finding suggests that drop-off recycling has a larger effect on yard waste (a material with high transportation and storage costs) than appears at first glance. As expected, the presence of curbside recycling has a positive and significant effect on intensity of recycling activity for all five materials. The magnitude of this varies

PAGE 81

72 substantially across materials, just as the magnitude of the effect of the drop-off option did. Table 4-7 shows that introducing a curbside recycling program increases the probability that the average household recycles over 95% of glass and plastic bottles by more than 50 percentage points; aluminum by more than 40 percentage points; and yard waste and newspaper by around 25 percentage points. The interpretation of the differences across materials is similar to that offered for the drop-off program variable. Bulky and potentially messy materials such as glass and plastic bottles are difficult to transport and thus more responsive to the introduction of curbside than are other materials. Also, the small percentage point response of yard waste to curbside recycling actually represents a fairly substantial percentage increase over baseline recycling levels. Table 4-7 also shows the marginal effects of replacing an existing drop-off recycling program with a curbside recycling program. The size of the difference is fairly similar for glass bottles, aluminum and plastic bottles. Replacing a drop-off program with a curbside program leads to roughly a 20% increase in the probability of recycling over 95% of these materials. For newspaper and yard waste, replacing a drop-off program with a curbside program increases the probability of recycling over 95% by about 5%. This is a small percentage change for newspaper (9%) and a slightly larger percentage change for yard waste (14%). Experience with a recycling program has a positive effect on recycling effort for newspaper; for yard waste, experience is significant only once the recycling program has lasted at least 2 years. Table 4-7 reports the marginal effect of having a program in place for more than two years versus having it in place between one and two years. The magnitudes are quite small. For yard waste, greater experience with recycling programs

PAGE 82

73 increases the probability that over 95% of it is recycled by less than 5%. In the case of newspaper, while program length has a positive effect on recycling effort, the coefficient on the indicator variable for a program of over two years in length is smaller than the coefficient on the indicator variable for a program of between one and two years in length. This means that the marginal effect of going from a program of 1 to 2 years in length to a program of over 2 years in length is actually negative, but only slightly so. The finding that recycling effort increases with experience is consistent with Reschovsky and Stone (1994) which finds that the probability of participating in recycling rises for newspaper, glass, plastic, cardboard, metal and composting when households feel knowledgeable about the recycling program. Our findings on the effects of other features of curbside recycling programs are mixed. The total number of materials collected curbside has a small, significant, positive effect on the intensity of newspaper recycling. Increasing the number of materials collected curbside by 1 leads to a 2.5% increase in the probability that a typical household will recycle over 95% of its newspaper waste. Making a curbside recycling program mandatory has no statistically discemable effect on intensity of recycling effort for any of the materials.'^ This finding is congruent with Kinnaman and FullertonÂ’s This finding could be attributable to a lack of enforcement of a mandatory recycling rule or law. If people perceive that the rule will not be enforced, then they have no incentive to comply. Unfortunately, we were unable to obtain systematic information on the enforcement of mandatory recycling requirements. The information we did locate was unclear about whether enforcement was for a mandatory recycling requirement or for curbside separation rules. Folz (Recycling policy and performance: Trends in participation, diversion, and costs, working paper, 2001) reports that the use of enforcement tactics increased from 37 percent of the cities in his sample in 1989 to 55 percent in 1996. The enforcement tactics used are described as refusing to pick up trash, tagging bins with instructions about proper recycling practice or issuing written warnings about improper separation of recyclables from other solid wastes. We conclude that there is evidence that mandatory requirements are sometimes enforced, however, we do not have a clear sense of how often they are enforced

PAGE 83

74 (2000) result that communities in states with mandated recycling do not recycle significantly greater quantities. Unit Pricing Policy VariablesÂ’Â’ The econometric results reported in Table 4-6 indicate that the price of disposal is not a significant determinant of intensity of household recycling effort for any of the materials. This finding suggests that increasing the price of disposal does not increase the intensity of recycling effort. There are several possible explanations why the data reveal no effect. First, the average price of disposal for the unit-pricing communities in our sample simply may be too low to create a response from our relatively high-income households. The sampleÂ’s median household income is approximately $40,000 per year, which equates to an hourly wage of roughly $20. At that wage level, if the amount of time associated with recycling 32 gallons of trash is more than 5.75 minutes then the household will have time costs of recycling that exceed the avoided $1.91 average disposal charge. Thus, as an incentive to recycle, unit pricing is ineffective. Second, a disposal price provides only an indirect signal to increase recycling, whereas it provides a direct signal to reduce trash. When faced with the prospect of paying a unit price for trash disposal, households may respond by changing their purchasing habits or making other changes in behavior that have a more direct impact on waste disposal. We return to this point in the conclusion of the paper. Â’Â’ Initially we set out to identify the effects of the level of the disposal price as well as other unit pricing program characteristics such as program type (bag/tag/sticker, subscription can) on the propensity to recycle. However, due to the small number of observations for bag/tag/sticker communities, we were unable to identify the effects of the type of unit-pricing program on the intensity of household recycling efforts.

PAGE 84

75 Finally, most of the unit pricing programs included in our sample are subscription can programs which provide a discontinuous signal to reduce disposal and therefore may provide only a weak incentive to households to recycle instead of disposing of solid waste (Nestor and Podolsky, 1998). Communities with bag/tag/sticker unit pricing programs may be more responsive to a given price level. Our finding of no effect of disposal price on recycling efforts is consistent with the findings of earlier studies by Kinnaman and Fullerton (2000), Fullerton and Kinnaman (1996) and Reschovsky and Stone (1994). All of these earlier studies find that unit pricing does not significantly affect the level of recycling or the probability of participation in recycling programs. However, our findings differ from those of Hong (1999), Callan and Thomas (1997) and Hong et al. (1993). For a sample of Korean households, Hong (1999) finds that a unit price has a significant positive effect on the recycling rate and that the elasticity of recycling with respect to price is approximately 0.5. Callan and Thomas (1997) finds that the presence of unit pricing increases the recycling rate by approximately 6.5 percent. Hong et al. (1993) indicates that unit pricing increases the probability that households will participate more often in recycling. Van Houtven and Morris ( 1 999) finds that the presence of unit pricing positively affects the probability that a household will participate in recycling but has no effect on the quantity of recyclables set out for collection. Socioeconomic Factors The econometric models also include a number of socioeconomic variables describing various characteristics of the households. The statistical significance and size of the effects of these variables on intensity of recycling effort vary substantially across materials. Below we discuss those variables that have a statistically significant effect.

PAGE 85

76 Household income has a significant and positive effect on intensity of recycling effort for newspaper only.'® We can calculate the “marginal” effects of moving from one income category to the next highest income category. For example, we find that for a typical household, moving from the “income between $35,000 and $49,999” category to the “income between $50,000 and $74,999” category leads to a 3.6% increase in the probability of recycling over 95% of all newspaper waste generated. The level of education attained by the most highly educated person in the household has a significant but small effect on intensity of recycling effort for all materials except plastic bottles and yard waste. The marginal effects for a typical household of moving from the “high school graduate” category to the “college graduate” category is to increase the probability of recycling over 95% of aluminum and newspaper by 0.1% and 1.5% respectively. Curiously, for glass bottles, the level of education has a small negative effect on intensity of recycling effort. A number of other socioeconomic variables also influence the intensity of yard waste recycling efforts. Increasing population density by 1000 persons per square mile leads to a 1.3% increase in the probability that a typical household recycles 10% or less of its yard waste. A likely reason is a growing scarcity of appropriate outdoor storage space as population becomes denser. Residents of single-family dwellings are substantially more likely to recycle larger quantities of their yard waste than residents of multi-family dwellings. Again, the reason might be a lack of outdoor or indoor storage This finding is in harmony with the theoretical results in Saltzman et al. (1993). They suggest that as long as newspaper is a normal good, because its recyclable content cannot be altered by the household, the impact of income on recycled newspaper should always be positive. The impact of income on other recyclable materials will be determined by whether the goods that are the source of the material are normal as well as to what extent the household can reduce the amount of a material associated with a good. For example, the glass content of beverage products can be reduced by switching to plastic or cardboard so that when income increases, glass recycling might decrease. Recall that the sample is more highly educated than one would expect of a randomly selected sample.

PAGE 86

77 space. Household size has a significant and positive effect on recycling efforts for glass bottles and yard waste. Increasing the number of occupants of the average household by 1 person leads to a 3% increase in the probability that the household will recycle over 95% of its glass bottle waste and a 2% increase in the probability of recycling over 95% of yard waste. This finding may be due to the fact that glass bottle and yard waste recycling are time intensive — bottles must be cleaned, yard waste must be bagged. As the number of occupants rises, the amount of time required from each occupant decreases thereby reducing the implicit cost on any one individual. Finally, age has a positive, but small, impact on intensity of recycling for all materials except glass bottles. Data Limitations A perennial problem with survey research is the nonresponse problem: are there systematic biases introduced into the data by the exclusion of those who failed to respond to the survey? A concern regarding our own data is that bias exists because respondents who are avid recyclers may have been more inclined to mail back the questionnaire. These individuals may recycle on their own initiative and thus, compared to the general population, be less responsive to recycling incentives. The methods for eorrecting such bias require information about the characteristics of the questionnaire recipients who did not respond, or, at a minimum, street addresses for those who did not respond.^® Unfortunately our efforts to obtain that information were futile. There is another reason why our data might mis-state the response to recycling incentives that would be expected from the general population. Avid recyclers may be better informed than others about the existence of recycling options. This problem is less Cameron et al. (1999) discuss econometric methods for dealing with non-response bias in mail surveys. Their method uses zip-code-level information for non-respondents combined with data collected from the survey to estimate a pooled-data probit model for response probability.

PAGE 87

78 of a concern for curbside recycling options that are quite visible to the least enthusiastic recyclers. The concern is larger for drop-off recycling options. If avid recyclers are more knowledgeable than other community members about all recycling options, their natural enthusiasm for recycling might make them appear less responsive to a drop-off program. In the absence of a drop-off option, avid recyclers are more likely than others to have already located recycling opportunities outside their own community. Thus, our estimate of the response to drop-off recycling incentives might be biased downward. However, it might also be biased upward. The general population might be less responsive than avid recyclers to drop-off programs simply because they are less likely to be aware of a programÂ’s existence. The net result of these two opposite sources of bias is unclear. Future research could clarify this uncertainty by analyzing all households rather than only households who report awareness of program existence. Our data set lacks information on community characteristics that might influence the intensity of household recycling effort for all members of a community. Communitylevel variables such as measures of recycling promotion activities or the general attitude toward environmental issues should be included in our equations because omitting these variables can cause a problem of endogeneity. Of particular concern is that the excluded community level variables might be correlated with the dependent variable, which would result in biased coefficient estimates for the included independent variables. To address these concerns, we test for the significance of regional indicator variables which are included to capture unobserved community-level heterogeneity. The results of F-tests To examine the sensitivity of the results to the use of an ordered logit-model speeification with fixed effects, we estimated a binomial logit model with fixed effects using a conditional maximum-likelihood estimator. For this model we partitioned the dependent variable into two groupsrespondents recycling between 0% and 10% of a material and those recycling greater than 10%. The sign and significance of the

PAGE 88

79 suggest that MSA level indicator variables as a group are significant determinants of recycling intensity for each material. Conclusion and Policy Implications This study uses a unique household-level data set, representing primarily middle and upper income households in 20 MSAs across the country, to examine the effect of two popular solid waste programs, curbside recycling and unit pricing, on the percent recycled of five different materials found in the municipal solid waste stream: glass bottles, plastic bottles, aluminum, newspaper, and yard waste. The study also assesses the impact of other attributes of recycling programs (e.g., mandatory or voluntary) along with socioeconomic characteristics of households on recycling activity. The results presented here provide new insights that could help policy makers to improve the costeffectiveness of a communityÂ’s recycling program and to design a program to achieve mandated recycling rate goals. Consistent with expectations, a curbside recycling program increases householdsÂ’ intensity of recycling and the results differ across recyclable materials. The effect of a unit pricing program, on the other hand, is less clear. The analysis indicates that drop-off and curbside recycling programs increase householdsÂ’ intensity of recycling for the five materials. The magnitude of the effect of these programs varies dramatically across materials with the largest impacts on glass and plastic bottles. The size of the impact on yard waste recycling effort is also large relative to the average intensity of recycling effort observed in the sample. We conclude that policy variables were very similar for all the equations except the one for newspaper. For it, many of the significant policy variables became insignificant under the binomial specification except the curbside indicator variable which remained significant. There were more differences in the sign and significance of the socioeconomic independent variables. In many cases, a variable that was significant under the ordered logit specification was insignificant under the binomial logit specification. The sign changes were only for insignificant variables. Overall, we chose to focus on the ordered logit results because the advantages from the third partition seemed to outweigh the disadvantages of modeling fixed effects with MSA indicator variables.

PAGE 89

80 introducing a local recycling option has a smaller impact on materials, such as newspaper and aluminum, for which there were recycling options such as charity drives or workplace or other away-from-home recycling stations even before the local drop-off program. We further conclude that introducing a local drop-off program has a greater impact on materials such as glass and plastic bottles whose transportation and storage would be most difficult for households. Local governments should take this finding into consideration when selecting which materials to include in a recycling program. Curbside recycling programs have a bigger effect on behavior than drop-off programs. For three of the materials, a curbside program increases the probability that the average household recycles over 95% by approximately 20% more than the increase generated by a drop-off program. Nonetheless, drop-off programs also are effective at increasing recycling. A budget-constrained community with no recycling program at all could see measurable waste diversion with the introduction of a less expensive drop-off alternative. Local governments considering implementing curbside recycling could compare the benefits of the expected increase in recycling activity to the incremental costs of implementing curbside as opposed to drop-off recycling. The impact of unit pricing on the intensity of recycling effort for specific materials is less clear. Unit pricing gives a direct incentive to decrease waste quantities. In response to such a program, households might adjust their consumption towards goods that generate easy-to-recycle wastes, likely those wastes eligible for collection by a local recycling program.^^ These easy-to-recycle wastes inerease in quantity; however, the Such an adjustment is suggested by Hong (1999) which finds that the price elasticity of total waste quantities is positive but the price elasticity of non-recyclable waste quantities is negative.

PAGE 90

81 percentage of that quantity that is recycled might not change. If unit pricing does increase recycling quantities by shifting consumption toward materials that are collected by a communityÂ’s recycling program, its impact on recycling will not be detected by examining the percent of a material a household recycles. Our findings indicate that the added convenience created by a recycling program creates a stronger incentive to recycle than having to pay at the margin for trash disposal. Of course, the levels of unit prices charged are important to the impact of the program. The mean fee for our sample was $1.91 per 32-gallon container. At these price levels, collecting more materials at curbside will produce greater waste diversion than will implementing unit-pricing. However, if the costs of adding a particular material to a curbside program exceed the waste diversion and recycling revenue benefits of doing so, then adding materials may not be worthwhile. Recycling programs appear to become more effective over time. Greater experience with a recycling program leads to increased recycling effort directed at newspapers and yard waste. However, the magnitudes of these effects are quite small. Of interest to policy-makers perhaps is that this effect is not negative; that is, households do not appear to become less enthusiastic over time about participating in recycling. Of course, which materials to include in a recycling program also depends on the market prices of recyclable materials and on collection and processing costs. For example, our findings suggest that introducing curbside recycling has a big effect on the recycling of plastic bottles, one of the highest valued materials of those we studied.^^ However, collection and transportation costs are also high for plastic bottles due to their The following are average prices recyclers were paying for materials in late January or early February, 2000 in 8 urban centers across the country: Aluminum cans -$750 per ton; Natural HOPE (a type of plastic container) $300 per ton; Newspaper number 8 $70 per ton; Amber glass $27 per ton (Truini, 2000).

PAGE 91

82 low density. Policy makers can combine the insights from this study with information on the material composition of their local waste stream, local collection and transportation costs and current market prices for recyclable plastic to decide if curbside recycling is a cost-effective means of managing plastic waste. The study suggests several issues for future research. First, due to a lack of variation in our data, we were unable to analyze the differences in responses to the two main approaches to implementing unit pricing for solid waste disposal services: bag/tag/sticker versus subscription can. Van Houtven and Morris (1999) and Nestor and Podolsky (1998) analyze data from Marietta, Georgia and find that there are differences and that a bag program causes larger reductions in waste quantities than a subscription can program. Future research could explore if the different program types affect recycling of different materials in different ways. Second, the nature of our data set has limited us to focusing on recycling intensity (percentage of each material type generated by the household that is recycled). However, policy makers and solid waste planners ultimately need more information about how recycling program characteristics and unit prices affect material-specific quantities of both recycling and waste disposal by households. Providing such information requires national household-level data on quantities of materials recycled and discarded. Third, research into the costs of implementing curbside recycling programs with different scopes compared to the costs of implementing a drop-off program as well as a unit pricing program would be useful to policy makers seeking to design effective and efficient waste management strategies.

PAGE 92

83 Finally, our data sample is focused on middle and upper income households in urban and suburban areas of the U.S. Therefore, our conclusions are applicable to these types of households. Future research is needed to identify the implications of solid waste and recycling policies for the recycling behavior of lower income households, of households in more rural areas in the U.S., and of households in other countries.^'* Hong (1999) has analyzed data for Korea, Sterner and Bartelings (1999) for Sweden.

PAGE 93

84 Table 4-1. Previous studies on effects of unit pricing and recycling programs on effort Author(s) (year) Dependent Variable Independent Policy Variables Data Aggregate or Material Specific Recycling Quantities Unit Price Recycling Program Attributes National or Regional Household Level Van Houtven and Morris (1999) Aggregate (quantity is weight, not volume) No, but dummy for presence of each of two types of unit pricing program No Regional Marietta, Georgia Yes Hong (1999) Aggregate Yes No National Korea Yes Hong and Adams (1999) Aggregate Yes No Regional Portland, Oregon Yes Sterner and Bartelings (1999) By Material (community proportion, not quantity recycled) No Yes Regional Southwest Sweden No Kinnaman and Fullerton (2000) Aggregate Yes Yes National No Nestor and Podolsky (1998) Aggregate Yes No Regional Marietta, Georgia Yes Callan and Thomas (1997) Aggregate (percent of total waste stream recycled) No, but dummy for presence of unit pricing program Yes Regional Massa-chusetls No Fullerton and Kinnaman (1996) Aggregate Yes No Regional Charlottes-ville Yes Rechovsky and Stone (1994) By Material (proportion, not quantity, recycled) No, but dummy for presence of unit pricing program Yes Regional upstate NY Yes Hong et al. (1993) Aggregate (recycling participation yes/no) Yes No Regional Portland Yes Salt 2 man et al. (1993) By Material No Yes Regional PA andNJ No

PAGE 94

85 Table 4-2. Metropolitan Statistical Areas sampled Boston/Hartford Corridor Detroit New York Metro (New Jersey side) Philadelphia Minneapolis/St. Paul Atlanta San Francisco Phoenix Houston Tampa New York City Metro (New York and Connecticut) Portland Camden, New Jersey Chicago Seattle St. Louis Los Angeles Dallas-Fort Worth Miami Denver Boston/Hartford Corridor

PAGE 95

86 Table 4-3. Unit pricing programs MSA Number of Communities with Unique Unit Price Number of Observations Program Type Los Angeles 1 1 Subscription San Francisco 8 20 Subscription Chicago 7 10 Bag/Tag/Sticker Detroit 1 1 Bag/Tag/Sticker Minneapolis/St. Paul 2 2 Subscription Portland 5 46* Subscription Philadelphia 1 2 Bag/Tag/Sticker Seattle 7 34** Subscription *Of the 46 households living in the Portland MSA, 37 are located in the city of Portland. **Of the 34 households living in the Seattle MSA, 1 8 are located in the city of Seattle.

PAGE 96

87 Figure 4-1. Distribution of unit prices

PAGE 97

88 Table 4-4. Summary statistics for independent variables in recycling logit regressions Variable Mean Standard Deviation PRICE-SW ($ per 32-gal.) $1.91 $0.86 PRICE-YW ($ per 32-gal) SI. 90 $0.86 NEWSPAPER CURB 0.916 0.278 GLASS BOTTLES CURB 0.886 0.318 ALUM CURB 0.853 0.355 PLASTIC BOTTLES CURB 0.775 0.417 YARD WASTE CURB 0.528 0.500 TOT MATERIALS CURB 3.900 1.200 NEWSPAPTER DROP 0.056 0.229 GLASS BOTTLES DROP 0.071 0.256 ALUM DROP 0.104 0.305 PLASTIC BOTTLES DROP 0.127 0.333 YARD WASTE DROP 0.057 0.232 MAND RECY PGM 0.528 0.500 RECY PGM 1 TO 2 3.900 1.200 RECY PGM > 2 0.056 0.229 POP DENS 5.820 5.923 INC 10000 TO 14999 0.068 0.251 INC 15000 TO 24999 0.133 0.339 INC 25000 TO 34999 0.135 0.342 INC 35000 TO 49999 0.208 0.406 INC 50000 TO 74999 0.258 0.438 INC > 75000) 0.140 0.347 HOUSEHOLD SIZE 2.700 1.400 HOUSEHOLD HEAD AGE 47.900 15.900 DETACHED HOUSE) 0.726 0.446 HOME OWNERSHIP 0.793 0.405 EDUC HS GRAD 0.511 0.500 EDUC COLL GRAD 0.247 0.431 EDUC BEYOND COLL 0.195 0.397 Note: In the logit equations, this PRJCE-SW and PRICE-YW are in dollars per gallon. Also, the price variables are only for the 116 households living in programs with unit pricing programs. The mean values for the material-specific variables use the numer of observations in the relevant logit regressions.

PAGE 98

89 Table 4-5. Proportions of materials recycled Material Percentage and Number of Respondents Recycling 0 to 10% Percentage and Number of Respondents Recycling 1 1 to 95% Percentage and Number of Respondents Recycling Over 95% Total Number Missing Newspaper 8.8% 16.6% 74.6% 100% 92 173 111 1042 7 Glass Bottles 11.3% 22.2% 66.5% 100% 117 229 687 1033 16 Aluminum 15.0% 21.8% 63.2% 100% 152 221 639 1012 37 Plastic Bottles 17.8% 28.0% 54.2% 100% 180 284 550 1014 35 Yard Waste 43.3% 22.8% 33.9% 100% 417 220 326 963 86

PAGE 99

90 Table 4-6. Recycling participation ordered logits Independent Variables Newspaper Glass Bottles Aluminum Plastic Bottles Yard Waste DROP 0.7886* 1.5184*** 0.8617** 2.1350*** 1.1074*** (0.4594) (0.4137) (0.3642) (0.3984) (0.1896) CURB 1.1423*** 2.1026*** 1.7808*** 2.9083*** 1.3111*** (0.4164) (0.4733) (0.3577) (0.4603) (0.1896) MAND RECY PGM*CURB 0.2467 -0.0379 0.2041 0.2226 0.1435 (0.2030) (0.1070) (0.1797) (0.1621) (0.2037) TOT MATERIALS CURB 0.1434* -0.0917 -0.1172 -0.0831 0.0677 (0.0772) (0.0612) (0.0914) (0.0790) (0.0659) RECY PGM 1 TO 2 0.6111*** 0.1668 0.1395 0.3272* 0.2684 (0.1907) (0.1312) (0.1712) (0.1670) (0.1821) RECY PGM > 2 0.5614*** 0.1737 0.2146 0.1300 0.4520** (0.2091) (0.1518) (0.1896) (0.1745) (0.2017) PRICE -11.2608 0.1227 -4.7538 1.6401 -4.6836 (7.8743) (3.3013) (6.1760) (5.0294) (5.1007) POP DENS 0.0056 0.0000 -0.0083 -0.0044 -0.0511*** (0.0179) (0.0106) (0.0151) (0.0140) (0.0188) INC 10000 TO 14999 0.9985** 0.4387 0.9363** 1.2334*** -0.3413 (0.4170) (0.2675) (0.4147) (0.3784) (0.4128) INC 15000 TO 24999 0.4590 -0.0654 -0.1214 0.1881 -0.1285 (0.3506) (0.1999) (0.3403) (0.2809) (0.3545) INC 25000 TO 34999 0.6408* 0.0380 -0.1043 0.6419** -0.3247 (0.3637) (0.2007) (0.3483) (0.2941) (0.3634) INC 35000 TO 49999 0.8246** 0.0075 0.0772 0.4555 -0.4338 (0.3575) (0.1972) (0.3309) (0.2807) (0.3568) INC 50000 TO 74999 1.0418*** 0.1268 0.0312 0.4673 -0.1164 (0.3762) (0.2112) (0.3457) (0.2977) (0.3651) INC > 75000 1.2203*** 0.0689 0.0915 0.5157 -0.2746 (0.4277) (0.2302) (0.3838) (0.3260) (0.3986) HOUSEHOLD SIZE 0.0437 0.0901** 0.0070 0.0717 0.1010* (0.0646) (0.0406) (0.0543) (0.0491) (0.0557) HOUSEHOLD HEAD AGE 0.0175*** 0.0049 0.0092* 0.0097** 0.0134** (0.0059) (0.0034) (0.0053) (0.0048) (0.0055) DETACHED HOUSE -0.0065 0.0710 0.1708 0.4021** 0.7399*** (0.2068) (0.1114) (0.1817) (0.1694) (0.1880) HOME OWNERSHIP 0.2651 0.3517** 0.3690* 0.2911 0.3779* (0.2118) (0.1433) (0.1964) (0.1841) (0.2165) EDUC HS GRAD 1.1600*** 0.4879** 1.0557*** 0.0862 -0.3421 (0.3265) (0.2095) (0.3112) (0.2662) (0.3466) EDUC COLL GRAD 1.2460*** 0.4711** 1.0612*** 0.0533 -0.4060 (0.3641) (0.2251) (0.3443) (0.2879) (0.3828) EDUC BEYOND COLL 1.2520*** 0.5269** 0.9415*** 0.2480 -0.2598 Constant (0.3802) (0.2388) (0.3531) (0.3037) (0.3894) -3.4140*** -2.1129*** -1.7563*** -2.8715*** -2.6362*** (0.7327) (0.6027) (0.6690) (0.6885) (0.6427) Number of observations 1042 1033 1012 1014 963 Log likelihood -693.146 -781.65 -849.992 -867.607 -848.263 Chi squared statistic 137.609*** 196.779*** 136.450*** 282.917*** 357.350*** Heteroskedasticity corrected Yes Yes Yes Yes Yes * 10% level of significance ** 5% level of significance Standard errors in parentheses. 1% level of significance

PAGE 100

91 Table 4-7. Marginal effects of significant policy variables from logits Policy Variable Newspaper Glass Bottles Aluminum Plastic Bottles Yard Waste Total Materials Cubside Recycle 0 10% Recycle 1 1 95% Recycle over 95% Drop-off -0.0090 [-0.1020] -0.0162 [-0.0974] 0.0252 [0.0338] Recycle 0 10% -0.0932 -0.5467 -0.1900 -0.5291 -0.2696 [-1.0559] [-4.8268] [-1.2647] [-2.9804] [-0.6226] Recycle 1 1 95% -0.0923 0.1284 -0.0027 0.1980 0.0743 [-0.5562] [0.5791] [-0.0123] [0.7069] [0.3252] Recycle over 95% 0.1856 0.4183 0.1926 0.3311 0.1953 Curbside (not mandatory) [0.2489] [0.6290] [0.3051] [0.6104] [0.5770] Recycle 0 10% -0.1198 -0.6427 -0.3210 -0.6404 -0.3143 [-1.3567] [-5.6743] [-2.1375] [-3.6075] [-0.7258] Recycle 1 1 95% -0.1347 -0.0028 -0.0941 0.0987 0.0718 [-0.8112] [-0.0126] [-0.4308] [0.3525] [0.3141] Recycle over 95% 0.2545 0.6455 0.4151 0.5417 0.2425 Drop-off to curbside [0.3413] [0.9706] [0.6574] [0.9986] [0.7164] Recycle 0 10% -0.0266 -0.0960 -0.1311 -0.1113 -0.0446 [-0.3008] [-0.8474] [-0.8728] [-0.6270] [-0.1031] Recycle 1 1 95% -0.0423 -0.1312 -0.0914 -0.0993 -0.0025 [-0.2550] [-0.5917] [-0.4185] [-0.3545] [-0.0111] Recycle over 95% 0.0689 0.2271 0.2225 0.2106 0.0472 Program length over 2 years Recycle 0 10% Recycle 1 1 95% Recycle over 95% [0.0924] 0.0027 [0.0310] 0.0053 [0.0318] -0.0080 [0.3415] [0.3524] [0.3882] [0.1394] -0.0440 [-.1016] 0.0059 [0.0258] 0.0381 Note: Numbers in brackets convert the marginal effect into a percentage change from the average intensity of recycling effort observed in the sample. For total materials curbside, the marginal effect is calculated assuming a one-unit increase in the total number of materials recycled curbside. For binary indicator variables, marginal effects are calculated by solving the model once with the significant indicator variable of interest set at one and all other variables set at their mean value, solving again with the indicator variable of interest set at zero and all other variables set at their means, and then taking the difference. The marginal effect for drop-off (curbside) is calculated with the curbside (drop-off) dummy variable set at zero. The “drop-off to curbside” marginal effect is defined as the difference between the marginal effect of curbside and the marginal effect of drop-off. For program length, the marginal effect gives the difference between having a program in place between one and two years and having a program for more than two years. The sum of marginal effects may not equal zero due to rounding.

PAGE 101

CHAPTER 5 CONCLUSION Chapters 2, 3, and 4 use applied econometric techniques and microeconomic analysis to investigate current economic issues in solid-waste management and recycling. Each chapter is specific in addressing a particular issue and has implications for public policy. I provide further explanation of the main ideas and results from each chapter and associated public policy concerns. Economic analysis is not only useful in analyzing the behavior of consumers and businesses; it also provides a framework for analyzing decisions made by policy makers. Chapter 2 makes a significant contribution in this direction within the literature of solidwaste and recycling. Absent in the economic literature is empirical analysis of the factors influencing states in adopting solid-waste management and recycling policies. My chapter fills this void by using a dataset for the U.S. states from 1988 to 1999 to examine the adoption of a wide range of solid-waste management and recycling policies during a time the U.S. Environmental Protection Agency implemented stricter guidelines on landfills. States have been aggressive in passing pro-environmental legislation in response to changes in federal policies and market changes in the solid-waste and recycling industry. As a direct method of limiting waste from being deposited in landfills, states are passing landfill material bans to prevent certain materials from being deposited in landfills. Also, states are passing “market development initiatives” such as recycling loans, grants, and tax incentives to stimulate recycling. In addition, states are passing 92

PAGE 102

93 laws to require local governments to implement recycling programs. Using a variety of covariates including socioeconomic controls, I use variables explaining the position of the median voter selecting these pro-environmental policies. A variable constructed to measure the preferences of state leaders for government regulation shows that greater Republican control of state government results in less proenvironmental solid-waste and recycling policies being implemented. One measure of environmental preferences of state citizenry is the percentage of the state population with membership in the National Audobon society. Greater membership generally results in more recycling policies being adopted. In addition, states with a higher percentage of the state population age 65 or older tend to provide more landfill bans and have a higher probability of having recycling grants or loans. Absent from my analysis is the evaluation of the costs and benefits associated with implementation of these policies. Whether or not the pro-environmental solid-waste management and recycling policies are adopted or kept in place when they result in net costs as the median voter shifts toward the aforementioned position remains open for further research. Chapter 3 uses applied microeconomics to analyze business behavior in the solidwaste industry by analyzing the broader economic questions are how government regulations, competition, and costs affect pricing behavior. Tipping fee regressions are estimated for a national dataset of landfills in the United States using yearly data from 1992 to 1999 with specific covariates for regulatory, competition, and cost effects. Implementing landfill bans for automobile tires and motor oil reduce tipping fees, while automobile battery bans, white good bans, and yard waste bans raise tipping fees.

PAGE 103

94 Competition, measured by the number of nearby competitors, surprisingly does not decrease tipping fees. A precipitation variable proxying the cost of leachate collection suggests tipping fees are higher in areas with greater leachate collection costs. In addition, labor costs and land costs positively affect tipping fees. Finally, landfills located in counties bordering coastal areas have much higher tipping fees. Collectively, the results from Chapter 3 provide a framework for policy makers to consider when locating new landfills in relation to expected tipping fees. If lower tipping fees are desired, landfills should be located in areas which are relatively drier and away from coastal areas. In addition, lower labor costs and land prices can reduce overall tipping fees. The ban on automobile tires seems to be the most effective landfill material ban resulting in a decrease in expected tipping fees. Further research into the costs of solid-waste management could use the results from Chapter 3 with empirical estimates on such externalities as noise and odor associated with landfilling to derive marginal social cost schedules for landfilling. Understanding the social costs of landfilling will enable researchers to find integrated solid-waste management and recycling policies that reach optimum economic efficiency. Chapter 4 extends the literature in recycling by examining the effectiveness of curbside recycling and unit pricing in encouraging household recycling of particular materials using a household-level data set covering 20 MSAs in the United States using ordered logit regressions. Both drop-off and curbside recycling programs are found to increase recycling behavior, with the largest effects coming from curbside recycling availability. The effect of a unit pricing program on recycling behavior is not clear. The magnitude of the policy effects vary across different materials, with the largest impacts

PAGE 104

95 on glass and plastic bottles. When examining the effects of particular attributes of the recycling programs, households recycle more under programs that have been in existence longer. For three (glass bottles, aluminum, plastic bottles) of the five materials recycled (plastic bottles, glass bottles, yard waste, aluminum, newspaper), curbside program availability increases the probability of recycling over 95% by about 20% more than if a drop-off program was available. Added convenience of recycling is found to be a greater incentive to recycle than households being charged more to dispose of more trash. With the mean unit prices about $1.91 per 32-gallon container in the sample, an increase in curbside collection of additional recyclable materials will increase waste diversion more than implementing unit pricing. This would be the reasonable policy option to pursue if the waste diversion and recycling revenue benefits exceed the cost of adding another material to curbside collection. On the recycling side, there are collection, processing, and transportation costs in addition to market prices for recyclable materials to consider. Cost implementation of unit pricing must also be considered. Common to both recycling and unit pricing programs are the effects of scope and scale of implementation and operation. Future research in the economics of solid-waste management and recycling could integrate results from the previously discussed chapters with new research on benefits and costs to examine optimal local landfilling and recycling policies in dynamic frameworks. Local governments need to have definitive answers about policy options resulting in positive net benefits in the long-run. This study provides some important results to proceed in this direction.

PAGE 105

LIST OF REFERENCES Aadland, D. & Caplan, A. (1999). Household valuation of curbside recycling. Journal of Environmental Planning and Management, 42, 781-799. Callan, S.J. & Thomas, J.M. (1997). The impact of state and local policies on the recycling effort. Eastern Economic Journal, 23, 41 1-23. Callan, S.J. & Thomas, J.M. (1999) Adopting a unit pricing system for municipal solid waste; Policy and socio-economic determinants. Environmental and Resource Economics, 14, 503-518. Cameron, T.A., Shaw, W.D., & Ragland, S.R. (1999). Nonresponse bias in mail survey data: salience vs. endogenous survey complexity. In Valuing recreation and the environment: Revealed preference pethods in theory and practice. Cheltenham, UK: Edward Elgar. Clayton, K.C. & Huie, J.M. (1973). Solid wastes management: The regional approach Cambridge, MA: Ballinger Publishing Company. Council of State Governments (1989). The book of the states 1988-1989. Lexington, KY : Council of State Governments. Council of State Governments (1991). The book of the states 1990-1991. Lexington, KY: Council of State Governments. Council of State Governments (1993). The book of the states 1992-1993. Lexington, KY : Council of State Governments. Council of State Governments (1995). The book of the states 1994-1995. Lexington, KY : Council of State Governments. Council of State Governments (1997). The book of the states 1996-1997. Lexington, KY : Council of State Governments. Council of State Governments (1999). The book of the states 1998-1999. Lexington, KY : Council of State Governments. Council of State Governments (2001). The book of the states 2000-2001 . Lexington, KY : Council of State Governments. 96

PAGE 106

97 Feiock, R.C. & West, J.P. (1993). Testing competing explanations for policy adoption: Municipal solid waste recycling programs. Political Research Quarterly, 46, 399-419. Glenn, J. (1992a). The state of garbage in America. BioCycle, 33, April, 46-55. Glenn, J. (1992b). The state of garbage in America (part II). BioCycle, 33, May, 30-37. Fullerton, D. & Kinnaman, T.C. (1996). Household responses to pricing garbage by the bag. The American Economic Review, 86, 971-84. Glenn, J. (1998a). The state of garbage in America. BioCycle, 39, April, 32-43. Glenn, J. (1998b). The state of garbage in America (part II). BioCycle, 39, May, 48-52. Glenn, J. (1999). The state of garbage in America. BioCycle, 40, 60-71. Goldstein, N. (1997a). The state of garbage in America. BioCycle, 38, April, 60-67. Goldstein, N. (1997b). The state of garbage in America (part II). BioCycle, 38, May, 7175. Goldstein, N. (2000a). The state of garbage in America. BioCycle, 41, April, 32-39. Goldstein, N. & Madtes, C. (2000). The state of garbage in America. BioCycle, 41, November, 40-48. Harvey, A.C. (1976). Estimating regression models with multiplicative heteroscedasticity. Econometrica, 44, 461-466. Hong, S. (1999). The effects of unit pricing system upon household solid waste management: The Korean experience. Journal of Environmental Management, 57, 1-10. Hong, S., Adams, R.M., & Love, H.A. (1993). An economic analysis of household recycling of solid wastes: The case of Portland, Oregon. Journal of Environmental Economics and Management, 25, 136-146. Hong, S. & Adams, R.M. (1999). Household responses to price incentives for recycling: Some further evidence. Land Economics, 75, 505-514. Hsaio, C. (1986). Analysis of panel data. Cambridge, UK: Cambridge University Press. Jakus, P.M., Tiller, K.H., & Park, W.M. (1996). Generation of recyclables by rural households. Journal of Agricultural and Resource Economics, 21, 96-108.

PAGE 107

98 Jenkins, R.R. (1993). The economics of solid waste reduction. Hants, England: Edward Elgar Publishing Limited. Kahn, M.E. (2000). Demographic change and the demand for environmental regulation. Journal of Policy Analysis and Management, 21, 45-62. Kenny, L. (1982). A model of optimal plant size with an application to the demand for congnitive achievement and for school size. Economic Inquiry, 20, 240-253. Kinnaman, T.C. & Fullerton, D. (1995). How a fee per-unit garbage affects aggregate recycling in a model with heterogeneous households. In Public economics and the environment in an imperfect world. Dordrecht, The Netherlands: Kluwer Academic Publishers. Kinnaman, T.C. & Fullerton, D. (2000). Garbage and recycling with endogenous local policy. Journal of Urban Economics, 48, 419-442. Kinnaman, T.C. (2000). Explaining the growth in municipal recycling programs: The role of economic and non-economic factors. Public Works Management and Policy, 4, 37-51. Kreith, F. (1994). Handbook of Solid Waste Management. New York, NY: McGraw Hill. McFadden, D. (1974). Conditional Logit Analysis of Qualitative Choice Behavior. In Frontiers in econometrics. New York, NY: Academic Press. Menell, P. (1990). Beyond the throwaway society: An incentive approach to regulating municipal solid waste. Ecology Law Quarterly, 17, 655-739. Miranda, M.L. & Aldy, J.E. (1998). Unit pricing of residential municipal solid waste: Lessons from nine case study communities. Journal of Environmental Management, 52, 79-93. Miranda, M.L., Everett, J.W., Blume, D., and Roy Jr., B.A. (1994). Market-based incentives and residential municipal solid waste. Journal of Policy Analysis and Management, 13, 681-698. Morris, G. & Holthausen, D. (1994). The economics of household solid waste generation and disposal. Journal of Environmental Economics and Management, 26, 215234. Mrozek, J.R. (2000). Changes over time in the decision to adopt curbside recycling. Atlantic Economic Journal, 28, 239-253.

PAGE 108

99 Nestor, D.V. & Podolsky, M.J. (1998). Assessing incentive-based environmental policies for reducing household waste disposal. Contemporary Economic Policy, 16, 401-411. Podolsky, M.J. & Spiegel, M. (1998). Municipal waste disposal: Unit pricing and recycling opportunities. Public Works Management and Policy, 3, 27-39. Ready, M.J. & Ready, R.C. (1995). Optimal pricing of depletable, replaceable, resources: The case of landfill tipping fees. Journal of Environmental Economics and Management, 28, 307-323. Reschovsky, J.D. & Stone, S.E. (1994). Market incentives to encourage household waste recycling: Pay for what you throw away. Journal of Policy Analysis and Management, 13, 120-139. Saltzman, C., Duggal, V.G., & Williams, M.L. (1993). Income and the recycling effort: A maximization problem. Energy Economics, 15, 33-38. Sterner, T. & Bartelings, H. (1999). Household waste management in a Swedish municipality: Determinants of waste disposal, recycling and composting. Environmental and Resource Economics, 13, 473-491. Steuteville, R. (1994a). The state of garbage in America. BioCycle, 35, April, 46-52. Steuteville, R. (1994b). The state of garbage in America (part II). BioCycle, 35, May, 30-36. Steuteville, R. (1995a). The state of garbage in America. BioCycle, 36, April, 54-63. Steuteville, R. (1995b). The state of garbage in America (part II). BioCycle, 36, May, 30-37. Steuteville, R. (1996a). The state of garbage in America. BioCycle, 37, April, 54-61. Steuteville, R. (1996b). The state of garbage in America (part II). BioCycle, 37, May, 35-42. Steuteville, R. & Goldstein, N. (1993). The state of garbage in America. BioCycle, 34, May, 42-50. Steuteville, R., Goldstein, N., & Grotz, K. (1993). The state of garbage in America. BioCycle (part II). BioCycle, 34, June, 32-37. Truini, J. (2000). Waste news commodity pricing report. Waste News, 5, February 14, 30.

PAGE 109

100 U.S. Environmental Protection Agency, Office of Solid Waste (1990). Charging households for waste collection and disposal: The effects of weight or volumebased pricing on solid waste management. EPA/530-SW-90-047, September. U.S. Environmental Protection Agency (1993). Office of Solid Waste guide to EPAÂ’s unit pricing database: Pay-as-you-throw municipal solid waste programs in the U.S. EPA/230-B-93-002, April. U.S. Environmental Protection Agency (1997). Office of Solid Waste, municipal solid waste factbook, version 4.0. Washington, DC. Van Houtven, G.L. and Morris, G.E. (1999). Household behavior under alternative payas-you-throw systems for solid waste disposal. Land Economics, 75, 515-537.

PAGE 110

BIOGRAPHICAL SKETCH Salvador A. Martinez obtained his B.S. in business eeonomics from Weber State University in Ogden, Utah during 1996 with minors in mathematics and communications He taught at Weber State University in the spring and summer of 1997 before entering the Ph.D. program at the University of Florida that fall. His research interests generally include public policy topics in applied microeconomics, with special interests in environmental regulation, public utilities regulation, and public choice. 101

PAGE 111

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. i awrence W. Kenny, Chair ^ / Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fiilly adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. id N. Figlio Associate Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. H (J ^ Steven M. Slutsky Professor of Economies I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Donna J. Lee;^ Associate Professor of Food and Resource Economies This dissertation was submitted to the Department of Economics in the College of Business Administration and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. August 2004 Dean, Graduate School