The effect of different waste collection systems on separation behaviour of households

(1)

behaviour of households

Job Papineau Salm

Economics and Finance, Supervisor: Vadim Nelidov January 31, 2017

Abstract

Hardly any academic attention has been devoted to the relationship between waste collection systems and separation behaviour. To our knowledge previous studies included only correlational analysis on these collection systems. This thesis would be the first to establish a causal link between the waste collection systems and the separation rate of Dutch municipalities. For the analysis three types of models are implemented: Logit regression, OLS-regression and Fixed Effect regression. The results suggest that the systems where households have to bring their recyclables to a nearby container are less likely to separate their recyclables than the systems where municipalities have to collect the recyclables at the homes. The combination of a system where households can bring their recyclables to a nearby container and where municipalities collect their recyclables at their homes seem to give the best result in terms of separation behaviour.

(2)

Statement of Originality

This document is written by Job Papineau Salm, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1 Introduction

Negative externalities of waste production are of growing concern with the constant rise of the quantity of waste produced by society (United Nations, 2007). Externalities like global warming, great Pacific garbage patch and habitat destruction belong to the most debated topics in the world. Increasing awareness of these negative externalities is reflected in the increase in separation ratios of waste in most of the country’s (European Environment Agency, 2013). Another important factor is the influence of the population growth on waste generation (United Nations, 2007). In view of the constant population growth of the last decades sustainable waste management is needed to manage the increasing amount of waste in the near future. Sustainable waste management should therefore be an important policy objective.

The Netherlands has an average separation rate of 50.9 per cent. This is high compared with the leading waste producing countries in the world. The United States of America and China have an average separation rate of 34 and 22 per cent (Climate Top, 2015). While the recycling rates are increasing, new targets are still being implemented. European countries are obliged to separate 65 per cent of their municipal household waste before 2030 (European Environment Agency, 2013). These targets were not strict enough for the Dutch government. Rijksoverheid (2016) established a target for the municipalities to increase the separation rate of municipal household waste to 75 per cent by 2020. This target is unlikely to be achieved unless decisive changes in waste management policies are taken.

Academic attention devoted to a causal relationship between the waste collection and the separation behaviour is limited. Scientific research on separations rates of waste focuses mainly on other determinants than collection systems of household such as the pricing systems and social-demographic values of individuals. For example, waste pricing per bag has -a significant effect on recycling behaviour and environmental values (Dijkgraaf & Gradus, 2009). Furthermore, Barr (2011) found in previous studies that young, female, high-income, well-educated and political liberal individuals tend to play an active part in reducing waste generation and recycling of waste.

The waste collection system in the Netherlands is distinguished into two systems: subsequent separation and separate collection (Nedvang, 2017). Subsequent separation refers to a collection system where machines separate the recyclables from all the household waste combined. Separate collection refers to systems where households separate recyclables at

(5)

their own homes. Municipalities either use a separate collection system where households have to bring their recyclables to a nearby container or the recyclables are collected by the municipality at their homes. Municipalities have also the option to have both systems at the same time (Nedvang, 2017).

The goal of this thesis is to provide a better understanding of the determinants related to the separation rates of waste in municipalities of the Netherlands. The focus of the thesis is on the determinant separate waste collection services. Subsequent separation is excluded, because it is irrelevant for understanding of separation behaviour of households. This will be discussed in the model section. Municipalities could use insights of this thesis to implement better policies in order to fulfil the national target. This thesis will analyse most of the determinants related to separation of waste but some of them will not be used in the model. The model section explains why these determinants are left out and why this thesis focuses on the determinant waste collection systems. The limited studies on the relationship between collection systems and separation behaviour and the discussed exclusion of other factors lead to the following research question:

How do waste collection systems of municipalities affect household decisions with regard to separation behaviour in the Netherlands?

Separate waste collection systems where households have to make less effort to separate their recyclables are likely to stimulate separation behaviour of households (Morris & Holthausen, 1994). Individuals have to walk a greater distance and thus make more effort in order to separate their waste in a system where individuals have to bring their recyclables to a nearby container. The hypothesis is therefore: The system where the recyclables are collected by the municipality at their homes is expected to stimulate separation behaviour more than a system where individuals have to bring their recyclables to a nearby container.

Logit regressions, Ordinary Least Squares regressions and Fixed Effect regressions are implemented to answer the research question. The Logit regressions are conducted to analyse the determinants of Dutch municipalities’ choice to adopt a certain waste collection system. The OLS and Fixed Effect regression are conducted to analyse the effect of waste collection system on the separation rates of municipalities. Analysing these regressions on this specific determinant should give a better understanding of this subject and therefore helps municipalities to reduce the environmental damage of waste generation.

(6)

First, it gives the theoretical background and explains the importance of recycling itself. Second, the literature review focuses on the empirical work that has been conducted with regard to the determinants that influence the separation rates. Third, it explains the determinants that influence policy decision with regard to sustainable waste management. The literature review is followed by an explanation of the used methods. It will cover the analysis for answering the research question and it will explain why some determinants are left out and why this thesis will focus on the determinant separate collection systems. After this, the results of the Logit regression, OLS regression and FE regressions are presented as well as the implications for further research. The last two sections discuss the possible violations of the assumptions and draw conclusions.

(7)

2 Literature

2.1 Theoretical background and importance

The Paris climate conference of 2015 has put recycling of municipal waste on the European agenda. New waste targets and the ambition to reduce emissions have resulted in new priorities for recycling (European Environment Agency, 2013) .For clarifying the importance of recycling it is essential to know what is meant with “recycling”. “Converting waste into reusable material” is the most common used definition of recycling. This definition will therefore be used throughout the whole thesis. Recycling includes mainly two parts: making recyclables available for reprocessing (separation and collection) and the process by which materials are made into substitutes for primary materials. This thesis focuses on the collection part of recycling. The main benefits of recycling municipal waste are the following. 1: Recycling reduces the energy use and emission of CO2 caused by the extraction of raw materials (e.g.oil, gas and coal);

2: The adoption of recycled materials instead of virgin materials reduces the energy use for the manufacturing process;

3: Raw material extraction disrupts ecosystems. Recovering these ecosystems allows vegetation to absorb carbon dioxide again;

4: Landfilling of organic materials is a huge source for methane into the atmosphere.

It is not often argued that the adoption of recycled materials reduces use of virgin materials or the amount of waste generated. In contrast to the argument that recycling leads to emission reduction (Nakamura, 1999). Although there is no evidence for all materials, according to the results of Nakamura (1999) recycling of old paper definitely reduces the total emission of CO2. Furthermore, waste disposal in landfill inevitable generates greenhouse gases, which stimulates the effect of global warming (El-fadel, Findikakis, & Leckie, 1997). Although the effect of recycling on the emission of CO2 is not fully proven yet, the effect of recycling on limited virgin materials is undeniable. Recycling is therefore important for promoting sustainability.

2.2 Determinants

Determinants for separation behaviour are complicated to fully understand. Therefore, the determinants are separated into two groups for the simplicity of the analysis. The first group of determinants is associated with the waste infrastructure of a municipality: pricing systems,

(8)

collection systems and the amount of waste generated. The second group of determinants is associated with individual characteristics such as environmental knowledge and social-demographic values.

Waste infrastructure

Dutch municipalities are obliged to supply a free collection system for paper, glass, textiles and compostable waste. This is in contrast to the financing mechanism for residual waste collection (Dijkgraaf & Gradus ,2009). The financing mechanism is distinguishable in two main systems: One Fixed Fee system (FF) and a Unit Based Pricing system (UBP). UBP includes four systems: volume, frequency, bag and weight-based. FF includes two systems: number of people and fixed fee (Nedvang, 2017).

Dijkgraaf & Gradus (2009) found that, based on their pooled cross-section analysis, UBP waste treatment has a significant effect on the quantity of collected waste and separation behaviour in the Netherlands Kinnaman (2006), (Linderhof, Kooreman, Allers, & Wiersma, 2001) and van Houtven and Morris (1999) show similar results in the United States of America and the Netherland in different time periods. The pooled cross-section analysis of Dijkgraaf & Gradus (2009) included a correction for environmental activism of the municipalities.In most of the municipalities that changed to a UBP system in an early stage people are more interested in environmental issues. The research showed that people are more interested in environmental issues when municipalities changed to a UBP system in an early stage. Consequently municipalities who did this have already a higher separation rate (Dijkgraaf & Gradus, 2004; 2009) To conclude, not correcting for environmental activism causes overestimation of the estimators.

Dijkgraaf & Gradus (2009) suggest that the introduction of a UBP system may have adverse effects as well. Individual incentive to dump waste illegally or in other municipalities may increase. This adverse effect has only been investigated in Oostzaan. This small municipality is the first municipality that introduced a UBP system on waste. Linderhof et al. (2001) found that UBP has a negligible effect on illegal waste dumping in Oostzaan. However, the validity of this study could be doubted because of its small size. This must be taken into account in final conclusions. .

As discussed in the introduction, the waste collection system in the Netherlands is distinguished into two systems: subsequent separation and separate collection (Nedvang, 2016). Subsequent separation refers to a collection system where machines separate the recyclables from all the household waste combined. Separate collection refers to system where households separate recyclables at their own homes. Municipalities either use a

(9)

separate collection system where households have to bring their recyclables to a nearby container or the recyclables are collected by the municipality at their homes. Quantitative research on the effect of these Dutch waste collection systems on separation behaviour of households has not been done so far. However the effect of implementing a subsequent separation system or a separate collection system has been studied in other countries. Jahre (1995) suggested that postponing separation to a subsequent separation system can enhance recycling’s performance by reducing cost and increasing service. CE Delft found similar results (CE Delft, 2013). Combining the two systems seem to give the best results in terms of recycling benefits. Although these studies are interesting for policy decisions with regard to implementing subsequent separation systems, it has little to do with the separation behaviour of households.

The studies of Morris & Holthausen are more relevant for separation behaviour. They found that transaction costs influence the effect of different pricing systems on waste generation and separation rates. Also, waste collection systems where households have to give less effort to separate their recyclables are stimulating separation behaviour of households (Morris & Holthausen, 1994). Consequently, the transaction costs are lower for citizens in a system where the recyclables are collected at their homes. Therefore, this could influence the effect of waste collection systems on separation rates. Furthermore, waste generation has a negative effect on separation rates. (Dijkgraaf & Gradus, 2004; 2009). Generating more waste increases the transaction cost of separation per household for recycling the same amount of waste. Additionally, waste generation is correlated with many variables. Income and employment have a positive effect on waste generation. In contrast, UBP has a negative effect on waste generation (Fullerton & Kinnaman, 2016).

Individual characteristics

Individual characteristics play an important role in the decision making of separation behaviour. For instance, environmental knowledge gives an incentive to recycle. The difficulty is how to measure the environmental knowledge of an individual. Although questionnaires give some insight, it is still difficult to measure. Oskamp et al., (1991) divided the explanatory variables for Environmental Knowledge and seven aspects of Environmental Attitudes in four subgroups: Demographic variables, Knowledge, Attitudinal variables and Behavioural variables. Questionnaires were used to measure these four subgroups. Education, Single-family houses and intrinsic motives had a significant correlation with separation behaviour. Environmental knowledge had a negligible and not significant correlation. Bar (2011) and Richardson and Havlicek (1978) found similar results. Bar (2011)

(10)

concluded that young, female, single-family dwelling, high-income earning, well-educated and political liberal individuals tend to play an active part in reducing waste generation and recycling of waste. Richardson and Havlicek (1978) found a negative effect of income on the generation of waste that has to be separated. This infers that citizens with higher income levels have less to separate, because the total amount of recyclables to separate decreases. Barr (2011) focused his research on the independent variables: environmental values, situational variables and psychological factors as independent variables. Environmental values had a negligible effect on either separation intention or behaviour.

2.3 Determinants influencing policy

Since this thesis is focused on the effect of collection systems on municipal levels it is important to know if municipalities influence each other with regard to policy decisions. Fredriksson & Millimet (2002) found a positive relationship between states and their environmental policy. Specifically, they found an asymmetrically response to abatement cost

in neighbouring states. Abatement cost is referred as the cost of reducing environmental negatives such as pollution. States with relatively higher abatement

costs are more responsive to policy changes in other states than states with relative low abatement costs. This asymmetrical response could be a factor in the relationship between the collection systems and their separation rates.

Various other determinants correlate highly with the policies with regard to recycling. Feiock & West (1993) found that a higher average disposable income levels in a municipality seems to positively influence the probability that it will provide separation services. This correlation influences the effect of collection services on separation rates. In particular if higher disposable income levels have an effect on separation rates. Another interesting finding is the influence of pressure group on the policy with regard to waste management. Municipalities with a higher percentage of pressure groups have a higher probability to alter their policy in a recycling friendly way. Finally, influence of the government cannot be left out. Governments implement targets that consequently influence the policy decision of a municipality.

(11)

3 Model

3.1 The model

As mentioned in the literature review research on the effect of the different Dutch waste collection systems has not been done so far. To analyse the effect of waste collection systems three models are implemented: Logit regression, Ordinary Least Squares (OLS)–regression and a Fixed Effect (FE) regression. The first part offers an explanation of the models and of independent variables that are included. The second and third part gives an argumentation for the included variables and a deepened description of these variables. Most of them are already briefly defined in the literature review. The fourth part covers the data sources and the fifth part covers the descriptive statistics. The section finishes with the hypotheses.

The independent variables:

Bring: Dummy variable for a separate collection system where households have to bring their recyclables to a nearby container.

Get: Dummy variable for a separate collection system where the recyclables are collected by the municipality at their homes.

UBP: Dummy variable whether a municipality has a UBP system or not. Citizens: Total population of a municipality.

Unemployment: Unemployment rate in percentage points.

Disposable Income: Average income in a municipality corrected for inflation.

Logit regressions are implemented in order to examine the determinants of Dutch municipalities’ choice to adopt a certain waste collection system or UBP system. The regression includes all the independent variables described in section 3.1. Variables related to the collection systems could influence the estimate of the coefficient. These variables are therefore important to mention in order the make a legitimate conclusion

An OLS- regression is implemented to study the correlation between the determinants and the separation rate. Though this regression gives a good impression of possible correlations, it does not show causal relationships. Example: Even though the waste collection service does not change in municipalities and over time, the OLS-regression could show a biased estimator of collection services. The OLS estimate is likely to be biased when environmental progressive municipalities with high separation rates have a certain type of waste collection

(12)

system. Not correcting for environmental activism of the municipalities causes overestimations of the parameter (Dijkgraaf& Gradus, 2009). A way to improve this model is the implementation of a FE regression on the same panel data.

FE regressions are implemented in order to acquire an estimate of the coefficient of the two collection systems based on changes in the municipalities over time. The first FE regression includes all the municipalities, also the ones where no change has happened with regard to their waste collection systems. The second FE regression includes only the municipalities with a policy change with regard to waste collection systems. FE regression analyses the relationship between the independent variables and the separation rate within the municipality. Each municipality has its own characteristic that could possibly influence the estimates of the coefficients of the independent variables. In my analysis this could be the earlier mentioned environmental values of individuals in municipalities. Environmental values of individuals are likely to stay stable over a short period of time and are therefore stated as time-invariant. FE regression controls for these time-invariant characteristics. This is possible because the assumption is made that the determinants can correlate with the municipality’s error term. The model holds if these time-invariant characteristics of a municipality are unique and are not correlated with other characteristics of other municipalities. Each municipality is different. Therefore, the municipality’s error term and the constant should no correlate with other municipalities. Violations of these assumptions and other potential drawbacks are discussed in section five. In section five various test are applied to control for the assumptions and these potential drawbacks.

3.2 The dependent variable

The dependent variable presents a percentage of the total of fine household waste that has been separated at the source. The source could be a home of a household or a container nearby. The separation in municipalities with the system subsequent separation is not included. This thesis will focus on the separate collection system, mainly because it is merely impossible to measure recycling behaviour of household in the subsequent separation system without using questionnaires. The exact formula used by the national data monitor of the ministry of infrastructure and environment:

𝑇𝑜𝑡𝑎𝑙 𝑟𝑒𝑐𝑦𝑐𝑙𝑒𝑑 𝑓𝑖𝑛𝑒 ℎ𝑜𝑢𝑠𝑒ℎ𝑜𝑙𝑑 𝑤𝑎𝑠𝑡𝑒

(13)

Compost, Paper, Glass, Small chemical waste, Metal cans, Metal packages, Drinking cartons, Synthetic waste, Pampers, Textiles, Oil (fried), and Mixing fractions are defined as fine household waste.

3.3 Determinants

In section one and two it is shown that the Netherlands has two main types of collection systems: subsequent separation and separate collection (Nedvang, 2017). As subsequent separation is not included in the model al the determinants in the regression represent separate collection. Separate collection refers to a system where households separate recyclables at the source. Nedvang (2017) distinguishes three systems of separate collection. Municipalities either use Bring-system, Get-system or Get/Bring-system. As a result of this distinction by Nedvang (2017) this thesis focuses on these three systems and solely measures an effect of choices between these three systems. The Get/Bring system is the basis of the model, because most municipalities alter their collection system from or to a Get/Bring-system.

As mentioned before, two pricing systems exist in the Netherlands: One Fixed Fee system (FF) and a Unit Based Pricing system (UBP). UBP includes four systems: Volume, frequency, bag and weight-based. The dummy variable UBP takes the value of 1 if the municipalities use one of these four systems. FF includes two systems: number of people and fixed fee (Dijkgraaf& Gradus, 2009). The dummy variable UBP takes the value of 0 if the municipalities use one of these two systems. A system where people have to pay for extra residual waste is likely negatively affecting the separation rates and waste generation. This prediction is substantiated with the fact that recycling leads to less residual waste.

3.4 Data sources

The database including waste collection services, pricing systems, citizens and unemployment rates of the years 2012, 2013 and 2014 are collected from the organization Nedvang and Afvalfonds Verpakkingen.. Some of the data are openly available on the sites of these organizations. However this database was not large enough to implement a FE regression. This database is therefore combined with the received data from a contact person and researcher of the Afvalfonds Verpakkingen. Nedvang is affiliated to the organization Afvalfonds Verpakkingen. Afvalfonds Verpakkingen is an organization that helps with building a waste management structure that stimulates recycling. Research on the topic of this thesis is one of the main tools to build desirable sustainable waste management structures. The data with dependent variables of separation rates of municipal waste with the same

(14)

time-period are collected from the national data monitor of the Ministry of Infrastructure and Environment (Afval Datamonitor). Most of the Unit Based Pricing System and Collection Systems data are cross-checked with the database of this monitor. Unemployment rates are verified with the unemployment rates collected from the Centraal Bureau van Statistiek (CBS), the Dutch statistical agency. The disposable income is measured as the average of the relevant population per municipality and is also collected from the CBS database. Especially the data of waste collection systems and UBP systems received from my contact person is hard to verify and could therefore lead to biased results. Throughout this thesis it is assumed that the data are correct and correctly interpreted. False data and incorrectly interpreted result and data are solely my responsibility.

3.5 Descriptive Statistics

Table 3.1: Descriptive statics of the included variables

Variable Obs Mean Std. Dev. Min Max

Separation rate 993 52,28827 12,60747 9 95 Citizens 1059 41305,92 60461,04 1501 810937 Unemployment 1059 5,887347 1,205136 3,5 12,6 Disposable Income 728 29385,32 3,668156 23,1 48,92 Waste Costs 1046 225,6447 48,40758 91,9 400,6 Bring-system 1009 0,280476 0,449454 0 1 Get-system 1009 0,401388 0,490422 0 1 Get/Bring-system 1009 0,318137 0,465984 0 1 UBP 1053 0,425451 0,494646 0 1

The numbers of observations are the observations over three years. In this table Separation rate includes data from 353 municipalities. The Netherlands has a decreasing number of municipalities. The total number of municipalities decreased from 415 in 2012 to 403 in 2014 (Centraal Bureau Statistiek, 2016). This decrease is caused by the merging of multiple municipalities. Omitted observations are either caused by these mergers or the lack of enough information to process the data in usable observations. Table 3.1 displays a couple of interesting observations. First, the separation rate of fine household waste is approximate 52%. Second, table 3.1 shows that 42,5% of the municipalities have a UBP system of waste collecting over the three years. In 2012 the percentage was 41,3 %. Municipalities are altering their policy to UBP systems to stimulate separation. This is a trend over the last fifteen years and this trend will likely continue over the next decade (Dijkgraaf & Gradus, 2009). Third, 40,1% of the municipalities uses -a system where the recyclables are collected at the homes.

(15)

This percentage was 42% in 2012. In these three years multiple municipalities altered their policy with regard to waste collection systems. Table 3.2 displays these alterations.

Table 3.2: Waste collection systems alterations

To

Get Bring Get/Bring Total Percentage

Get 1 21 22 40

From Bring 5 14 19 34,5

Get/Bring 10 4 14 25,5

Total 15 5 35 55 100

Percentage 27,3 9,1 63,6 100

63.6 per cent of the municipalities that altered their collection system, switched from one mode of collection to a combination of a Bring-system and Get-system. An explanation could be that new containers have to be bought and new waste collection services have to be hired. The Get/Bring-system as a transition phase is therefore logical. It is interesting to know if an alteration to or from a one- modus collections system has an effect on the separation rates. Only five of the municipalities altered directly from a Bring-system to a Get-system and only one altered directly from a Get-system to a Bring-system. All the other municipalities waste system alterations involved the system of Get/Bring. The Get/Bring-system is therefore the default in the OLS- regression and the FE regression.

All the observations do not include the municipalities with subsequent separation. The observations could therefore be biased. All the subsequent separation systems are in the provinces Groningen and Friesland (Nedvang, 2017). Groningen and Friesland have lower average of disposable incomes and a higher unemployment rates (CBS, 2016).This could influence the estimates of the coefficient. This potential drawback will be discussed further in section five.

(16)

4 Hypothesis

The hypothesis: separate waste collection systems where households have to make less effort to separate their recyclables are likely to stimulate separation behaviour of households. The system where the recyclables are collected by the municipality at their house is therefore expected to stimulate recycling behaviour more than the other two separate collection systems. However 30.1 % of the municipalities in the Netherlands use a combination of the Get and Bring system. The Get-system is expected to have a higher positive effect on the separation rate than the Get/Bring-system. However the structure of the Get/Bring systems is not clear. It is almost impossible to measure for every municipality the number of waste containers available and the frequency of recyclables collection at the homes. Bring/Get systems may have numbers of containers comparable to Bring-systems and the frequency of recyclables collection could be comparable to Get-systems in the same municipality at the same time. Municipalities that have best of both are probably more stimulating for separation of recyclables. Individuals may have more incentive to separate in Bring-systems, but others may have more incentive to separate in a Get-system.

Based on previous research mentioned in section two, UBP is expected to have a positive effect on the separation rate of municipalities. Separating recyclables results in less residual waste and a system where people have to pay for extra residual waste is likely to positively affect the separation rates (Elbert Dijkgraaf & Gradus, 2009).

Collections systems of municipalities with large number of citizens are expected to have lower separation rates. Kandel & Lazear (2017) found that peer pressure is an important factor for social good behaviour. Peer pressure is higher in small communities and thus in municipalities with lower numbers of citizens (Kandel & Lazear, 2017). Municipalities with large number of citizens are therefore less likely to separate recyclables.

The effect of unemployment is probably ambiguous. Unemployment is highly correlated with education (Mincer, 1974). People with higher education levels tend to participate more in recycling (Barr, 2015) . Based on these studies unemployment should have a negative effect on the separation rate. However unemployed citizens have also more time to separate recyclables than working citizens, which could result in higher separation rates.

(17)

5 Results

This section is structured as follows. First, the correlation between the independent variables is briefly explained with the implementation of logit regressions. Second, the OLS regression results are analysed. Finally the Fixed Effect regression results are analysed.

5.1 Logit regression

Several variables and entities influence the effect of policy decisions on environmental issues. The main problem with analysing the effect of these policy decisions is the probability that individuals respond differently to a certain policy choice by the government. These differences are recognised, but cannot be fully implemented in the model. To measure the effect of altering a waste collection system in a Dutch municipality, the following variables are taken into account: separation rate, waste collection systems, citizens, unemployment, year 2013 and 2014. It is important to know if these variables influence each other. Environmental progressive municipalities are likely to have UBP systems and a certain waste collection system. Logit regressions are implemented to examine these correlations.

Table 5.1: Logit regression on the collection systems and UBP

(1) (2) (3) (4)

VARIABLES Get Bring Get/Bring UBP

Unemployment 0.197 -0.155 0.0168 0.0564

(0.202) (0.200) (0.195) (0.205)

Citizens -1.74e-05*** 1.33e-05** -1.88e-06 -1.14e-05*

(6.75e-06) (5.25e-06) (3.57e-06) (6.33e-06)

Disposable income -0.134*** -0.000529 0.118*** -0.167*** (0.0485) (0.0410) (0.0415) (0.0500) Get 0.0613 (0.342) Bring -1.058*** (0.384) UBP 0.574** -1.121*** 0.388 (0.288) (0.326) (0.313) Constant 2.937* -0.0968 -4.652*** 5.020*** (1.759) (1.614) (1.651) (1.817) Observations 240 240 240 240

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(18)

From table 4.1 several conclusions can be drawn. First, unemployment has no correlation with the dependant variables at a significance level of p<0.1. Unemployment seems to have no influence on the decision-making with regard to waste management. Second, the number of citizens in a municipality has a significant negative correlation with the Get-system, a significant positive correlation with Bring-system and a significant negative correlation with UBP. A possible explanation could be that in densely populated cities it is easier to implement containers instead of recyclables collection services at homes. It is also harder to implement a UBP system in densely populated cities. Maintaining a UBP system gives administrative work and for larger numbers of citizens it is harder to control for. Third, average disposable income has a negative correlation with the Get-system, positive correlation with the Get/Bring-system and a negative correlation with UBP. Most of the municipalities that use a Get and UBP system are located outside the rich urban provinces Noord-Holland and Zuid-Holland (table 5.2). Therefore, it makes sense that poorer municipalities correlate with the Get-system. Finally, UBP correlates positively with the Get-system and correlates negatively with the Bring-system. Municipalities with a UBP system seem to also use a Get-system instead of Bring-system.

Table 5.2: Correlation table

Noord-Holland Zuid-Holland Get Bring Get/Bring UBP Disposable income

Noord-Holland 1 Zuid-Holland -0,17 1 Get -0,07 -0,06 1 Bring 0,13 0,19 -0,53 1,00 Get/Bring -0,05 -0,12 -0,54 -0,43 1,00 UBP -0,30 -0,33 0,21 -0,26 0,03 1,00 Disposable income 0,29 0,18 -0,23 0,07 0,18 -0,23 1

(19)

5.2 OLS- Linear Regression

In this section the OLS- regression results are examined. The waste collection systems are regressed on the separation rate and step by step other independent variables are added.

Table 4.3: Stepwise OLS regression

VARIABLES Separation Rate Separation Rate Separation Rate Separation Rate Separation Rate Bring -7.404*** -4.936*** -4.107*** -4.274*** -4.085*** (0.971) (0.831) (0.793) (0.787) (0.775) Get 3.291*** 1.142 0.726 0.725 1.004 (0.897) (0.766) (0.728) (0.722) (0.711) UBP 13.50*** 12.72*** 12.84*** 12.87*** (0.671) (0.641) (0.636) (0.626)

Citizens -5.17e-05*** -4.10e-05*** -3.11e-05***

(4.98e-06) (5.55e-06) (5.76e-06)

Unemployment -1.190*** -2.372*** (0.280) (0.351) year13 3.283*** (0.869) year14 5.311*** (0.896) Constant 53.13*** 47.26*** 49.73*** 56.29*** 59.81*** (0.663) (0.630) (0.644) (1.669) (1.762) Observations 983 977 977 977 977 R-squared 0.120 0.379 0.441 0.451 0.471

Table 4.2 displays several interesting results. First the Bring-system has a negative correlation with the separation rate for a significance level of p<0.01. This correlation decreases substantially when UBP is included. Altering from a Get/Bring system to a Bring-system seems to affect the separation rate negatively. This is in conformity with hypothesis. Second, the Get-system has only a significant correlation in regression one. Including UBP causes insignificant results of the Get-system. An explanation for this could be the positive correlation between UBP and the system. Another possible explanation is that the Get-system and The Get/Bring –Get-system are too identical. The structure of the Get/Bring Get-systems is not very clear. There is a possibility that the Get/Bring systems are identical to Get-systems with a few additional containers for recyclables. Alterations to a Get-system will than have

(20)

little effect on the separation rate. However the FE regression is a better instrument to test for a potential causal relationship. These results are compared in section 4.3. Third, a UBP system has as expected positive a significant correlation with separation rate. The size of the impact is high compared with the other variables. Based on this correlation introducing a UBP system increases the separation rate with thirteen percentage points. This result is in conformity with the results presented by Dijkgraag& Gradus (2009). Fourth, the correlation of citizens and unemployment are in conformity with the hypothesis stated in section 3.7.

5.3 Fixed Effect Regressions

In this section the Fixed Effect regression results are examined. The FE regression is divided in two parts: 1: FE regression with all the municipalities and 2: FE regression with only municipalities that altered their waste collection system. In both regressions the municipalities that altered their waste pricing system are not taken into account.

Table 4.3: FE regression with all municipalities

(1) (2) (3) (4) VARIABLES Separation Rate Separation Rate Separation Rate Separation Rate Bring -2.652*** -2.654*** -1.957** -1.596* (0.937) (0.940) (0.929) (0.908) Get -1.790** -1.790** -1.430** -1.360** (0.713) (0.713) (0.701) (0.684)

Citizens -3.43e-06 -8.80e-05 -0.000119

(0.000120) (0.000118) (0.000115) unemployment 0.862*** 0.963* (0.162) (0.543) Year 2013 -0.989 (0.789) Year 2014 0.565 (0.855) Constant 53.72*** 53.86*** 52.03*** 52.76*** (0.427) (5.109) (5.013) (5.601) Observations 968 968 968 968 R-squared 0.019 0.019 0.062 0.112 Number of gemeente1 337 337 337 337

(21)

From these results several conclusions can be drawn. The Get-system and Bring-system have a negative fixed effect on the separation rate at a significance level of p<0.05. The results are in contrast with the earlier found correlation of the Get-system in the OLS regression and the hypothesis. If the results are correct, municipalities that alter their Get/Bring-system to a Get-system are facing a negative effect on their separation rates. This means that combinations of the two one modus systems give the best result in term of separation behaviour. An explanation could be that Bring/Get systems may also have numbers of containers comparable to Bring-systems, but frequency of recyclables collection could also be comparable to Get-systems in the same municipality. Municipalities that have best of both are probably more stimulating for separation of recyclables. In addition, individuals may not want keep their recyclables at their homes for a long time and the option to also bring it to a nearby containers lowers their transaction costs for separating recyclables. However the structure of these Get/Bring systems is not very clear. Further research is needed to analyse the specific differences of the three distinguished waste collection systems.

The most surprising result is the positive effect of unemployment. Although it is only significant for a significance level of p<0.001 in regression four and p<0.1 in regression five, it is in both regressions positive. This is in contrast with the hypothesis and the OLS correlation. In section 3.7 it is mentioned that the negative correlation between education and unemployment and the positive correlation between education and separating of recyclables could be the reason for a negative effect of unemployment on separation rate. Unemployment increases over the years 2012-2014 as a result of slow economic growth and the euro crisis (CBS, 2015). Level of education on the other hand increased also in the same period (CBS, 2016). In 2012 1.44% of the population had a bachelor degree or higher in the Netherlands. This number increased to 1.512 % of the population in in 2014 (CBS, 2016). FE regression analyses the relationship between the determinants and the separation rate within the municipality in these three years. Consequently the reasoning in the hypothesis does not hold for a FE regression, because the education levels increased instead of decreased in the years 2012-2014. Another explanation could be that unemployed citizens have more time to separate and because of the increased education level they still have the same or more incentives to separate their recyclables. This explanation seems doubtful and probably other omitted variables are relevant, but not included in the present research. Further research has to be done to substantiate the findings of this regression.

The variables time and number of citizens do not have any causal effect on the separation rate. In this regression the environmental values and other not included variables of the

(22)

municipality are stated as fixed. The assumption is made that these environmental values and omitted variables of the households do not substantially change. A substantial change in these variables causes inconsistent estimators of the FE regression. It seems reasonable to conclude that environmental values of households do not fluctuate much on average in a municipality in the period 2012-2014. The insignificant results of year 2013 and year 2014 substantiate this reasoning.

All the estimators are significantly lower than the estimator of the OLS-estimator. The OLS regression does not include environmental correction and is therefore causing overestimation of the parameters (Elbert Dijkgraaf & Gradus, 2009). The OLS- regression has also a high probability for correlation between the independent variables and the error term. This endogeneity could cause biased estimates of the coefficient.

Table 5.4: FE regression with only municipalities that altered their waste collection system (1) (2) (3) (4) VARIABLES Separation Rate Separation Rate Separation Rate Separation Rate Bring -2.807*** -2.940*** -2.288*** -1.948** (0.767) (0.788) (0.820) (0.817) Get -2.255*** -2.261*** -1.909*** -1.687*** (0.584) (0.585) (0.591) (0.581) Citizens -0.000246 -0.000356 -0.000454 (0.000321) (0.000318) (0.000312) Unemployment 0.844** 4.058*** (0.363) (1.293) Year 2013 -4.835*** (1.813) Year 2014 -4.688** (1.935) Constant 53.13*** 66.04*** 66.45*** 55.06*** (0.348) (16.89) (16.50) (16.80) Observations 147 147 147 147 R-squared 0.206 0.211 0.255 0.311 Number of gemeente1 51 51 51 51

Regression two is comparable with regression one. The slopes of the coefficients and the significance levels are similar to the results of regression 1. However, some coefficients are remarkable. The coefficients of the collection system and unemployment rate increased

(23)

substantially. Especially the increase of the coefficient of unemployment is notable. Furthermore the coefficients of the year variables not only increased but also turned significant for α<5. The increase of the coefficient is understandable if you look at the sample size. In regression two only municipalities that altered their waste collection system are included. These results are likely to be biased, because there could be other reasons that only these 51 municipalities altered their waste collection system.

6 Assumption Statistics

Although the results look promising several assumptions and potential drawbacks have to be taken into account.

The most essential drawback is that the dataset provides an incomplete representation of the real structures of the Get/Bring systems. From the dataset it is possible to know if municipalities use a Bring-system and a Get-system at the same time. A Get/Bring system could have many containers or very little. The frequency of recyclables collection is also not known. This incomplete representation of the real structures of the collection systems could lead to biased results.

Another potential drawback is the number of municipalities that altered their waste policy. Six of the 331 municipalities changed their FF pricing system to a UBP system or vice versa and 50 changed their waste collection system. These municipalities are not clustered in one area or province. This could be a potential drawback for the generalization of the results for the whole country. The effect of policy changes probably varies per province or area with a similar profile.

For unbiased estimates of the coefficients of the following FE regression several assumptions have to hold.

1. Strictly exogenous, no feedback 2. The constant should be normalized 3. Homoscedasticity/ Serial Correlation

Strictly exogeneity is violated if the independent variables contain feedback. This infers that separation rates in year 2012 are influencing the policy with regard to waste management in year 2013. For example: a municipality has low separation rates and therefore alters its FF pricing system to a UBP system. This violation is tested in table 5.1 with the included future dummy variables of Bring and Get. D_bring takes a value of 1 in 2012 if the dummy Bring in

(24)

year 2013 takes a value of 1. A significant coefficient causes a violation of assumption strictly exogeneity.

Table 6.1: Feedback test

(1) (2) (3) (4) VARIABLES Separation Rate Separation Rate Separation Rate Separation Rate Bring -3.210*** -2.899*** -3.036*** -2.976*** (0.896) (1.000) (0.964) (1.006) Get -1.854*** -1.500 -1.596 -1.590 (0.712) (1.574) (1.578) (1.580) D_Bring 0.538 0.307 (1.414) (1.450) D_Get -0.573 -0.540 (0.718) (0.737) Constant 53.95*** 53.09*** 53.53*** 53.42*** (0.418) (0.833) (0.775) (0.948) Observations 983 642 642 642 R-squared 0.026 0.034 0.036 0.036 Number of gemeente1 342 335 335 335

From table 5.1 one conclusion can be drawn. The dummy variables D_bring and D_Get are both insignificant for P<0.1. Based on this test, violation of strict exogeneity can be ruled out.

Heteroscedasticity results in biased test statistics and biased confidence intervals. The homoscedasticity is tested with the General Least Squares method and serial correlation is tested with the Wooldridge test for autocorrelation panel data. Homoscedasticity and serial correlation are not violated with a significance level of p<0.1. The results of these tests indicate that all the violations of the assumptions can be ruled out.

(25)

7 Conclusion

To my knowledge, this thesis is the first quantitative study on the influence of differences in waste collection systems of Dutch municipalities on the separation behaviour of households. This thesis distinguishes three waste collection systems: Bring-system, Get-system and Get/Bring-system, two main pricing systems: Unit Based Pricing (UBP) and Fixed Fee system. Furthermore, several other variables such as unemployment, number of citizens and disposable income relevant for separation behaviour are included in the analysisr. For the analysis three types of models are implemented: Logit regression, OLS-regression and Fixed Effect regression.

The OLS- regression and the FE regression show contradictory results. The OLS- regression suggest that alterations from a Get/Bring system to a Bring-system have negative effect on the separation behaviour of households. An alteration from Get/Bring to Get-systems has only a positive correlation if the independent variable UBP is not included.

In contrast the results of the FE regressions indicate different effects of the waste collection systems on the separation behaviour of households. The results suggest that the Get/Brings gives the best results in term of separation behaviour. Altering waste collection systems to either Bring or Get-systems- has a negative effect on the separation behaviour. Municipalities that have best of both are probably more stimulating for separation of recyclables. Get/Bring systems may also have numbers of containers comparable to Bring-systems and at the same time the frequency of recyclables collection could also be comparable to Get-systems in the same municipality. An alteration from Get/Bring to Bring-system decreases the separation rates more than an alteration to a Get-Bring-system. This means that the Bring systems stimulate separation behaviour less than Get-systems. This is in conformity with the hypothesis.

The effects of the other variables are less evident. The most surprising finding is the effect of unemployment. Unemployment is expected to affect separation rates negatively. This hypothesis is based on the negative correlation between unemployment and education and the positive correlation between education and separation behaviour. However, both education and unemployment increased in the years 2012-2014. Therefore, this hypothesis is rejected. An explanation for the positive effect of unemployment could be that individuals have more time to separate their recyclables when they are unemployed. Probably other omitted variables are relevant too, but not included in the present research. Further research on these findings is

(26)

The explanatory variables for Dutch municipalities’ choice to adopt a certain waste collection system are understandable. The rich and densely provinces Zuid-Holland and Noord-Holland have a positive correlation with the Bring-system and Disposable income. Bring-systems are easier to maintain in densely neighbourhoods. Municipalities that adopt UBP systems tend to use also Get-systems. This correlation could influence the results. Another potential drawback could be that the structure of the Get/Bring systems is unclear at the moment. The collected data provide an incomplete representation of the real structures of these combination systems. Some structures could be similar to the Bring-systems or others could be similar to Get-systems. Further research on these combination systems is therefore needed.

Although, as explained above, the results of the resent research must be treated with caution and although further research on several questions is needed, it seems possible to conclude that the combination of the Bring and Get systems gives the best results in term of separation behaviour. The Bring system gives the worst results in term of separation behaviour.

(27)

References

Barr, S. (2015). Factors Influencing Environmental Attitudes and Behaviours. Environment and

Behaviour, 39(4), 435–473.

Bergsma, G. C., Bijleveld, M. M., Otten, M. B. J., & Krutwagen, B. T. J. M. (2011). LCA: recycling van kunststof verpakkingsafval uit huishoudens.

Centraal Bureau voor de Statistiek (2017). Inflatie; CPI, vanaf 1963. Den Haag: CBS. Retrieved from: http://statline.cbs.nl/StatWeb/publication/?VW=T&DM=SLNL&PA=70936NED&D1=0&D2

=(l-34)-l&HD=081020-1258&HDR=T&STB=G1

Centraal Bureau voor de Statistiek (2016). Gemiddeld inkomen van personen naar kenmerken en naar regio. Den Haag: CBS. Retrieved from:

http://statline.cbs.nl/Statweb/publication/?DM=SLNL&PA=80500NED&D1=2-3&D2=a&D3=0&D4=101-543&D5=a&HDR=G4,G2,G1,T&STB=G3&VW=D

Centraal Bureau voor de Statistiek (2016). Leerlingen, deelnemers en studenten; onderwijssoort, woonregio. Den Haag: CBS. Retrieved from:

http://statline.cbs.nl/Statweb/publication/?DM=SLNL&PA=71450ned&D1=0&D2=0&D3=0 &D4=a&D5=0&D6=0,6-18&D7=12,14&VW=T

Centraal Bureau voor de Statistiek (2012). Retrieved from: https://www.cbs.nl/nl-nl/nieuws/2012/21/werkloosheid-stijgt

Centraal Bureau voor de Statistiek (2017). Retrieved from: https://www.cbs.nl/nl-nl/nieuws/2017/03/in-2016-sterkste-daling-werkloosheid-in-tien-jaar

Dijkgraaf, E., & Gradus, R. (2009). Environmental activism and dynamics of unit-based pricing systems. Resource and Energy Economics, 31(1), 13–23.

https://doi.org/10.1016/j.reseneeco.2008.10.003

Dijkgraaf, E., & Gradus, R. H. J. M. (2004). Dutch cost savings in unit-based pricing of household waste. The Waste Market: Institutional Developments in Europe, 26, 111–130.

Duggal, V. G., Saltzman, C., Williams, M. L., Duggal, V. G., Saltzman, C., & Williams, M. L. (2016). Recycling : An Economic Analysis. Eastern Economic Journal, 17(3), 351–358.

El-Fadel, M., Findikakis, A. N., & Leckie, J. O. (1997). Environmental Impacts of Solid Waste Landfilling. Journal of Environmental Management, 50(1), 1–25.

Feiock, R. C., & West, J. P. (1993). Testing Competing Explanations for Policy Adoption: Municipal Solid Waste Recycling Programs. Political Research Quarterly, 46(2), 399–419.

Fredriksson, P. G., & Millimet, D. L. (2002). Strategic Interaction and the Determination of Environmental Policy across U.S. States. Journal of Urban Economics, 51(1), 101–122. Fullerton, D., & Kinnaman, T. C. (2000). Garbage and Recycling with Endogenous Local Policy .

Journal of Urban Economics, 48, 419–442.

Fullerton, D., & Kinnaman, T. C. (2016). American Economic Association Household Responses to Pricing Garbage by the Bag. The American Economic Review, 86(4), 971–984.

(28)

Korean experience. Journal of Environmental Management, 57, 1–10.

Jahre, M. (1995). Household waste collection as a reverse channel: A theoretical perspective.

International Journal of Physical Distribution & Logistics Management, 25(2), 39–55.

Jenkins, R. R., Martinez, S. A., Palmer, K., & Podolsky, M. J. (2003). The determinants of household recycling: A material-specific analysis of recycling program features and unit pricing. Journal of

Environmental Economics and Management, 45(2), 294–318.

Kandel, E., & Lazear, E. P. (2017). Peer Pressure and Partnerships. Journal of Political Economy,

100(4), 801-817.

Kinnaman, T. C. (2006). Examining the Justification for Residential Recycling. Journal of Economic

Perspectives, 20(4), 219–232.

Linderhof, V., Kooreman, P., Allers, M., & Wiersma, D. (2001). Weight-based pricing in the

collection of household waste: The Oostzaan case. Resource and Energy Economics, 23(4), 359– 371.

Mincer, J. (1991). Education and E mployment. National Bureau of Economic Research, 3838. Morris, G. E., & Holthausen, D. M. J. (1994). The economics of household solid waste. Journal of

Economics and Management, 26, 215–234.

Nakamura, S. (1999). An interindustry approach to analyzing economic and environmental effects of the recycling of waste. Ecological Economics, 28, 133–145.

Oskamp, S., Harrington, M. J., Edwards, T. C., Sherwood, D. L., Okuda, S. M., & Swanson, D. C. (1991). Factors influencing household recycling behavior. ProQuest Information and Learning

Company, 23(4), 494–519.

Richardson, R. A., & Havlicek, J. (1978). Economic analysis of the composition of household solid wastes. Journal of Environmental Economics and Management, 5(1), 103–111.

Rijksoverheid (2016). Retrieved from:

https://www.rijksoverheid.nl/onderwerpen/afval/inhoud/huishoudelijk-afval.

Saltzman, C., Duggal, V. G., Williams, M. L., & Farrell, O. (1993). Income and the recycling effort : a maximization problem. Energy Economics, 33–38.

Shah, R. (2007). Waste statistics Session 8 : Waste statistics. In Session 8: Waste statistics (pp. 16– 20).

(29)