Municipal Waste Kuznets Curves: Evidence on Socio-Economic Drivers and Policy Effectiveness from the EU

1 Introduction

Indicators of ‘decoupling/delinking’ are used to measure improvements in environmental/ resource efficiency with respect to economic activity. The European Union’s (EU) ‘thematic strategies’ on resources and waste, include reference to ‘absolute’ and ‘relative’ delink-

ing indicators ( EC 2003a , b ; Jacobsen et al. 2004 ). The former are a negative relationship between economic growth and environmental impacts associated to the descending side of an inverted U shape according to the Environmental Kuznets Curves (EKC) framework. The achievement of delinking is of prime importance for waste, which plays as important a role

in environmental impact and economic costs as climate change (Fig. 1 ).

The EEA ( 2007 ) acknowledges that “It is increasingly important to provide answers to these questions because waste volumes in the EU are growing, driven by changing produc- tion and consumption patterns”. EEA ( 2006 ) also highlights the importance of market-based instruments for achieving a stronger degree of delinking in waste indicators.

From a ‘waste accounting’ perspective, waste generation (WG), or the amount of col- lected waste, can be managed in three ways: recycling (R) and materials recovery, including composting (C), and incineration (I) with energy recovery; incineration (without energy recovery); and landfilling (L, which might also involve energy recovery). Incineration is a hybrid option. If we exclude the possibility of incineration without energy recovery, the net social benefits of incineration should be positive, categorising it within waste recovery rather than final disposal. Treated amounts of waste from the recycling and incineration processes will then need to be disposed of in (regulated) landfills; this is defined as disposal although

(million tonnes) 100

50 landfilling

municipal solid waste generation/landfilling 1980 1985

Fig. 1 Projected generation and landfilling of municipal waste in the EU25. Source EEA ( 2007 ), Figures from 1980–2004 are data from Eurostat. Figures from 2005–2020 are projections. BMW (Bio degradable Municipal waste)

Municipal Waste Kuznets Curves 205 100%

a Latvia

Cyprus Denmark

Greece Poland Germany

Lithuania Netherlands

Sweden Belgium

Czech Rep. Slovak Rep. landfill

Luxembourg

Estonia ‘03

calculated material recovery incineration with energy recovery

other recovery operations Fig. 2 Landfilling, incineration and materials recovery in 2004. Source EEA ( 2007 ), Eurostat Structural

Indicators on municipal waste generated, incinerated and landfilled, supplemented with national statistics

it involves some energy recovery. The mass balance equation is: WG = WR(C + R + I) + L, or WG = WR(C + I) + I + L. 1

EEA ( 2007 ) shows that countries can be categorised under three waste management ‘groupings’, according to their strategies for diversion of municipal waste away from land- fill, and the relative shares of landfilling, materials recovery (recycling and composting) and incineration. The first group comprises countries with high levels of materials recovery and incineration, and relatively low landfill levels. The second group includes countries with high materials recovery rates, medium incineration levels and medium level dependence on land- fill. The third group includes those countries with low materials recovery and incineration

levels and relatively high dependence on landfill (see Fig. 2 ).

The environmental impacts of landfilling are massive ( Pearce 2004 ; El-Fadel et al. 1997 ; Eshet et al. 2004; DEFRA 2005 ). Landfilling should not become the default best economic practice in all situations; its costs and benefits are influenced by economic and technolog- ical factors. For examples of economic assessments of different waste disposal strategies, see, among others, Pearce ( 2004 ); Dijkgraaf and Vollebergh ( 2004 ). It should be noted that reducing waste generation at source through the imposition of policy targets in terms of waste generated per capita, is probably the most effective and efficient way of handling the problem in the long run. Given its potential high cost in the short run, the first phase of policy implementation at EU level has focused on landfill diversion and increased levels of recycling/recovery, including incineration. There is a need to analyse empirically whether

1 If we include the possibility of waste prevention/reduction occurring through the implementation of ‘closed loop’ strategies in household and production processes, WG could be considered the net output of total waste

generated by economic growth and the part that is integrated in economic processes and decisions before col- lection. Waste flows recovered by closed loop integration may fall outside the core definition of waste flows as materials without any value for the economic agents that ‘produce’ them. It is clear that market evolution and policy may change the ‘economic’ notion of waste, such that does not always coincide with the legal counterpart.

206 M. Mazzanti, R. Zoboli

these policies are being effective in terms of changing the endogenous relationship between economic growth and waste trends. 2

This paper provides empirical evidence on delinking trends and Waste Kuznets Curves (WKC) for municipal solid waste (MSW). The primary aim is to provide preliminary robust empirical evidence for a vast regional area (the EU) on the economic and policy drivers of waste dynamics. Although waste policies have been in force for some time in the EU, and are a pillar of EU environmental policy, they have not been studied using quantitative meth- ods. Empirical evidence on WKC dynamics for waste is scarce. Research on delinking for materials and waste is far less developed than research on air pollution and greenhouse gas emissions. This limited research could be a problem for policy.

In addition, analyses using cross country, highly disaggregated panel data on waste are also very scarce, with most studies involving single countries at the regional, provincial or municipal levels. In spite of the significant environmental, policy and economic relevance of waste issues, there is very little empirical evidence on delinking even for major waste streams, such as municipal waste, packaging and other waste.

Analyses of policy effectiveness are also scarce, and generally dominated by studies of waste management optimization schemes or evaluation of externalities. Economic analy- ses are predominantly focused on cost benefit assessment of waste site externalities (fixed and variable) and on their acceptability to local communities, with or without compensation ( Gallagher et al. 2008 ). This focus on costs and benefits of siting decisions is due in part to the lack of reliable country level and within country data ( Pearce 2004 ). 3

Our first objective is to provide new empirical evidence on WKC and waste policy effec- tiveness in the EU, distinguishing between socio-economic and policy factors. The WKC is appropriate for assessing the effects of socio-economic and structural factors in the business as usual scenario, and the role of policy levers in explaining the eventual delinking between environmental pressure and growth, without which we can expect a lower or even no delink-

ing. A secondary objective is to identify differences between the EU 15 group of countries and new entrants to the EU, with respect to waste drivers, delinking trends and policy effects, depending on the actions taken in response to implementation of the 1999 Landfill Directive 4 and even earlier ‘policy actions’ by some countries (e.g. Germany).

The paper is structured as follows. Section 2 presents evidence on WKC, which high- lights the urgent need to intensify research in this area to enable policy evaluation. Section 3 defines the conceptual framework, the empirical model and comments on the empirical results regarding waste generation, incineration, and landfill options. Section 5 concludes.

2 EEA ( 2007 ) provides a qualitative assessment of landfill and incineration policy tools at EU and country level.

3 We quote, among others, Powell and Brisson ( 1995 ); Miranda et al. ( 2000 ); Eshet et al. ( 2004 ); Brisson and Pearce ( 1995 ); Dijkgraaf and Vollebergh ( 2004 ). Recently Caplan et al. ( 2007 ) provides an example of

how economic evaluation techniques inform landfill siting process, while Jenkins et al. ( 2004 ) provides sound and very interesting econometric analyses of the socio-economic factors explaining the levels of monetary compensation paid to communities hosting landfills.

4 “The Landfill Directive pursues two approaches: firstly to introduce stringent technical requirements for landfills; and secondly, to divert biodegradable municipal waste (BMW) from landfills by setting targets for

the landfill of BMW in 2006, 2009 and 2016. Even more ambitious targets for the post-2016 period have recently been proposed by the European Parliament. The targets are based on the quantity generated in 1995, and the main implication of this approach is that there is an absolute limit placed on the quantity of biode- gradable municipal waste (in tonnes) that can be landfilled by the specific target dates” (EEA 2007, P. 7). The Incineration Directive (Directive 2000/76/EC on the incineration of waste) is an ancillary and complementary piece of the EU waste policy strategy which also includes the Waste Framework Directive, legislation that was revised at the end of 2008 ( EC 2008 ), but which still does not identify clear policy targets.

Municipal Waste Kuznets Curves 207

2 Delinking Analysis and WKC: Empirical Evidence

Since the pioneering works of Grossman and Krueger ( 1994 ), the World Bank ( 1992 ), Holtz-Eakin and Selden ( 1992 ), interest in the so-called Environmental Kuznets Curve (EKC) has increased.

Theoretical studies of EKC are rather scarce although there are some contributions which try to establish some empirical foundations for EKC. These studies generally try to explain EKC dynamics in terms of types of technological investments, endogenous growth spillovers, changing preferences, and policy factors ( Chavas 2004 ). Andreoni and Levinson ( 2001 ) is a seminal work that suggests that EKC dynamics quite simply may be technologically micro- founded, and not strictly related to growth and externality issues. At a more macroeconomic level, Brock and Taylor ( 2004 ) integrate the EKC framework and the Solow model of eco- nomic growth. Notwithstanding the increasing relevance of theoretical studies on EKC, it is quantitative analysis that dominates, but there is scope for more research at the margins. In the waste realm, Highfill and McCasey ( 2001 ) and Huhtala ( 1997 ) are some out of a few studies which provides a theoretical underpinning for WKC dealing with dynamic optimal ‘investments’ in waste management and waste disposal options.

We argue that future empirical work should concentrate on using newly constructed, more heterogeneous and longer datasets, either at country level or for samples of countries in homogenous relevant areas, rather than at the cross country international level, where the results of analyses are rather general and whose ‘average’ figures can hide some important features arising from the econometric analysis.

Below we briefly review the evidence on waste delinking, WKC, and model-based waste policy evaluation. Table 1 presents a summary of the main works.

The report that gave birth to the EKC literature ( World Bank 1992 ) presented some evi- dence based on cross country regression analysis of data from the 1980s; no WKC were found. A more recent report ( DEFRA/DTI 2003 ) highlights the positive elasticity of waste generation to income as a primary concern for policy: waste generation seems still to be characterised by a strict positive relationship with economic drivers.

One of the earliest WKC studies was by Cole et al. ( 1997 ), who found no evidence of an inverted U-shape for municipal waste. They found no turning point (TP) in their study, which exploited data on MSW, for the period 1975–1990, for 13 OECD countries. Similarly, Seppala et al. ( 2001 ), in a study of five industrialised countries including Japan, the US and Germany, over a similar time period (1970–1994), also found no evidence of delinking regarding ‘direct material flows’. However, there is some emerging evidence of delinking, although for quite specific (waste) indicators. Raymond ( 2004 ) presents evidence of a WKC based on a waste/consumption indicator derived from environmental sustainability indexes (ESI). Berrens et al. ( 1998 ) and Wang et al. ( 1998 ), focusing on stocks of hazardous waste in the US and exploiting a county-based cross sectional dataset also find evidence in favour of a negative elasticity.

A 2004 study by Johnstone and Labonne using panel data on solid waste in the OECD coun- tries, provides evidence of economic and demographic determinants of rates of household solid waste generation, regressed over consumption expenditure, urbanisation and popula- tion density. This study finds positive elasticities, in the range 0.15–0.69, while Mazzanti and Zoboli ( 2005 ), in a study of a group of European countries, find evidence of neither absolute nor relative delinking. Using European panel data they found no WKC evidence for either municipal waste or packaging waste respectively for 1995–2000 and 1997–2000. Estimated elasticities of waste generation with respect to household consumption were close to unity.

Table 1 Literature survey on waste-related studies 0 2 8 Author(s),

Turning point (per (publication year)

Countries/

Time period

Waste typology

Panel/cross-

EKC evidence

capita) Anderson et al. 2007

geographical focus

section data

EU 15 and EU 10 Past years before

Waste and

Model allowing for

Waste generation is

countries

material flows

trendwise changing

linked to economic

coefficients

activities by non-constant trend ratios, Waste Kuznets Curve reasoning

Beede and Bloom 1995

Thirty-six countries

Various years

Solid waste

Cross-section and

MSW generation is

(several case

associated, but inelastic, with respect to per capita income

Berrens et al. 1998

US, county level

Hazardous waste

Cross-section

Negative elasticity;

20,253$ and 17,679$

inverted-U shape

counties)

Cole et al. 1997

OECD countries

Municipal waste

Panel

No evidence of an

No TP

inverted U-shape

Fischer-Kowalski

Absolute decoupling – and Amann

OECD countries

Landfilled waste and

Panel

for landfilled waste but 2001

waste generated

not for waste gener- ated

Johnstone and

– Labonne 2004

OECD countries

Municipal solid waste

Panel

Positive elasticities,

but lower than 1

Karousakis 2009

OECD countries

1980–2000 (4 years)

Municipal solid waste

Panel

MSW increases

– M monotonically with

income, with an

az

n za Mazzanti and Zoboli

elasticity around

ti, 2005 , Mazzanti 2008

European countries

Municipal and

Panel

Neither absolute nor No TP . R

packaging

relative delinking

(elasticity close to o Z

waste

o b li

unity)

Municipal Table 1 continued

Author(s),

Turning point (per (publication year)

Countries/

Time period

Waste typology

Panel/cross-

EKC evidence

geographical focus

section data

capita) W aste

Mazzanti et al. 2008

Relative delinking;

24,000–27,000 e

weak sign of

uznets K

absolute delinking (some richer provinces)

Curv Raymond 2004

es Seppala et al. 2001

International data

Waste/consump-

Cross-section

Evidence for EKC

tion indicator

No evidence of delinking No TP Wang et al. 1998

Five industrialised countries

Direct material flows

Panel

US, county level

Hazardous waste

Cross-section

Evidence for EKC

23,000 (1990 US$)

210 M. Mazzanti, R. Zoboli

Few WKC studies incorporate waste policy analyses. Karousakis ( 2009 ) deals with policy evaluation, and presents evidence on the determinants of waste generation and the driving forces behind the proportions of paper/glass recycled, and the proportion of waste going to landfill. The data are for a panel of 30 OECD countries and the results show that MSW increases monotonically with income, and that urbanisation exerts an even stronger effect on waste generation, while the time-invariant policy index is not significant.

Other studies have investigated policy actions, but at the level of single countries, exploit- ing rich regional data, which however allows moderate generalisation of results. Mazzanti et al. ( 2008 ) find some WKC evidence and signs that waste management instruments do have an effect on reducing waste generation in Italy, where trends are affected by economic, policy

and structural geographical differences. 5 There are also some studies on specific evaluations of the Landfill Directive—the main driver of regulatory actions in the EU, and the UK landfill tax—introduced in 1996 and one of the few cases of a real environmental tax based on eval- uation of the marginal external costs. They include, among others, work by Martin and Scott ( 2003 ), which stresses that tax aimed at reducing landfilling of waste in favour of recovery, recycling, re-use and waste minimisation has failed to significantly change the behaviour of domestic waste producers.

For countries outside the EU, analyses are rare. Taseli ( 2007 ) recently presented an assess- ment of the EU Landfill Directive on Turkey, a potential EU member that is comparable to some of the eastern European country new member states. This study highlights the immense difficulties in such countries of achieving even long run targets. Outside Europe, there are numerous studies on landfill diversion and waste generation in Far Eastern countries where land values are especially high and population densities are peaking at world levels ( Lang 2005 ; Ozawa 2005 ; Yang and Innes 2007 ). Population density and waste policies emerge as complementary drivers of stronger delinking of both waste generation and landfilling, through more active recycling.

The literature on waste drivers and WKC, therefore, underlines that waste generation still generally tend to increase with income or other economic drivers, such as population, and that, in general, an inverted U-shape curve is not in line with the data. Some authors have suggested that for stock pollution externalities, e.g. waste, the pollution-income relation- ship difficulty shows a WKC shaped curve, with pollution stocks monotonically rising with income ( Lieb 2004 ).

A decreasing trend (negative elasticity) for land filling is found in industrialised countries with more developed waste management and waste policies. Nevertheless, there is a risk that WKC trends (absolute delinking) are associated with a few rich countries or areas, and can divide countries in terms of waste performance indicators. Another structural reason for the lack of evidence on waste might be that the change in the sign of income elasticity of the environment/income function will occur at relatively lower income levels for pollutants whose production and consumption can be easily spatially separated, e.g. by exporting the associated pollution or relocating the activities. This will likely be more difficult for waste flows.

The literature survey highlights the need for and value of in-depth investigations on driv- ers of waste policy and its effectiveness: what is required is WKC analyses combined with studies of policy effectiveness and an extensive evaluation of waste drivers. The general added value of delinking analysis is not (only) to show whether economic drivers produce decoupling effects, but more especially, to assess whether and to what extent, there are addi- tional factors that influence this core relationship, and increase the explanatory power of the

5 See the collection of WKC policy oriented works in Mazzanti and Montini ( 2009 ).

Municipal Waste Kuznets Curves 211

model proposed. Our work on the EU allows some generalisation since it focuses on a large regional area that is relatively homogenous and has common waste policy references in the form of EU Directives. Most existing studies analyse circumscribed local situations, with a focus on specific waste streams and waste management instruments and unit based pricing among other aspects ( Sterner and Bartelings 1999 ; Ferrara and Missios 2005 ).

3 Empirical Analyses

3.1 The Empirical Model and the Data In order to verify the delinking relationships between waste indicators (from generation to

disposal) and the economic, socio-economic/structural 6 and policy drivers, we refer to the established EKC framework.

The main methodological problem for applied analysis in this delinking-related frame- work is how to specify the WKC functional relationship. Some authors estimate second order polynomial, others estimate third or even fourth order polynomials, comparing different spec- ifications for relative robustness. N shapes are not relevant here given that the evidence for waste shows that even bell shapes are rare. 7

We test our hypotheses by specifying the proper reduced form usual in the EKC field

( Stern 1998 , 2004 ; Cole 2005 ; Cole et al. 1997 ; Dinda 2005 ) 8 :

log (WI) = β 0i + β t + β 1 log (C) it + β 2 log (C) 2 it + β 3 ( Xi) + β 4 ( Zi) + e 9,10 it (1)

where the first two terms are intercept parameters, which vary across countries and years. X refers to all other structural and socio-economic drivers that are added to the baseline specifi- cation in order to correct for the omission of relevant variables. Z is a vector of policy related

variables (see Tables 2 , 3 for the variables). 11 In order to mitigate collinearity flaws, when highly correlated to each other, the variables included in the vectors X and Z are tested sepa- rately. The error term has the usual properties in panel settings, although obviously differing between fixed effects (FE) and Random Effects Model (REM) settings.

6 We define these factors as structural since, with respect to waste trends, they are a set of exogenous potential drivers that are influenced by the historical, institutional and cultural development of the country and also

are relevant for waste management and disposal based on idiosyncratic geographical aspects (e.g. population density).

7 Cubic specifications for waste are not conceptually or empirically relevant and, as would be expected, are not significant.

8 The panel data model presented here refers to a baseline fixed effects model (FEM) with FE and time dummies.

9 Waste Indicator: MSW—GEN, MSW—INC, MSW—LAND—see Table 2 . 10 All the analyses specify consumption as the main economic driver, since it is the most plausible in waste dynamics given its strict link to MSW ( Mazzanti and Zoboli 2005 ). However, the results do not change if we

use GDP. We took household expenditure (consumption) per capita as the main economic driver, based on the hypothesis that consumption is a better independent variable for waste collection and disposal ( Rothman 1998 ; Jacobsen et al. 2004 ; Gawande et al. 2001 ).

11 The model is based on a framework derived from the EKC literature. All variables are specified in loga- rithmic form using per capita values, to provide elasticity values and to smooth the data. Except where it is

not feasible, logarithmic transformations are used for all covariates.

Table 2 Descriptive statistics and a summary of research hypotheses 1 2 2

Notes and research hypotheses Dependent variables

MSW Collected/generated

Descriptive stats are calculated (kg per capita)

MSW–GEN

for EU 25 over 1995–2005 MSW Landfilled (kg per

MSW–LAND

capita) MSW Incinerated (kg per

73.47 MSW–INC

capita) Independent variables

1. Economic drivers Hypothesised correlation a Final consumption expenditure of

households (euro Per inhabitant—at 1995 prices and exchange rates)

Eventual inverted U + I

Gross domestic expenditure on

0.19 4.25 1.37 RD

R&D (% of GDP)

2. Structural and socio-economic variables

Population density (inhabitants/km 2 )

Urban population (% of total) b 50.60 97.20 71.36 URBPOP

?I

. M az

za n ti, R . Z

o b o li

Municipal Table 2 Continued Household size

− G W Single households (%)

1.9 3.4 2.62 SIZE

+ G aste Age index or ‘elderly ratio’ (population 60 and over to population 20–59 years)

10.12 38.30 25.04 SINGLE

?G K uznets Share of manufacturing value added

OLDNESS

− G 3. Policy variables

9.10 36.30 18.54 VAMAN

Curv Decentralised waste management policy drivers (dummy)

0 1 0.24 DECPOLIND ?G,L,I es Incineration directive (dummy: years/country in which Directive is transposed)

0 1 0.24 INCDIR

Landfill directive (dummy: years/country in which directive is ratified)

− G Waste strategy policy index (range 0–1)

0 1 0.27 LANDIR

+ I Landfill strategy policy index

0.00 0.95 0.34 POLIND

0.00 0.25 0.09 LANDPOLIND − L All values in non log format a

The sign on the hypothesised correlation is shown, as well as the level at which this is most relevant (G for generation, L for land-filling, R for recycling, I for incineration). The element (?) means that the hypothesis is ambiguous either because opposing forces may be influencing the link or because economic theory and other scientific fields do not b provide clear insights Given high correlation, population density and urban population are used alternatively in estimations

214 M. Mazzanti, R. Zoboli

Table 3 Main figures (1995–2005 values) Countries

MSW–LAND MSW–INC Austria

C MSW–GEN

599.6–653.1 0–0 Czech Republic

136.1–23.3 148.6–242.1 United Kingdom

414–374.9 45–48.9 Source Eurostat

In order to test the hypotheses, we exploit information on waste collected and waste land- filled and incinerated in a group of European countries for 1995–2005 (Eurostat sources). 12 The standard WKC specification includes two groups of variables—socio-economic/struc- tural variables and policy indexes—to control for inter-country heterogeneity. The first group controls for the socio-economics factors that might differ between countries, such as pop- ulation density, urban population, household size, share of manufacturing in the economy based on data mainly from EUROSTAT structural indicators datasets. Variables for policy

indexes are constructed based on the country fact sheets available at EIONET, 13 and public information on the ratification of the EU Landfill and Incineration Directives.

12 Table 3 provides a summary of country heterogeneity for the main variables. 13 EIONET is the information network of the European Environment Agency (EEA) and its member countries, and collects and disseminates data and information on the European environment.

Municipal Waste Kuznets Curves 215 Below we provide specifications for the three waste indicators investigated, 14 highlighting

which factors and hypotheses are linked to the three different levels of the waste chain, from generation to disposal. A concise discussion of the main hypotheses and the variables used

in specific regressions is provided for each level (see also Table 2 ).

3.1.1 Waste Generation This level of analysis provides direct evidence regarding the usual WKC hypothesis in terms

of the waste generation-income relationship. Waste generation reduction (i.e. ‘waste preven- tion’) is the ultimate objective of any social policy targeted towards waste flows, although for the most part explicit policy actions (targets for waste generated per capita) do not exist. Although waste prevention is at the top of the waste hierarchy in the EU, there are no directives so far that include actions oriented specifically towards waste prevention. Waste management (separated collection, recovery/recycling) and landfill diversion are the focus of existing waste policies, probably because of their presumed relatively lower implementation and compliance costs.

Let us analyse the various research hypotheses we may formulate for waste generation. Regarding economic drivers, a WKC oriented structure of the model allows the estimation of an eventual TP for waste generation. This is generally not observed. The TP hints at the GDP/consumption level beyond which the relationship (in this case, between waste produc- tion and income) turns negative. Existing econometric and descriptive empirical evidence shows that only a relative delinking is observed. We aim to provide new evidence based on official EU data.

We also test various hypotheses on the possible effects exerted by socio-economic and

structural variables, which represent a diverse set of factors (see Table 2 ). Population density (or urban population) 15 is likely to have a positive impact on waste generation. In more densely populated areas, only economies of scale spurred by urbanisation could invert the trend and reduce generation. Household features may matter at this level. In fact, we expect that the larger the size of the household, the less waste will be generated per capita. Nevertheless, even a positive link could be plausible if waste management at the domestic level (composting) is poorly developed on average. Thus, more single person households should increase waste generation.

We use age as a control variable. From a socio-economic point of view, there may be opposite forces at play: if, on the one hand, older people produce less waste than younger residents, it is also true that older people may be less accustomed or committed to collection and recycling of waste. On the other hand, the opportunity cost of time is lower for older people, and waste collection/recycling efforts require time. The sign of the relationship, then, is unpredictable. Interaction with data on education level would be interesting, but this would

be the subject for micro-based studies, which would probably test these other factors more robustly. We use them here (given their availability) as controls to mitigate for problems of omitting relevant variables. Some of these socio-economic variables capture more than one of the structural/institutional elements characterising a country.

We also include in our analysis two types of policy proxies. The first is related to the European Landfill and Incineration Directives and their implementation in member states.

14 Waste generation, incineration and landfill disposal. Waste generation is measured by waste collection, the observable factor. We do not present results for recycling (although they are available) because recycling is

only calculated as a residual (MSW generation—MSW landfilled—MSW incinerated).

15 Given their high correlation, these two variables can be used alternatively in the econometric exercises.

216 M. Mazzanti, R. Zoboli

Table 4 MSW generation regression (EU 25 )

Model FEM FEM

– − 0.022 Hausman test

0.005 0.006 0.009 prob. > chi sq N

264 264 (–) means not included. Significance at 90%, 95% and 99% is denoted by *, ** and ***; TP (e, consumption

per capita); F test show overall significance for all regressions at 1% significance; R squared present reason- ably high value for panel settings. All variables are in logarithmic form when possible. The constant term in REM and the overall constant/individual fixed effects in FEM are not shown

These proxies are built as dummy variables that take the value 1 in a given year between 1995 and 2005 if a country has transposed these Directives into its national law. We expect implementation to be positively correlated with delinking performance.

The second group of policy indexes is more country specific. We first exploit a ‘decen- tralised waste management index’ that reflects the degree of waste policy decentralisation across countries. 16 Decentralisation may positively affect waste generation, via prevention and better waste management performance, because of higher flexibility and specificity in policy implementation, which may account for local idiosyncratic cost and benefit elements

related to policy ( Pearce 2004 ). 17 Although decentralisation may improve policy implemen- tation in the EU, including policies for waste prevention, it may have some drawbacks in terms of exploitation of local rents by public and private agents. In principle, rents are neither good nor bad in the environmental realm. What matters is their effects on static and dynamic elements such as value creation and innovation. Waste ‘markets’, such as land-filling and even recycling, may be associated with rents that could lock in a local system to less than optimal equilibrium. This aspect required further research. Watson et al. ( 2008 ) discuss some policy decentralisation and environmental policy integration issues, focusing on waste management and governance in the UK.

Finally, we also include an environmental policy index. This is a proxy for national policies implemented over the time period examined. It captures all possible information regarding national implementation of waste related policies (MSW, biodegradable solid waste, pack- aging waste, end of life vehicles, other). We used the country studies available on EIONET

16 This discrete index variable captures the extent to which a country is decentralised in (waste) policies, and more generally is structured as a federal state. Actually, 4 countries are associated to the value 1: Italy,

Germany, Austria, Spain, 2 have the value 0.5 (UK and Belgium), all others have the value 0. 17 Fredriksson ( 2000 ) studies the pros and cons of decentralisation versus centralised management options

regarding the siting of waste facilities.

Municipal Waste Kuznets Curves 217

as our information source. The index is both very comprehensive with regard to Landfill Directive related variables, 18 and also may capture some of the waste prevention features of national policies. 19 It should be noted that, contrary to the decentralised policy index, all other proxies vary across countries and over time. Introducing policy proxies is crucial in the waste arena and could constitute a contribution of our work. Their role is very relevant because many European policies have been enacted quite recently, and their inclusion in a WKC framework could produce a sort of ex-post effectiveness evaluation. Both structural indicators and policy variables could be important drivers of WKC shapes; their omission could overestimate the ‘pure’ economic effect.

The specification we test in this case then is:

log (MSW−GEN) = β 0i + β t + β 1 log (C)

it + β 2 log (C) it + β 3 ( Xi) it

(2) where X denotes the above discussed socio-economic/structural factors (DENS or URBPOP,

+ β 4 ( Zi) it + e it

SIZE, SINGLE, OLDNESS, VAMAN) and Z the policy levers (DECPOLIND, INCDIR, LANDIR, POLIND). For a summary of the covariates related to the level of waste genera-

tion see Tables 2 , 3 .

3.1.2 Landfill Diversion The second level of the empirical model focuses on the disposal stage in the waste chain, i.e.

landfilling. The economic driver is hypothesised to have an impact on landfilled waste and to show a bell shape. In fact, although some countries are continuing to increase their share of landfill and there is wide heterogeneity across Europe, on average, shares of landfilled waste and landfilled waste per capita have been constantly decreasing over the last decade. We can also expect to find a negative relationship, not a bell shape. From the viewpoint of the EU average, the period 1995–2005 may already be at the ‘right’ descending side of an inverted U shaped landfill-economic growth relationship. It is possible that testing for N shapes will be more relevant in the future. For example, a major country, such as Italy, following a decade of focusing on landfill diversion, has seen a slight increase in waste landfilled per capita in 2006 with regard 2005.

In terms of socio-economic and structural factors, we assume that population density and urbanisation will be negatively related to landfilled waste. Both the opportunity costs linked to the higher value of land in densely populated and urban areas (value of land, of commer- cial activities crowded out by landfill sites, and other public investments), and the higher externality costs that arise in areas of higher residential occupation should, ceteris paribus, drive down the use of landfill as a disposal option and as a real ‘investment’ opportunity in terms of rents, compared to other market rents ( André and Cerda 2004 ). However, since such

18 Thus, in any given year, each country is associated with an index value, which assigns 1 to the maximum potential value (among all the considered policies). We differentiate between the presence of only strategy

(low value), and an effective regulatory policy (high value). The latter is assigned a stronger weight (0 for no policy, 1 only strategy, 2 policy). Prominent examples of overall environmental policy performance index setups for many countries based on a synthesis of diverse policy performances can be found in Eliste and Fred- rikkson ( 1998 ). Cagatay and Mihci ( 2003 ; 2006 ) provide an index of environmental sensitivity performance for 1990–1995, for acidification, climate change, water and even waste management.

19 We can hypothesise that the backward effects of landfill policies and waste management actions on the MSW generated are not significant. Nevertheless, since our synthetic policy index also captures the variety of

national measures implemented on waste in addition to landfill diversion, some effects may emerge.

218 M. Mazzanti, R. Zoboli

opportunity costs and environmental costs may vary across regions (e.g. EU 15 versus EU 10 ) we need a careful case by case investigation of this hypothesis.

The core factors we test at this level are policy levers. This is a crucial step in our analysis. By exploiting both the binary indexes and the continuous synthetic index, which it should be noted both vary over time and across countries, we can verify the extent to which the policy experiences of European countries in the last decade (primarily the 1999 Landfill Directive), have affected the waste-income relationship. Since the synthetic policy index also captures the general commitment of a country over the period, actions that occurred prior to 1999 and the actual implementation of the Directive, are taken account of in the econometric investigation.

The specification tested then is: log (MSW−LAND) = β 0i + β t + β 1 log (C) it + β 2 log (C) 2 it + β 3 ( Xi) it

(3) where X (DENS or URBPOP) and Z (DECPOLIND, INCDIR, LANDIR, POLIND) refer to

+ β 4 ( Zi) it + e it

the socio-economic and policy variables linked to this stage of the investigation, as summa- rised in Table 2 . 20

3.1.3 Incinerated Waste The WKC-like relationship here assumes a different character. In fact, we can expect that

incineration, at least in the EU 15 (data availability confines our analysis to the EU 15 , given that EU countries show incineration level close to zero), is positively related to economic growth over recent years. This is evident from the descriptive analysis and is obviously linked to the trend towards landfill diversion. Nevertheless, the non-linearity of the relation- ship could assume an exponential or a bell-shaped dynamics. We would expect the latter as being more plausible given that incineration is associated, to some extent, with dimin- ishing marginal returns and increasing external costs. The interpretation is different from the usual EKC hypothesis for waste generation and landfilling. It is related to the fact that our analysis includes other options, such as recycling and composting, which are treated as ‘residuals’ here. The increasing role of these options, common to most countries, should lead to and make compatible scenarios of landfill diversion and increasing, but not exponentially, incineration, independent of the delinking dynamics of waste generation at source. As many quantitative applied studies show, in the waste arena corner solutions are seldom (never) optimal strategies, given the increasing marginal costs—financial and social—of landfilling, incineration and recycling.

Linked to the above comments, we assume here that we cannot formulate a clear hypoth- esis on the role of density. The resulting sign is ambiguous depending on the alternative to landfill disposal, if any, that is chosen. If we assume that densely populated areas move away from landfilling, the opportunity costs and environmental costs of incineration could represent a relatively better scenario. If, instead, the country moves away from landfilling primarily to recycling and incineration then population density/degree of urbanisation may

be negatively linked.

20 For landfill diversion and incineration we test two stage regressions, including the predicted values of waste generation as a covariate in specifications (3) and (4), replacing the economic driver. This accounts for the

unchained and consequential nature of waste flows from generation to disposal. The results do not change; predicted significance values reflect the significance of consumption in the model (2), with elasticities around unity or even higher.

Municipal Waste Kuznets Curves 219

We also test R&D intensity, exploited as a country specific effect. Although some outliers may exist, i.e. countries with high R&D/GDP shares and low incineration levels, eventually because of their more determined pursuit of recycling strategies, on average we would expect

a dynamic, strong correlation between a country’s technological intensity and its incineration level. The example of Germany may be a paradigmatic case in Europe. We test whether this anecdotal evidence may represent a general statistical regularity. Given the potential corre- lation with GDP, R&D is tested as an alternative to consumption. Eventual non-linearity is tested to account for the diminishing marginal effects of R&D on incineration technology strategies.

The reasoning related to policy levers is similar, but with the opposite expected sign, as for landfill diversion. At this stage we can expect a robust specific link between implementation of the Incineration Directive and incinerated waste dynamics.

The specification for incinerated waste then is: log (MSW−INC) = β

0i + β t + β 1 log (C) it + β 2 log (C) it + β 3 ( Xi) it + β 4 ( Zi) it + e it (4a)

or log (MSW−INC) = β 0i + β t + β 1 log (R&D/GDP) it + β 2 log (R&D/GDP) 2 it

(4b) where X (DENS or URBPOP) and Z (DECPOLIND, INCDIR, LANDIR, POLIND) refer to

+ β 3 ( Xi) it + β 4 ( Zi) it + eit

the socio-economic and policy variables.

3.2 Methodological Issues and Estimation Procedures We present the regression results for EU 25 , given their higher statistical robustness and gen-

erality. Specific outcomes for the sub samples (EU 15 and EU 10 ) are commented on. Because of the large number of cases involved, Table 4 reports only the EU 25 regressions. All analyses are available upon request.

The analyses are conducted in three steps: (i) a traditional WKC specification, including only the main economic driver; (ii) the specification with the structural and socio-economic drivers relevant at each level of the waste chain; (iii) the regression including the economic driver, the significant structural factors and the different policy indexes.

For all levels, the variables included in vectors X and Z are inserted separately to mitigate collinearity flaws. Only the more robust results are shown. The exception is for incineration, which is 0 for most east European countries; we therefore focus on the EU 15 .

All regressions are estimated first by FEM and REM, and the best specification is selected following the Hausman test. 21

21 REM is generally more appropriate for datasets derived from larger statistical populations, while the FEM is preferred when there are data for all population units. The choice about which to use is often based on

these assumptions without implementing the Hausman test. In this case we decided to perform the Hausman test and follow its results, which largely provide support for the FEM, a specification that is obviously not used in presence of time invariant covariates. A key consideration in our choice was whether the (unobserved) individual effects and the vector of covariates are correlated. The test, under the assumption of strict exoge- neity for βs in both the null and the alternative, compares an inefficient but consistent model (FEM, which is costly in terms of loss of degrees of freedom when N grows large), and an efficient model, REM. The latter is by assumption more efficient given homoskedasticity regarding unobserved effects ( Wooldridge 2003 ). If

the null hypothesis (difference in coefficients B FEM − B REM is not systematic between the two models) is not rejected, then the REM provides similar and more efficient estimates compared to FEM, i.e. between GLS

220 M. Mazzanti, R. Zoboli

In addition, we test the extent to which random coefficient models and dynamic models

add insights. 22 First, we test the relevance of the Swamy ( 1970 ) random-coefficients lin- ear regression model, and then we focus on two different types of dynamic model, GMM (General Method of Moment) and Least Square dummy variable correction (LSDVC, Bruno 2005 ). This further level of analysis aims to test whether models that take account of differ- ent aspects omitted in the simple FEM and REM analysis (e.g. heterogeneity of slopes and dynamics) confirm our results.

3.2.1 MSW Generation The analyses (Table 4 ) do not show overall WKC evidence for MSW generation. The linear

term shows a significant and positive coefficient for consumption, with an elasticity in the range 0.114–0.23 across specifications. 23 The EU 15 and EU 25 analyses are similar, but with higher elasticity (around 0.70–0.80) in the case of the former. These elasticities are slightly lower than found in previous analyses of Europe using waste data for the 1990s ( Mazzanti

2008 ), and also seem to imply more active delinking for the EU 10 countries. 24 In general, then, our evidence is of a ‘relative delinking’ in the relation. This result is interesting and partially confirms our expectations, but with some positive signs in terms of currently lower elasticities.

The introduction of socio-economic controls does not alter the results. The results pro-

vide evidence of relative delinking, and elasticity equal to 0.72 in the EU 15 sub-sample and lower than 0.20 for EU 25 (see Table 4 ). The most significant and robust control variables are population density (or share of urban population) and share of manufacturing in total economic activities. Share of manufacturing has a negative sign which is in line with our expectations: wealthier and more services oriented economies produce more waste per capita. This questions whether relatively better environmental performance is always associated with services, especially if we focus on pollution-related aspects: composition effects for waste (MSW includes some commercial and business waste), could show counterintuitive results. Population density or urban population in these economies instead impacts positively on waste generation: scale effects prevail over possible economies of scale in waste prevention and waste management/collection activities. All other factors are never significant.

To sum up, socio-economic structural factors provide useful pointers to what drives waste generation although they do not impact on the core relationship. Household related structural

Footnote 21 continued (General Least Squares) and LSDV (Least Square Dummy Variables) estimators. As noted by List and Gallet ( 1999 ), the parameter estimates of FEM and REM converge as Time and cross section observations increase, which also confirms our findings.

22 We estimated the relationship in first differences to take account of possible serial correlations in the data, but the results were similar to those for the FE estimation. A FEM with time FE proved to be not significant

in terms of time dummies and overall robustness. 23 Table 4 presents linear specifications only. The results do not change overall if we run quadratic specifi-

cations, showing no EKC evidence, but with a linear term with a negative coefficient and the quadratic term with a positive one. This is due to the nature of the data, and captures the effect of the low-income countries (e.g. Lithuania, Czech Republic, Slovakia, Poland and Slovenia), which registered a reduction in total MSW collected in the period under study. This slight anomaly generates a downwardly sloped relationship between MSW collected and consumption, in the first phase.

24 This may be interpreted in terms of the (observed even in other contexts) delinking in some east European countries (Hungary among others), which may be driving the general empirical picture. Even the

EU 25 quadratic analysis shows a U shaped curve affected by delinking in low income eastern European coun- tries. Bluffstone and Deshazo ( 2003 ) present a case study of an east European country coping with EU waste management challenges.

Municipal Waste Kuznets Curves 221

variables seem not to have a major influence. Average household behaviour and characteristics are not correlated (either negatively or positively) to the amount of waste generated/collected. This might indicate that policy efforts are having more effect on post-consumer activities, such as recycling and landfilling operations.