Abstract
With worldwide concern for how and where to dispose of household waste, policy makers are increasingly looking for tools to efficiently and effectively reduce the amount of waste households produce. Using a comprehensive household-level data set involving 10,251 respondents from a crosssection of 10 countries (Australia, Canada, Czech Republic, France, Italy, Korea, Mexico, Netherlands, Norway, and Sweden), we examine waste policy, recycling behavior, and waste prevention. Unlike previous work, we empirically make comparisons across countries, incorporate attitudinal characteristics and a wide range of policy instruments, and allow for interdependence of decisions about recycling different materials. (JEL H23, Q58)
I. Introduction
In recent years, the issue of how a society should deal with municipal solid waste has become an important policy problem. Despite an increasing awareness of the external effects of waste generation and a growing resistance by society to the development of new landfills and incineration facilities, municipal solid waste has grown drastically over the last decades as a result of higher incomes, more intensive use of packaging materials and disposable goods, and increased purchases of durable material goods.1 Projections suggest solid waste will continue to grow despite current efforts to reduce the material content of products and to stimulate the reuse of products and packaging and the recycling of materials and substances.
In response to the increasing environmental pressures of municipal waste, many countries have begun to explore ways of reducing and disposing of it more effectively. In targeting one of the main sources of municipal waste, household or residential waste,2 municipal governments (which tend to be responsible for carrying out waste management and recycling services and for developing waste management programs and can thus have much influence on waste reduction through policies and legislative measures) have grown particularly interested in experimenting with unit pricing systems and improving recycling services. In the United States, for example, the number of jurisdictions with some sort of pay-as-you-throw or unit pricing program increased from about 1,000 in 1993 to almost 7,100 in 2006, or about 25% of all U.S. communities (Skumatz and Freeman 2007); in Canada, the share of households with access to at least one type of recycling program increased from about 70% in 1994 to 93% in 2006 (Statistics Canada 2008).
To assist policy makers in the design of efficient policies that effectively induce households to minimize waste through recycling and/or waste prevention, a better understanding of household behavior is necessary. To this end, a new activity on Household Behavior and Environmental Policy was initiated in 2005 by the OECD that covered not only waste generation and recycling but four other areas of household consumption identified as important environmental policy targets, namely, energy use, organic food consumption, personal transport, and water use. As part of the activity, a questionnaire on environment-related household behavior covering each of the above five areas was designed and a web-based access panel was used in early 2008 to implement the household survey in 10 countries representing the three OECD regions (North America, Europe, and Asia-Pacific): Australia, Canada, Czech Republic, France, Italy, Korea, Mexico, Netherlands, Norway, and Sweden. Approximately 1,000 households per country participated in the study (10,251 in total) providing information on sociodemographic and attitudinal characteristics and on policy variables in each of the areas under consideration.
In this paper, the 2008 OECD householdlevel data set is employed to examine several questions pertaining to recycling behavior and waste prevention, including (1) whether user fees for waste disposal have significant effects on waste recycling rates relative to flat fees, and whether these effects vary significantly by material and/or by type of unit pricing; (2) whether the presence of a recycling program strengthens or weakens the impact of a user fee system on recycling, and, if so, whether there is significant variation across materials; (3) the extent to which household waste recycling decisions depend on attributes of recycling programs, and whether there is significant variation across materials; (4) how general attitudes toward the environment influence waste recycling levels, and whether the presence of economic incentives and/or other forms of governmental intervention erodes or enhances the relevance of intrinsic motivation; (5) whether user fees have significant effects on waste prevention relative to flat fees.
The empirical literature on municipal waste management is mostly concerned with waste production and recycling decisions and focuses on the effects of sociodemographic variables and unit pricing systems on such decisions, although there are some recent attempts to quantify the role of attitudes and the importance of cultural and social influences in the decision-making process. In general, there is some agreement that user fees for waste disposal, mostly bag-based systems, are effective at reducing waste and/or increasing recycling (Callan and Thomas 1997; Dijkgraaf and Gradus 2004; Ferrara and Missios 2005; Fullerton and Kinnaman 1995; Hong, Adams, and Love 1993; Jenkins 1993; Linderhof et al. 2001; Nestor and Podolsky 1998; Podolsky and Spiegel 1998; Skumatz and Freeman 1997; Van Houtven and Morris 1999), although there are instances in which they have no impact on waste disposal decisions (Hong 1999; Hong, Adams, and Love 1993; Jenkins et al. 2003; Kinnaman and Fullerton 2000; Reschovsky and Stone 1994; Sterner and Bartelings 1999). While the impact of unit pricing on waste disposal or recycling is well documented, little to nothing is known about its impact on source reduction and on consumption and/or consumption patterns. Furthermore, although different types of unit pricing (for example, bag-or volume-based, subscription-based, frequency-based) are considered in the literature,3 there exists minimal evidence about their relative effects. In most of the available comparative analyses, a bagbased program is more effective at reducing waste and, to a lesser extent, at increasing recycling than a block payment system (Dijkgraaf and Gradus 2004; Kinnaman and Fullerton 2000; Nestor and Podolsky 1998; Van Houtven and Morris 1999). In a more comprehensive analysis of different types of unit pricing, weight-and bag-based programs have comparable effects on waste management decisions but perform better than frequency-and subscription-based programs (Dijkgraaf and Gradus 2004). In terms of its impact on recycling, a frequency-based program is, however, equivalent to a weightbased system (Sterner and Bartelings 1999).
In addition to user fees, governments often rely on recycling programs as a means of diverting waste from landfills. There is evidence that communities with recycling programs have higher recycling rates but not necessarily for every type of recyclable. Furthermore, curbside recycling programs do tend to be more effective in the presence of a unit pricing system, and vice versa (Callan and Thomas 1997; Reschovsky and Stone 1994). In general, households are sensitive to the time intensity of recycling activities and respond favorably to initiatives intended to make collection more accessible (Jenkins et al. 2003; Judge and Becker 1993; Reschovsky and Stone 1994) or to reduce sorting requirements (Judge and Becker 1993). Households are also responsive to changes in collection frequency and recycle more as collection becomes more frequent (Ferrara and Missios 2005; Judge and Becker 1993). Knowledge about recycling programs has a positive effect on whether households recycle (Reschovsky and Stone 1994), but experience with recycling programs may not contribute to increasing the probability of recycling consistently across different types of recyclables (Jenkins et al. 2003). If curbside recycling is based on mandatory participation, independently of whether unit pricing is in place, it is, however, not clear whether and how households' decisions over recycling are affected.
Another policy instrument that is often employed, although limited to particular waste items, is a refundable deposit system. Very little is known about the empirical impact of such a policy on households' waste disposal and recycling activities in spite of the extensive theoretical work that supports their implementation (Atri and Schellberg 1995; Dinan 1993; Dobbs 1991; Ferrara 2003; Ferrara 2011; Fullerton and Kinnaman 1995; Fullerton and Wu 1998; Palmer, Sigman, and Walls 1997).
Although educational programs are not commonly considered policy instruments, there is some evidence suggesting that they can be of assistance in waste diversion efforts. Environmental activism or awareness and knowledge about available management options do in fact contribute to less waste discarding and more recycling (Dijkgraaf and Gradus 2004; Hornik et al. 1995; Jenkins et al. 2003; Linderhof et al. 2001; Reschovsky and Stone 1994). Particularly interesting is the finding that knowledge and social influence from neighbors, friends, and family members are the most effective predictors of recycling, implying that, once educated about recycling (importance, availability, and how to recycle quickly and conveniently), individuals tend to recycle more (Hornik et al. 1995). Less commonly studied are attitudinal elements of influence, and the limited evidence seems to suggest that they can play a role in waste disposal decisions. A positive attitude toward composting does in fact lead to a lower demand for garbage collection services, while the perception that recycling is difficult induces households to recycle less (Sterner and Bartelings 1999). Some recent evidence also suggests that moral and social motivations can positively affect households' recycling decisions (Berglund and Matti 2006; Brekke, Kipperberg, and Nyborg 2010; Halvorsen 2008), although it is not clear whether and how the presence of economic incentives or mandatory recycling affects their relevance.
The question about possible interaction effects between policy instruments and sociodemographic or attitudinal characteristics is an important one and as relevant to policy makers as the question about possible substitution or complementary effects among different policy instruments. Although limiting, the available evidence does suggest that policy-induced changes in waste disposal and recycling are affected by sociodemographic variables. Specifically, the effect of unit pricing is smaller in low-income households, in households that subscribe to more daily newspapers, for households with infants, and for married couples (Fullerton and Kinnaman 1996). Unit pricing is also more effective in larger households but is less effective among homeowners (Van Houtven and Morris 1999). As for the effect of external intervention (through economic incentives or regulatory measures) on intrinsic motivation, there could either be a crowding out, if the intervention is perceived to be controlling, or a crowding in, if it is perceived to be acknowledging (Frey 1999). Although there exists some evidence suggesting that, when households have strong moral motives for environmentally responsible behavior, policies relying on economic incentives may be ineffective as they may undermine individuals' sense of civic duty (Frey and Oberholzer-Gee 1997), there is support to date in the specific household waste area for a crowding in, with either no erosion of personal motives in the presence of economic incentives or perceived mandatory recycling (Halvorsen 2008), or a large proportion of the positive effect of unit pricing on recycling and composting attributable to personal norms and self-efficacy beliefs (Thørgeson 2003).
II. Empirical Framework
In this paper, two separate but related decisions are considered: recycling and waste prevention. For each decision, two questions are posited: whether to participate and, if so, to what extent to participate. For the recycling decision, the relevant dependent variables are (1) indicators for recycling particular materials (glass, plastic, aluminum, paper, and food) and (2) proportions of materials recycled as captured by integers 1 to 5 (approximately 0%, 25%, 50%, 75%, and 100%). For the waste prevention decision, the dependent variables considered are (1) an indicator for taking a recycling logo into account in purchasing decisions and (2) the regularity of purchasing/using refillable containers through the assignment of 1 to the Never option, 2 to the Occasionally option, 3 to the Often option, and 4 to the Always option.
Data and Variables
The data set employed in this study was gathered by Lightspeed Online Research, Inc., for the OECD in February 2008 through an international web-based panel that involved 10,251 respondents.4 The explanatory variables used are listed with descriptions in Table 1; their means and standard deviations are given in Table 2 for the entire data set as well as for a subset comprising only observations under unit pricing (volume-or weight-based). Aside from variables commonly considered in the empirical study of household waste decisions (age, household size, education, etc.), the list includes attitudinal characteristics (rank of environmental concerns, relevance of waste generation as an environmental concern, environmental attitude, and concern for environmental issues), three four-point Likert variables to capture the extent of relevance (from not at all important with a value of 1 to very important with a value of 4) of recycling motives (environmental benefits, the belief that recycling is a civic duty, and the desire to be seen as a responsible citizen), an indicator to reflect the perception that recycling is mandatory, and a four-point Likert variable (from not at all important with a value of 1 to very important with a value of 4) to measure the importance of mandatory recycling, whenever it is an applicable motive, in motivating recycling.
To capture possible linkages between unit pricing and recycling services and between different types of motivation (economic vs. moral and/or social), as well as possible differences in how unit pricing affects different segments of the population, various interaction indicators are constructed. For possible complementary effects among policies, an indicator is created for each material to allow for any interaction between the presence of unit pricing and the presence of any type of recycling service (door-to-door service, drop off, refundable deposit system, and nonre-fundable deposit system). For possible crowding in or out effects of unit pricing, interaction variables are constructed between the presence of unit pricing and whether recycling is perceived to be mandated, the importance of taking the environmental benefits of recycling into account, and the extent to which recycling is motivated by a sense of civic duty or by a desire to be seen as a responsible citizen. Interaction variables are also created between the presence of unit pricing and income, the number of rooms, and whether the residence is owned, is a house, has a garden, or is located in an urban/suburban area.
Some summary statistics pertaining to policy instruments (collection services, financing methods, collection frequency, and recycling intensities for the five recyclable materials) are provided for the overall data set, as well as for each country, in Appendix B, Tables B1–B4. In general, and not unexpectedly, door-to-door and drop-off programs are more common than refundable deposit and bring back with no refund systems, although curbside collection is more widespread in Australia, Canada, and Korea, while collection at drop-off centers is more prevalent in the Czech Republic, France, Italy, and Sweden (Table B1). Across the five materials, refundable deposit systems are mostly implemented for glass, particularly in Canada and the Czech Republic, and plastic, particularly in Netherlands, Norway, and Sweden. Of the 10 countries, Mexico has the highest proportion of households reporting no service across the five materials.
In terms of charges, if any, households pay for the collection and management of mixed waste, flat fees are widely used in every country but Korea, where most households pay according to volume, and Mexico, where over 40% of the respondents report facing no charge (Table B2). Of the remaining systems, charging households according to their size is common in Italy and, to a lesser extent, in the Czech Republic. Mixed waste is mostly collected at least once a week, with France, Italy, Korea, and Mexico showing higher rates for more frequent collection (Table B3). Recycling participation tends to be relatively high for each material but food and, to a lesser extent, aluminum, and in each country but Korea (Table B4). Korea has a lower participation rate for glass, plastic, aluminum, and paper recycling but a higher participation rate for food recycling. Aluminum recycling participation is also quite low in France, Italy, and Norway (0% recycling reported by around 40% of respondents) and extremely low in the Czech Republic and Netherlands (0% recycling reported by around 80% of respondents). Although the lower participation rates in Korea do not seem to be linked to an absence of services (which is noticeably greater than in the other countries), the lower participation rates for aluminum recycling in the Czech Republic and Netherlands and, to a lesser extent, in France and Italy are consistent with the higher proportions of respondents in these countries reporting having no service.
Methodology
To assess the impact of the above defined explanatory variables on household recycling and waste prevention behavior, probit specifications are employed. Specifically, the decisions about whether to recycle and whether to engage in waste prevention are studied through multivariate and univariate (binary) probit analyses, while the decisions about the proportion of recyclable materials to recycle and the regularity of purchasing/using refillable containers are studied through ordered probit analyses. The multivariate probit analysis for the recycling participation decisions allows for the decision of recycling a particular material to be correlated with the decision of recycling a different material. Although the costs and benefits of recycling a particular material depend upon its volume and weight characteristics as well as upon the type of service available for its collection, thus suggesting that the decision of recycling that material is independent of recycling participation decisions pertaining to other materials, there are valid arguments suggesting otherwise. Among such arguments: (1) waste management policies targeting recycling may be introduced simultaneously for different types of materials and, in the specific case of drop-off service, centers may be placed in the same area even when comingling different materials is not permitted; (2) recycling may entail fixed costs associated, for example, with the collection of information about available collection services and the purchase of additional bins, so that the incremental cost of recycling an additional material may decrease when another material is already being recycled; (3) economies of scope may exist, as sorting a particular material is equivalent to sorting, at least partially, the other materials (see Appendix A for a brief outline of the univariate, multivariate, and ordered probit models).5
III. Results
The results of the empirical models estimated with a probit (binary, multivariate, or ordered) procedure are presented and discussed below according to whether the questions are about recycling (Tables 3, 4, and B5) or waste prevention (Table 5). For the recycling participation and intensity questions, the results are also summarized and compared to those in the related literature in Table B7.
Determinants of Recycling
The results of the multivariate analysis (Table 3) suggest that the decisions of recycling different materials are correlated. In fact, the hypothesis that the 10 off-diagonal coefficients of the variance-covariance matrix are simultaneously equal to zero is rejected, with v2(10) = 1414.57, at 1%.6 Association is always positive (that is, recycling a particular material has a positive influence on the decision to recycle another material) and is stronger among glass, plastic, aluminum, and paper, which tend to be recycled together with the same type of collection service. Association between food and any of the remaining materials is still positive but rather low, ranging from 0.15 (between food and glass) to 0.23 (between food and paper).7
In general, country effects are significant for both the decision about whether to recycle and about how much to recycle, suggesting that institutional and cultural factors, which tend to be country specific, play an important role in household recycling behavior. More specifically, Sweden tends to enjoy a higher recycling participation rate than the Czech Republic and France for all materials, Mexico for all materials but food, Italy for all materials but plastic and food, and the Netherlands and Norway for all materials but paper and food. With the exception of paper, Sweden has, however, a lower participation rate than Korea. Relative to Australia, Sweden has a comparable recycling participation rate for glass and aluminum, a lower rate for plastic and paper, and a higher rate for food. Relative to Canada, Sweden has a comparable recycling participation rate for glass, aluminum, and paper, a lower rate for plastic, and a higher rate for food. Not only does Sweden have a higher recycling participation rate but also a higher recycling intensity; in fact, with the exceptions of Italy, Korea, and Norway for the recycling of food, households in Sweden tend to recycle greater proportions of their recyclable materials. No significant difference in intensity exists, however, for aluminum recycling between Sweden and Australia, Canada and Korea, for paper recycling between Sweden and the Netherlands and Norway, and for food recycling between Sweden and the Netherlands.
Sociodemographic and Household Characteristics
Being married or living as a couple has a positive effect on recycling intensity only for plastic and on recycling participation for glass, plastic, aluminum, and paper. Men have higher recycling participation and intensity for aluminum. In general, young individuals participate less in recycling and recycle less, with plastic being the only material for which they have higher recycling participation and intensity. Household size matters only in terms of the number of children below five years of age, which reduces glass, plastic, and paper recycling. Education is mostly a factor at low levels, with individuals without a high school diploma recycling less glass, aluminum, and paper than individuals with a postgraduate degree; for glass recycling, however, individuals with a university degree tend to have higher participation and intensity than individuals without high school, with high school, or with some postsecondary education. Employment status is most relevant for the recycling of aluminum, with those working on a full-time or part-time basis, retirees, househusbands/wives, and students recycling more. Among individuals with a job or in retirement, middle/senior executives and salaried (office) employees participate less in the recycling of plastic and aluminum and recycle smaller proportions of aluminum and food. Income seems to matter only for the recycling of glass and, to a lesser extent, for the recycling of plastic and aluminum, although its marginal effect is very low, suggesting that differences in recycling are most noticeable when very rich individuals are compared to very poor individuals; specifically, richer individuals are more likely to recycle glass and tend to recycle larger proportions of glass, plastic, and aluminum.
Household characteristics (whether the primary residence is owned, is a house, has a garden or is in a suburban/urban area, how many rooms, excluding bathrooms, it has, and how long it has been the primary residence) are mostly significant. Ownership and size of residence as captured by the number of rooms tend to increase recycling, with the former variable having, however, no significant effec on food recycling and the latter having no significant effect on plastic and aluminum recycling. The presence of a garden has a positive effect only on recycling intensity for food Having a house (detached or semidetached) tends to reduce the probability of recycling glass, plastic, and paper but to increase recycling participation and intensity for food. Living in an urban or suburban area has a negative effect only on food recycling. Finally, having lived in the primary residence for more than 15 years, which would imply a stronger neighborhood attachment, has a positive effect on both recycling participation and intensity.
Attitudinal Characteristics
The evidence supporting the importance of attitudinal characteristics in recycling decisions is strong.8 Of the four variables included in the empirical analysis that capture individuals' attitudes toward waste generation and, more generally, toward the environment, the index measuring the level of concern for environmental problems (ENVCNCRN_INDX) has a positive effect on glass, plastic, aluminum, and food recycling. The index summarizing individuals' environmental attitude based on the extent of agreement or disagreement with five statements about the environment (ENVATTIT_INDX) increases recycling for glass, plastic, and aluminum. Of the remaining two variables describing environmental attitudes (ENVRANK and WSTE_ CNCRN), WSTE_CNCRN, an indicator that records whether waste generation is of concern, has generally no relevance in recycling decisions, although it reduces aluminum recycling and increases the intensity of glass recycling. The ranking of environmental concerns (ENVRANK), on the other hand, reduces recycling participation for aluminum and paper and recycling intensity for every material but plastic and food.
The results for the effects of the indices included in the analysis to capture different types of motivation for recycling (MTVRCY-LENVR_LRKT, MTVRCYLDUTY_LRKT, and MTVRCYLRESP_LRKT) suggest that one of the most important factors motivating recycling in general (that is, across the five materials) is whether and the extent to which it is considered to be beneficial for the environment. This finding resonates with the conclusion that attitudes toward the environment are far more important in recycling decisions than attitudes toward waste generation. Of the two indices reflecting personal motives for recycling based on social considerations, MTVRCYLDUTY_LRKT, which signals the presence (and measures the importance) of a sense of civic duty, is quite relevant in inducing individuals to recycle more, independently of the material; furthermore, there is no evidence that the positive effect of MTVRCYLDUTY_LRKT decreases as user charges for waste disposal are introduced (that is, there is no crowding out). On the other hand, MTVRCYLRESP_LRKT, which signals the presence (and measures the importance) of a social pressure to act responsibly (and thus a desire to be seen as a responsible citizen), is mostly insignificant but has a negative effect on food recycling intensity that, however, tends to be lower under a unit pricing system, as reflected in the positive and significant effect of the interaction term between the presence of unit pricing and MTVRCYL-RESP_LRKT.
Policy Variables
The presence of a unit pricing system for waste disposal, whether based on volume or weight or frequency, does not have a strong effect on the decision of whether to recycle. Weight-and frequency-based charges have quite a positive effect on food recycling but are otherwise ineffective; on the other hand, volume-based charges have significant and positive effects on recycling intensity for every material but plastic. Although the results do suggest that economic instruments (that is, user fees for waste disposal) can promote recycling, the evidence is not as convincing as one would expect, especially for charges levied according to weight and frequency, based on theoretical predictions and empirical findings in most of the studies on recycling (e.g., Dijkgraaf and Gradus 2004; Ferrara and Missios 2005). As in Jenkins et al. (2003), one of the very few studies that reports no significance for unit pricing, most of the observations (about 80%) come from communities without some form of unit pricing (see Table B2) that either charge a flat fee or a fee based on the size of the household or do not charge at all for waste collection. Of the remaining observations, 1,108 (or 12.6%) are from communities with a volume-based fee, 405 (or 4.6%) from communities with a frequency-based fee, and 241 (or 2.7%) from communities with a weight-based fee.
When unit pricing is assessed in conjunction with collection services for recyclables, the evidence based on the estimated coefficients does not suggest that the presence of collection programs for recyclables tends to increase the effectiveness of unit pricing rather than to decrease it. Hence, from a policy point of view, collection services for recyclables and unit pricing may be substitutable as opposed to being complementary approaches, a result that may counter the finding in the literature that unit pricing is more effective if combined with curbside recycling and vice versa (Callan and Thomas 1997). The absence of any evidence supporting complementarity between unit pricing and collection services for recyclables (mostly curbside and drop off), along with the presence (in the cases of plastic and food) of some evidence supporting substitutability between the two policies, may explain why user charges (particularly, volume-and weight-based fees, as frequency-based fees do not directly provide incentives to recycle) are not found to be effective at increasing recycling participation. Indeed, when observations with curbside collection for recyclables are excluded from the empirical analysis, the results of the ordered probit estimation, which are reported in Table B6 only for the relevant variables (namely, curbside collection of recyclables and weight-, volume-, and frequency-based fees), show that unit pricing based on volume or frequency has always a positive effect on recycling independently of the material, while unit pricing based on weight has a positive effect on glass, aluminum, and food recycling.
Although a unit pricing system for mixed waste and a door-to-door collection program for recyclables seem to be substitutes, the former may be the redundant policy in that its positive effect on recycling disappears when observations with curbside collection for recyclables are excluded from the analysis; the effect on recycling of a door-to-door collection program is, however, always present and positive independently of whether observations with a user fee system are excluded (see Table B6). This difference is likely attributable to the different channels through which the two policies affect recycling. A door-to-door collection program has a direct effect on recycling through a reduction in its time cost; a unit pricing system has an indirect effect on recycling through a reduction in the cost of disposing of mixed waste. Hence, the time cost of recycling in the absence of curbside collection may be a more relevant consideration in recycling decisions than its money benefit in the presence of unit pricing. When curbside collection for recyclables is introduced, individuals tend to recycle more in light of the reduced time cost of recycling; a unit pricing system for mixed waste may then provide no additional incentive (in the form of money saving) for recycling. When curbside collection for recyclables is not available, the money saving aspect of recycling in the presence of a unit pricing system may be sufficient to offset the time cost of recycling (either through a drop-off program or a deposit system with or without refund), thus inducing individuals to recycle.
Among the variables that are most consistently significant across the five materials and influence recycling participation and intensity in accordance with theoretical predictions, are the following: whether some type of collection service is in place, the frequency of curbside collection if available, whether recycling is mandated by the government as captured by the applicability of mandatory recycling as a factor motivating recycling, and the frequency of mixed-waste pickup. In general, having any type of service (door-to-door, drop off, bring back with refund, bring back with no refund) results in more recycling. In terms of marginal effects (Table B5), the availability of curbside recycling has its greatest impact on the probability of recycling 100% for aluminum, which increases by approximately 0.48 compared to 0.35 for plastic, 0.29 for paper, and 0.28 for food and glass. Under a dropoff system, the largest impact is detected for aluminum, with a 0.29 increase, followed by plastic with a 0.21 increase, food and paper with a 0.19 increase, and glass with a 0.17 increase. Under a refundable deposit system, the probability of recycling everything increases by 0.29 for plastic, 0.26 for aluminum, 0.19 for paper, 0.13 for food, and 0.09 for glass. Finally, under a bring back with no refund system, food experiences the largest increase in the probability of recycling 100%, followed by aluminum, plastic, paper, and glass; specifically, the probability of recycling 100% increases by 0.22 for food, 0.18 for aluminum, 0.09 for plastic, 0.07 for paper, and 0.04 for glass. Hence, collection programs for recyclables seem to be most effective for aluminum and food and least effective for glass and paper. Furthermore, of the four types of collection programs, curbside recycling is the most effective independently of the material, while a refundable deposit system is the least effective for food and a bring back with no refund system is the least effective for the remaining materials (glass, plastic, aluminum, and paper).
While the presence of curbside collection for recyclables increases recycling, the frequency of collection has a negative impact on both recycling participation and intensity. In particular, moving from less than once a week to once a week collection reduces recycling participation for paper and recycling intensity for glass, plastic, and aluminum; moving from less than once a week to more than once a week collection reduces recycling participation for plastic, aluminum, and paper and recycling intensity for every material but food. Increasing how often mixed waste is collected also reduces recycling, although recycling participation for glass, plastic, and paper tends to be statistically responsive only to a shift from less than once a week to once a week collection.
Under mandatory recycling, individuals tend to exhibit higher recycling participation and intensity, particularly for glass, plastic, and paper; however, individuals for whom mandatory recycling is an important factor in motivating their recycling decisions are not likely to recycle more, with the exception of food and, for the participation decision although with the opposite effect, glass. In terms of marginal effects, the probability of recycling 100% when mandatory recycling is an applicable factor motivating recycling is higher by 0.11 for glass and 0.07 for plastic and paper; furthermore, as mandatory recycling becomes a more relevant consideration in recycling decisions, the probability of recycling 100% increases by 0.02 for food but decreases by 0.03 for glass and by 0.01 for plastic, while it remains unchanged for the other materials. The presence of unit pricing reduces the effect of mandatory recycling on recycling participation for glass, plastic, and aluminum and on recycling intensity for glass and paper.
As for the interaction terms between unit pricing and some sociodemographic characteristics, the evidence is a bit scattered, with the presence of unit pricing strengthening the effect of (1) income on plastic recycling participation, (2) living in a house on glass and plastic recycling participation, (3) living in an urban or suburban area on glass recycling participation, and (4) having a garden on aluminum recycling participation, while weakening the effect of (1) owning the primary residence on glass recycling intensity, aluminum recycling participation, and paper recycling participation and intensity, (2) living in a house on food recycling participation and intensity, and (3) size of primary residence as reflected in the number of rooms (excluding bathrooms) on food recycling participation.
Determinants of Waste Prevention
In the absence of consumption figures and, more specifically, information on the waste content of consumption, the question about waste prevention and the factors contributing to waste prevention is addressed indirectly through a binary question about the importance of recycling logo information in purchasing decisions and a question about using refillable containers involving an ordinal choice over regularity of use.
Based on the country effects, with the Czech Republic excluded from the binary probit estimation because of unavailability of a recycling logo, there exist institutional and cultural factors that yield differences across countries. For example, while the results of the binary probit estimation suggest that the probability of engaging in waste prevention as captured by the probability of taking into account recycling logo/label information in purchasing decisions is higher in Sweden than in any of the remaining eight countries, the results of the ordered probit estimation suggest that the intensity of waste prevention as captured by how regularly refillable containers are used is, for the most part, lower in Sweden than in the other countries (see Table 5).
Sociodemographic and Household Characteristics
In terms of sociodemographic characteristics, relevant factors include gender, age, education, employment status, income, presence of a garden, and type of area of residency. In the estimation of the intensity of use for refillable containers, only four of these variables matter, namely, gender, age, garden, and urban/suburban indicators. To be precise, older and/or male individuals, individuals without a garden, and individuals living in an urban or suburban area tend to use refillable containers less regularly. At the same time, younger and/ or male individuals as well as individuals with access to a garden are more likely to take recycling logo information into account in purchasing decisions; individuals living in an urban or suburban area, on the other hand, do not display a behavior, in terms of considering recycling logo information in purchasing decisions, that is significantly different from the behavior of individuals living in other types of areas. Furthermore, individuals working full-time, those working part-time, and house-husbands/wives are less likely to consider recycling labels in their purchasing decisions by 0.08, 0.07, and 0.09, respectively. Richer individuals are less likely to pay attention to recycling labels when shopping, although any noticeable difference in behavior between rich and poor individuals requires a substantial income gap, as the marginal effect of income is quite negligible.
Attitudinal Characteristics
Of the variables characterizing attitudes toward the environment or motivation for recycling, most are significant and have the expected effect on waste prevention. In particular, individuals who show a greater concern for environmental problems are more likely to engage in waste prevention both in terms of accounting for recycling labels in purchasing decisions and using refillable containers more regularly. Individuals who rank environmental concerns high in order of importance or show a stronger attitude toward the environment are more likely to take recycling labels into account in purchasing decisions, but there is no evidence that they make more extensive use of refillable containers. As with the case for the recycling participation and intensity decisions, concern for waste generation is not an important determinant of waste prevention decisions. However, individuals who believe that recycling is beneficial for the environment or that it is a civic duty tend to engage more in waste prevention activities, while individuals who believe that recycling is a social responsibility tend to use refillable containers more often but are not more likely to account for recycling labels when shopping. Finally, individuals who face mandatory recycling are more likely to account for recycling labels in their purchasing decisions but tend to become less likely to do so as mandatory recycling becomes a more important consideration in their recycling decisions.
Policy Variables
Unit pricing, whether based on weight, volume, or frequency, does not seem to affect whether recycling labels are taken into account in purchasing decisions but does increase the probability of using refillable containers more regularly. The presence of recycling services is, for the most part, statistically insignificant in both the binary probit estimation and the ordered probit estimation. There exists some evidence, however, suggesting that individuals would pay greater attention to recycling labels if (1) their waste were to be collected less frequently, (2) their recycled glass were to be collected less frequently, (3) a bring back with no return system were not available for glass, (4) curbside collection were available for tin and steel cans (aluminum), (5) tin and steel cans were collected at the curb more than once a week, and (6) a drop-off system were available for tin and steel cans. Individuals would also use refillable containers more often if a drop-off system or a refundable deposit system were in place for tin and steel cans.
IV. Policy Implications
A result that is common to the two issues addressed in this paper about recycling and waste prevention relates to the presence of institutional and cultural elements, as captured by the country dummy variables, explaining variation in household behavior across countries. An important implication of this finding is that policy makers may derive some useful lessons by looking closely at countries, such as Sweden, that tend to consistently exhibit a more environmentally friendly behavior. Among factors to consider are countries' approaches to waste management and views on environmental problems when the whole product chain is taken into account. As the empirical analyses in this paper are based on partial equilibrium models that focus on the interaction between households and the government, variation across countries may result from differences in policies, regulations, and actions taken at different stages of the product chain, as well as differences on the supply side of collection services. Sweden in particular takes a holistic approach to waste management (and environmental problems in general) in that it holds producers and distributors of goods responsible for the waste they produce; in other words, companies are responsible by law for the collection of the entire waste stream resulting from their products, either directly or through public or private contractors.
Of the four variables describing environmental concerns and attitudes, the one specific to waste generation has no impact on waste prevention and recycling efforts, with the exception of the decision about how much glass to recycle. The remaining variables do matter, almost consistently across the five types of recyclables. The importance of attitudes toward the environment in general (as opposed to waste generation) has implications for the design of effective informational measures targeting recycling and waste prevention. Informational measures presuppose a more psychological perspective of human behavior and aim at changing perceptions, motivations, and knowledge levels. Hence, informational campaigns that stress how waste production contributes to environmental deterioration may be quite helpful in inducing individuals to recycle more and produce less waste. Unlike other environmental issues (e.g., car pollution, climate change) to which individuals can relate more closely, waste generation may not be perceived as a major environmental problem and may in fact be viewed more as a practical nuisance because of its space requirement than as an environmental problem. That a favorable response may ensue from an increased awareness of the environmental implications of waste generation is also supported by the finding that individuals who believe that recycling is beneficial for the environment are more likely to recycle and engage in waste prevention.
As social considerations constitute an important determinant of both recycling and waste prevention decisions, informational measures that focus on social aspects may also assist in promoting recycling and waste prevention. Although there are two sources of social motivation considered in this study, namely, a belief that recycling is a civic duty and a desire to be seen as a responsible citizen, findings suggest that the social dimension of waste management comes from the former and thus from a desire to act responsibly as opposed to being seen as acting responsibly. Informational measures that build upon social considerations may then be more effective if they present waste reduction as a moral responsibility rather than as a social pressure.
The evidence gathered in this study also suggests that information-based measures can coexist with pricing-based schemes, that is, governments can simultaneously implement both types of intervention. In fact, policies relying on economic incentives do not tend to reduce (or crowd out) individuals' intrinsic motivation for environmentally responsible behavior. In some instances (for example, food recycling participation), economic incentives may actually increase (or crowd in) individuals' intrinsic motives for environmentally responsible behavior and are thus more likely to be perceived as communicating norms and responsibilities (that is, acknowledging).
In spite of the significant body of evidence in the waste generation and recycling literature that points to the contrary, user charges are not found to be very effective, particularly in the participation decisions. That frequencybased pricing schemes are not effective is not surprising given that individuals facing charges based on mixed-waste collection frequency do not pay per unit of waste generated but per collection and can avoid some of their disposal costs simply by storing more waste at home without necessarily recycling more. That weight-based and volume-based charges are insignificant is, however, quite unexpected. Unfortunately, despite the greater policy heterogeneity resulting from the international setting, very few observations in the data set are drawn from communities with some form of unit pricing.
An important result regarding a unit pricing system for the collection of mixed waste is that it is not a complement of a door-to-door program for the collection of recyclables, as previously gathered evidence suggests (Callan and Thomas 1997). When unit pricing is assessed only for households without access to curbside collection of recyclables, findings reveal that weight-and frequency-based charges can be effective at inducing individuals to recycle more, although they remain insignificant in the decision about whether to recycle and continue to perform below volume-based charges (see Table B6). In contrast, when households with access to curbside collection of recyclables are not excluded from the analysis, volume-based charges do not matter in the paper recycling intensity decision, and frequency-and weight-based charges matter only in the food recycling intensity decision. As a door-to-door collection program for recyclables is always effective independently of whether user charges are implemented and in both the decision about whether to recycle and the decision about how much to recycle, increasing recycling may be more easily achievable with policies that focus on the time cost of recycling as opposed to policies that stress the money benefit of recycling.
The evidence supporting the importance of the presence of a collection program for recyclables is quite strong. In general, the presence of any type of service (door-to-door, drop off, bring back with refund, or bring back without refund) increases recycling participation and intensity but has mostly no effect on waste prevention. Based on the marginal effects estimated for the decision of recycling everything, a door-to-door program is preferable to a drop-off system for any recyclable. As curbside collection may be more costly to administer than collection by means of a dropoff center, the benefits from the additional recycling under the former would have to be weighed against its potentially higher provision cost. Hence, plastic and aluminum, for which curbside collection brings about a larger benefit over and above the benefit from a drop-off system, may be better candidates for curbside collection. A drop-off system performs better than a refundable deposit system for any material but plastic and is particularly appealing for glass and paper, both of which experience a larger increase in recycling everything than the other materials. Finally, when a deposit refund system is compared to a bring back without refund system, the former is preferable to the latter for any material but food. Needless to say, significant administrative cost differences may exist among the four types of services, which policy makers would have to account for before deciding on which program to implement and for which material.
Although policy makers may consider mandating recycling to get individuals to recycle or to recycle more, mandatory recycling may work for some materials (glass, plastic, and paper) but not for others (aluminum and food) and may yield smaller benefits than a curbside or drop-off program. Furthermore, mandatory recycling may have stronger side effects than recycling programs on whether and how unit pricing impacts recycling intensity. In light of these considerations, mandatory recycling may not be a desirable policy option, and policy makers may be able to achieve better results by focusing on improving accessibility of recycling services.
In implementing a collection program for recyclables, policy makers should keep in mind that such a program may only succeed at targeting a particular aspect of waste management. Ideally, a policy-induced behavioral adjustment should include both an increase in recycling and a decrease in waste generation through a shift in consumption patterns in favor of products with less waste content and/ or reusable products. The evidence in this study points to the conclusion that the provision of recycling services does not encourage individuals to produce less waste, so that, for waste prevention, policy makers may have to resort to additional mechanisms that may involve incentive structures at other stages of the product chain. There is certainly significant variation in waste prevention across the 10 countries, as reflected in the country dummy variables, which may be attributed to differences in waste management policies at the production stage.
When a collection program is being contemplated, an important feature to consider is pickup frequency. While the findings of this study suggest that individuals take into account both the frequency of collection of mixed waste and the frequency of curbside collection of recyclables, the behavioral response to an increase in the latter is inconsistent with theoretical predictions. In fact, individuals tend to recycle less as mixed waste is collected more frequently, as expected, but to recycle less as recyclables are collected more frequently, which is unexpected. To increase recycling, policy makers should thus consider a less frequent collection of mixed waste and, if they implement a door-to-door collection of recyclables, should not necessarily opt for a more frequent collection of recyclables. It is quite possible, although further investigation is necessary to confirm this interpretation, that there exist economies of scale in recycling activities, partly because of the preparation that is required every time recyclables are placed at the curb. Because of this preparation or start-up cost, individuals may find the process of getting recyclables ready for collection less time consuming if they have to engage in it every two weeks as opposed to once (or even more than once) a week. Individuals may thus react to a more frequent collection of recyclables simply by recycling less to avoid incurring additional time costs, which, from previous discussions in relation to unit pricing, seem to constitute a more relevant factor in recycling decisions than monetary benefits.
V. Concluding Remarks
The present study widens the scope of previous analysis and improves upon our understanding of household recycling and waste prevention behavior in two main ways: (1) by bringing together key aspects of household behavior (sociodemographic characteristics and attitudinal variables), thus capturing a broader spectrum of policy influences and allowing for a more accurate assessment of the direct effects of sociodemographic factors and for the investigation of complementary effects among strategies that differ in the assumptions about how behavior can be changed; (2) by examining the effects of a broad range of policy instruments (pricing, informational, and regulatory). Although waste prevention is only measured indirectly through a couple of proxies, this study represents the first attempt to assess the effects of sociodemographic and attitudinal characteristics as well as policy instruments (unit pricing, in particular) on activities related to waste prevention.
One of the key insights relates to the presence of strong intrinsic motivations for a more environmentally responsible behavior. There is indeed a positive relation between environmental concern and/or attitude and recycling or waste prevention, which suggests that more environmentally sensitive individuals tend to recognize the environmental deterioration that results from waste production and to exhibit a greater commitment to recycling and waste prevention activities such as taking recycling labels into account when shopping and using/ purchasing refillable containers.
Correspondingly, individuals who are simply concerned about waste generation do not tend to adjust their recycling and waste prevention efforts. Intrinsic motivations are also present in the form of moral/social considerations, as individuals with a stronger sense of civic duty tend to engage more extensively in recycling and waste prevention. On the other hand, individuals who are motivated by a desire to be seen as acting responsibly are not more environmentally responsible. An important implication of these findings is that there may be benefits from sensitizing individuals to environmental problems and educating them about the environmental impact of waste production and the moral dimension of recycling and waste prevention.
The evidence gathered in this study is quite conducive to the conclusion that individuals do respond favorably, in terms of recycling efforts, to the presence of recycling services and that, the more accessible such services are, the more responsive they become. Hence, curbside collection of recyclables yields better results than the other three programs considered in the analysis (drop off, refundable deposit, and bring back with no refund). Under curbside collection of recyclables, however, more frequent collection of recyclables is not necessarily desirable, contrary to findings in previous studies (Ferrara and Missios 2005; Judge and Becker 1993). The adverse effect of frequency on individuals' recycling behavior also transpires in the collection of garbage, as the more frequently their garbage is collected, the less likely to recycle individuals are and the less they recycle.
The evidence on unit pricing is not as strong as one would expect based upon theoretical predictions and previous empirical findings. Nevertheless, of the three types of unit pricing examined in the study (weight-, volume-, and frequency-based), the volumebased system appears to be the most effective. An important result is that unit pricing may have little to contribute when curbside recycling is in place, especially if weight-based or frequency-based charges are being considered, and curbside recycling may have additional benefits over and above those brought about by unit pricing. In terms of waste prevention, the presence of unit pricing does not affect the decision about whether to take recycling labels into account in purchasing decisions but does have a positive effect on the decision about how often to use/purchase refillable containers instead of their alternatives. The evidence thus points to some role that user charges may have in inducing individuals to alter their consumption patterns in order to reduce waste, as theory predicts.
While the evidence may suggest stronger support for curbside recycling than for unit pricing, there are a few considerations about unit pricing that deserve mention. First, unit pricing does not seem to crowd out intrinsic motivations for recycling but does reduce the positive effect that mandatory recycling has on recycling participation and intensity (equivalently, mandatory recycling reduces the effects of unit pricing). Second, there are specific sociodemographic segments of the population that may respond to unit pricing more or less favorably (for example, the effect of unit pricing on paper recycling is smaller among homeowners; the effect on food recycling is smaller among those living in a house; the effect on aluminum recycling participation is larger among those who have a garden; the effect on glass recycling participation is larger among those living in an urban or suburban area). Third, the number of observations in the data set that fall under each of the three types of unit pricing (weight-and frequency-based, in particular) is very small, and as a result, the effects of each system may not be fully captured in the analysis.
Appendix A
Binary and ordinal regression models can be derived from a latent-variable model that relates a latent or unobserved variable y ∗ ranging from -∞to ∞to the observed independent variables according to the structural equation
where β is the vector of coefficients estimated by maximum likelihood estimation (MLE), xis the vector of independent variables, ε is a random term,9 and i denotes the observation. The idea of a latent y ∗ is that the underlying propensity (for example, to recycle or to recycle a particular proportion) generates the observed state; although the propensity itself cannot be observed, a change in what is observed is triggered by a change in y ∗. The probability of an event occurring is thus given by the cumulative density function (cdf) of ε evaluated at given values of the independent variables. A simple measurement equation is then used to link the observed y with the latent y ∗. In the binary case,
so that positive values of y∗ are observed as y=1 while negative values of y∗ are observed as y=0, and the probability of the event occurring is given by
where Φ denotes the normal cdf. In the ordinal case,
where J is the number of categories and μj is the cut point for j =1,...,J -1, such that 0<μ1<μ2< ...<μJ-1, and the probabilities of household i falling into the J categories are given by
that is, the areas under the normal cdf between pairs of cut points.
For household i, the multivariate probit model can be written as
and the resulting measurement equation as
for m =1,...,M, where Mis the number of equations (or materials in the present case). The error terms are distributed as a multivariate normal variable with a mean of 0 and a covariance matrix Vwith Vmg =1if m = g, where m,g =1,...,M, and Vmg=ρmg=ρgmif m ≠ g . Estimation of multivariate probit models requires the computation of multivariate normal probability distribution functions that, for integrals of level greater than three, can be accomplished with simulation methods. One of such methods is the Geweke-Hajivassiliou-Keane (GHK) algorithm, which is based upon the idea that the joint probabilities can be written as a succession of conditional probabilities.
Appendix B
Additional tables (Tables B1–B4) are provided with summary statistics about collection services, financing methods, collection frequency, and recycling intensities. Table B5 gives the marginal effects of key variables (e.g., user fees and recycling services) on recycling intensity, while Table B6 provides results for key variables from the ordered probit estimation for the recycling decision in the presence of both unit pricing and door-to-door collection of recyclables and under only one of the two policies. Finally, Table B7 summarizes and compares the findings of this study with those in the related literature.
Footnotes
The authors are, respectively, associate professor, York University, Toronto, Ontario, Canada; and associate professor, Ryerson University, Toronto, Ontario, Canada.
↵1 Within the OECD region, municipal waste generation increased by about 58% from 1980 to 2000 and 4.6% between 2000 and 2005; under the assumption of no new policies, total municipal waste is projected to increase by 38% from 2005 to 2030 and per capita municipal waste by 25% (from 557 kg to 694 kg) over the same period (OECD 2008).
↵2 In 2005, for example, households produced over 75% of municipal waste in Korea, Germany, the United Kingdom, Mexico, Belgium, the Netherlands, the Slovak Republic, Luxembourg, Denmark, and Spain (OECD 2008).
↵3 A subscription-based program entails households to precommit to a certain number of bags over a given period of time for which they pay independently of whether they use them. In terms of the unit of measurement upon which a payment is established, a subscription-based program is not much different from a bag-based program, although variation in average and marginal fees is likely to result across blocks or levels of commitment under the former but not under the latter.
↵4 In light of possible sample bias (due to the means of implementation of the survey) and strategic bias (due to the nature of the survey), the OECD performed several qualitative data checks on sociodemographic variables and other variables specific to the five areas of household consumption considered in the survey. The results of the data corroboration can be found at www.oecd.org/dataoecd/55/19/ 44101274.pdf. The results of the study are summed up in OECD (2011).
↵5 To account for the possible endogeneity of environmental attitudes (ENVATTIT_INDX and ENVCNCRN_ INDX, in particular), a two-step procedure was used to estimate the IV probit model with the above attitudinal variables instrumented by the four education indicators (AGE_CLASS_1, AGE_CLASS_2, AGE_CLASS_3 and AGE_CLASS_4). An important consideration in the choice of the instrumental variables is that they are correlated with the potentially endogenous variables but do not influence recycling participation. In all cases but glass, the hypothesis about the validity of the instruments could not be rejected based on the Amemiya-Lee-Newey minimum v 2(2) statistic (p-values associated with the statistic were 0.03 for glass, 0.90 for plastic, 0.76 for aluminum, 0.73 for paper, and 0.11 for food). In all cases but paper, the hypothesis of no endogeneity could not be rejected based on the Wald test of exogeneity (p-values associated with the Wald v 2(2) statistic were 0.22 for glass, 0.93 for plastic, 0.16 for metal, 0.01 for paper, and 0.86 for food). The results were taken to conclude that the model did not suffer from an endogeneity problem, and the IV probit approach was dropped from consideration.
↵6 These coefficients (ρij for i,j =1,...,5 with i≠ j and ρij = ρ ji) measure the strength of linear association between recycling a particular material and recycling any of the remaining materials.
↵7 In addition to the reasons stated above, correlation of errors across the various materials may result from unobserved personality traits that affect behavior with respect to all materials alike.
↵8 In relying on this evidence, a caveat to keep in mind is that stated behavior does not necessarily correspond to actual behavior. The strong correlation between stated behavior and attitudes found in the paper does not necessarily imply a strong correlation between actual behavior and attitudes.
↵9 The error term could be distributed normally (probit specification) or logistically (logit specification). The two distributions differ only in spread, with the latter having thicker tails: var(ε)=π2 /3 with the logistic cdf and var( ε)=1 with the normal cdf. The two distributions can give different results if the sample is unbalanced (that is, most of the outcomes are similar with only few differences).
