Waste Management 34 (2014) 2239–2250

Contents lists available at ScienceDirect

Waste Management journal homepage: www.elsevier.com/locate/wasman

Does industrial waste taxation contribute to reduction of landfilled waste? Dynamic panel analysis considering industrial waste category in Japan Toshiaki Sasao ⇑ Faculty of Humanities and Social Sciences, Iwate University, 3-18-34, Ueda, Morioka 020-8550, Japan

a r t i c l e

i n f o

Article history: Received 14 April 2014 Accepted 15 July 2014 Available online 22 August 2014 Keywords: Dynamic analysis Panel data Industrial waste tax Landfilled waste

a b s t r a c t Waste taxes, such as landfill and incineration taxes, have emerged as a popular option in developed countries to promote the 3Rs (reduce, reuse, and recycle). However, few studies have examined the effectiveness of waste taxes. In addition, quite a few studies have considered both dynamic relationships among dependent variables and unobserved individual heterogeneity among the jurisdictions. If dependent variables are persistent, omitted variables cause a bias, or common characteristics exist across the jurisdictions that have introduced waste taxes, the standard fixed effects model may lead to biased estimation results and misunderstood causal relationships. In addition, most existing studies have examined waste in terms of total amounts rather than by categories. Even if significant reductions in total waste amounts are not observed, some reduction within each category may, nevertheless, become evident. Therefore, this study analyzes the effects of industrial waste taxation on quantities of waste in landfill in Japan by applying the bias-corrected least-squares dummy variable (LSDVC) estimators; the general method of moments (difference GMM); and the system GMM. In addition, the study investigates effect differences attributable to industrial waste categories and taxation types. This paper shows that industrial waste taxes in Japan have minimal, significant effects on the reduction of final disposal amounts thus far, considering dynamic relationships and waste categories. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Waste taxes, such as landfill and incineration taxes, have emerged as a popular option in developed countries to promote the 3Rs (reduce, reuse, and recycle). Most member states of the European Union (EU) have introduced landfill taxes (BIO Intelligence Service, 2012). For example, the United Kingdom has introduced a landfill tax since 1996. The standard rate that applies to all active waste has been raised annually, and the rate amounts to 80 pounds (127 dollars) per ton as of 2014.1 Some countries, such as Demark and France, have gone further to implement incineration taxes. In areas where the central government is not involved in waste taxes, some local governments have introduced such taxes of their own volition. For example, 20 states in the United States have introduced landfill taxes (Kinnaman, 2006). In Italy, landfill taxes

⇑ Tel./fax: +81 19 621 6751. E-mail address: [email protected] The British pound is converted into US dollars assuming an exchange rate of approximately 1 pound to 1.59 dollars (the average rate in 2013). 1

http://dx.doi.org/10.1016/j.wasman.2014.07.014 0956-053X/Ó 2014 Elsevier Ltd. All rights reserved.

have been implemented at the regional level since 1996, and the tax rate varies by region (BIO, 2012; Mazzanti et al., 2012). A similar pattern emerges for Japan’s industrial waste taxes. More than half of Japan’s prefectures enforce industrial waste taxation, which the tax base refers to as transporting waste to intermediate treatment facilities and landfills at the prefecture level, though the Japanese government is not involved in such enforcement. Many economic studies have assessed the effectiveness of unit-based pricing on the curbside collection of municipal solid waste (Kinnaman, 2009). However, few studies have examined the effectiveness of landfill taxes or incineration taxes. In addition, quite a few studies have considered both dynamic relationships among dependent variables and unobserved individual heterogeneity among the jurisdictions. If dependent variables are persistent, omitted variables cause a bias, or common characteristics exist across the jurisdictions that have introduced waste taxes, the standard fixed effects model, which is a popular estimation in panel data analysis, may lead to biased estimation results and misunderstood causal relationships. In addition, most existing studies have examined waste in terms of total amounts rather than by categories. Even if significant reductions in total waste amounts

2240

T. Sasao / Waste Management 34 (2014) 2239–2250

Table 1 Trend of enforcement of industrial waste taxation and total amounts of industrial waste generated and disposed of in landfills in Japan. Source: Ministry of the Environment and Ministry for Internal Affairs and Communications. 2003FY

2004FY

2005FY

2006FY

2007FY

2008FY

2009FY

Number of prefectures enforcing Industrial waste taxation Type A 0 0 1 Type B 0 0 0 Type C 0 0 0 Total 0 0 1

2000FY

2001FY

2002FY

1 3 0 4

2* 9* 0 11*

2 13 6 21

2 18 6 26

2 19 6 27

2 19 6 27

2 19 6 27

Total amount of industrial waste generated (thousand tons) 406,037 400,243 393,234

411,623

417,156

421,677

418,497

419,425

403,661

389,746

25,827

24,229

21,799

20,144

16,701

13,591

Total amount of industrial waste disposed in landfills (thousand tons) 45,000# 42,000# 39,561 30,440 * #

Although industrial waste taxes in a part of prefectures were enforced in January 2004, they are counted in 2004FY in accordance with this analysis method noted later. Approximation.

are not observed, some reduction within each category may, nevertheless, become evident. Therefore, this study analyzes the effects of industrial waste taxation on quantities of waste in landfill in Japan by applying the bias-corrected least-squares dummy variable (LSDVC) estimators (Kiviet, 1995); the general method of moments (difference GMM) (Arellano and Bond, 1991); and the system GMM (Arellano and Bover, 1995; Blundell and Bond, 1998). In addition, the study investigates effect differences attributable to industrial waste categories and taxation types.2 Industrial waste is generated as a byproduct by economic activity. In Japan, most of the generated waste passes through intermediate disposal sites such as incinerators before it is brought to landfills. The rate of landfilled waste relative to the total amount of industrial waste generated has gradually decreased, and was about only 3.5% in 2009. However, the total amount of industrial waste disposed of in landfills is still nearly 10 million tons (Table 1), and therefore it is not negligible. This paper proceeds as follows. Section 2 outlines industrial waste taxes in Japan. Section 3 provides an overview of existing relevant studies on waste taxes. Section 4 explains the estimation methods and data used in the analysis. Section 5 presents the estimation results, and Section 6 offers concluding remarks. 2. Industrial waste taxation in Japan Currently, 27 out of Japan’s 47 prefectures enforce industrial waste taxes following their introduction in the Mie Prefecture in 2002. Industrial waste taxation is an objective tax that includes two elements: raising financial resources and promoting incentives for adhering to the 3Rs principle, with proper industrial waste disposal methods. The three types of taxation employed at the prefecture level are described below. Type A taxation involves declaration payments made by waste generators. Waste generators are taxed on their direct transport of waste to landfills and to intermediate treatment facilities except recycling facilities. This type of industrial waste taxation is based on declarations by the waste generators. Two prefectures use this type of taxation by imposing a tax of 1000 yen (10 dollars) per ton of waste transported to landfills.3 This tax rate is common among the three types of industrial waste taxation in Japan. Unique to Type A, however, a specific coefficient is multiplied for waste transported to intermediate treatment facilities, considering the environmental impact of final disposal. For example, an additional tax rate of 100 yen (1 dollar) per ton is imposed on transporting 2 In Japan, industrial waste includes both hazardous and nonhazardous waste, particularly, ash, sludge, waste oil, waste acid, waste alkali, waste plastics, and others, identified by a cabinet order among all the wastes left as a result of business activity. 3 The Japanese yen is converted into US dollars assuming an exchange rate of approximately 100 yen to 1.02 dollars (the average rate in 2013).

the waste to incinerators. Type A taxation is said to generate the strongest incentive for waste reduction among the three types of taxation because waste generators are taxed on direct waste transport to both landfills and intermediate treatment facilities except recycling facilities (Morotomi, 2003). Type B taxation is a special levy on contractors involved in the final disposal of waste. This tax is similar to the landfill tax in the United Kingdom and other countries, and it is now the most popular type of industrial waste taxation in Japan. In this type of taxation, waste generators, including intermediate treatment facilities, are taxed on the amount of waste transported to landfills. This type of taxation has two major characteristics: It is based on transporting waste only to landfills, and final disposal contractors must levy the tax on waste generators on behalf of a prefecture’s government. All prefectures that have introduced this type of industrial waste taxation have applied a tax of 1000 yen (10 dollars) per ton.4 Type B taxation costs the least among the three types of taxation because the tax is only applied to waste transported to landfills. However, Type B is widely considered to provide weaker incentives for waste reduction than Types A and C, although we can still expect a reduction of waste sent to landfills (Morotomi, 2003). Type C is also a special levy on intermediate disposal contractors with incinerators and final disposal contractors. This type of industrial waste taxation, implemented in six prefectures on Kyushu Island, exhibits two major characteristics. First, waste generators are taxed on transporting waste not only to landfills but also to incinerators. Second, intermediate disposal contractors with incinerators are also taxed on transporting waste (e.g., post-incineration ashes) to landfills. All prefectures that have introduced this type of taxation have placed taxes of 1000 yen (10 dollars) per ton on transporting waste to landfills and 800 yen (8 dollars) per ton on transporting waste to incinerators. It is anticipated that this type of tax will generate incentives to reduce the amount of waste stronger than those of Type B, but weaker than those of Type A. Table 1 shows that the number of prefectures enforcing industrial waste taxation increased for several years prior to 2007, but has remained steady thereafter. Table 1 also shows trends in the total amounts of industrial waste generated and disposed of in Japan’s landfills over the past decade. As indicated, in this time frame, the total amount of industrial waste generated in Japan has remained almost constant, while the total amount of industrial waste disposed in landfills has gradually decreased. Given this pattern, this paper analyzes whether industrial waste taxation contributes to the observed reduction of landfilled waste in Japan.

4 Some prefectures introduced industrial waste taxation at a lower level, and followed by a gradual increase to 1000 yen per ton.

2241

T. Sasao / Waste Management 34 (2014) 2239–2250

Some prefectures apply industrial waste tax reductions or exemptions to specific types of waste or size of generators; for example, waste transported to a generator’s own landfill. Moreover, the two prefectures that have adopted Type A taxation have exempted smaller generators to reduce the cost of tax collection. Furthermore, prefectures that have adopted Type C taxation have exempted waste recycled in cement kilns or used for thermal recovery.

3. Literature review This section reviews the existing literature that examines actual waste management data and compares data before and after taxation. Sigman (1996) used panel data for local waste taxes in the United States to examine the impact of state taxes on levels of chlorinated solvent waste from metal cleaning. The result showed that the estimated overall impact of state hazardous waste disposal taxes currently in place was small, although the amount of solvent waste generated may respond elastically to changes in incineration costs. Bartelings et al. (2005) used panel data to examine the impact of a landfill tax in the Netherlands. They showed that the landfill tax contributed to an increase in the service sector’s share of recycling, although it did not affect the amount of waste generated. Mazzanti et al. (2012) analyzed the relationship between income and generated or landfilled municipal solid waste (MSW). They also analyzed endogenous changes in waste management and the spatiality of policy implementation (including landfill taxes) at the provincial level in Italy. The authors showed that the landfill tax has not significantly affected generated and landfilled waste per capita. In addition, they estimated a regression that the dependent variable is the variation in the growth of the log dependent variable against the lagged variable. They indicated the presence of convergence in both waste generation and landfill diversion trends. Nicolli and Mazzanti (2013) used panel data to examine the effectiveness of Italian landfill taxes in terms of landfill diversion. They showed that the tax did contribute to landfill diversion in Italy. Sasao (2014c) used panel data to analyze the effect of industrial waste taxation on waste generation and on final disposal in Japan. He showed that although such taxes exert a significant effect by reducing the amount of final waste disposed, the reduction effects to date have been offset by the time trend that is not directly observable. From a different perspective, other studies have examined the relationship between the introduction of waste regulation (including waste taxes) and waste shipments (or even illegal dumping) from an economic viewpoint. Levinson (1999a, 1999b) links the theoretical and empirical literature on interjurisdictional tax and regulatory competition, focusing on state hazardous waste disposal taxes in the United States. The study theoretically showed that ‘‘not in my backyard’’ (NIMBY) taxes in that states are incentived to set inefficiently high rates for imported waste. Moreover, the study used panel data to empirically show that the taxes do make a difference as they significantly decrease shipments of waste to high-tax states. Ley et al. (2002) evaluated the potential economic effects of public policies proposed to restrict flows of the MSW. They showed that restricting interstate trade with import surcharges or volume-based constraints reduces the aggregate surplus, although the producer surplus can increase. Sasao (2014a, 2014b) showed that industrial waste taxation in neighboring prefectures was a significant factor behind increased waste inflows at the prefecture level in Japan. He also showed that Type B taxation, noted in the previous section, could result in increased outflows to other prefectures for intermediate disposal, although industrial waste taxation did not restrict waste inflows for either intermediate or final disposal. In addition, Sasao (2014a) indicated

that prohibitions on imports, the most stringent regulation on waste inflow, could actually increase illegal dumping. D’Amato et al. (2014) investigated how illegal disposal of waste is affected by the decentralized waste management of local governments and by enforcement policies in Italy. They suggested that more diffuse incentive-based waste policies do increase illegal disposals. As noted above, some studies have used panel data to examine the effects of waste taxation on waste flows. However, quite a few studies have considered possible endogeneity. Levinson (1999a) is a pioneer study that addresses possible endogeneity of waste tax rates. He uses a fixed effects model to control for the unobserved time-invariant intrinsic effect. However, the fixed effects model cannot consider possible reverse causality, though it can control omitted variable bias. He also uses a two-stage instrumental variable (IV) estimator to control endogenetity. However, he does not consider dynamic relationships among dependent variables. Nicolli and Mazzanti (2013) is a rare study considering possible endogeneity and dynamic relationships among dependent variables by means of a two-stage IV estimator. However, as Baltagi (2008) noted, the IV estimation method may lead to consistent but not necessarily efficient estimates of a model’s parameters, since it does not use all available moment conditions and does not consider the impact of structural differences upon residual disturbances. 4. Methodology 4.1. Estimation methods When we examine the effects of waste taxes on waste flow in detail, we should consider some econometric problems, or possible endogeneity in a wide sense (Mazzanti et al., 2012; Mileva, 2007): First, is possible reverse causality, that is endogeneity in a narrow sense. If common characteristics exist across the jurisdictions that have introduced waste taxes, causal relationships may be misunderstood. For example, the prefectures that have a large or small amount of landfilled waste originally may tend to introduce waste taxes. Second, is a fixed effects. If there are unobserved time-invariant prefecture characteristics, or omitted variables such as average disposal fees, they may be correlated with the explanatory variables. Third, is the presence of the lagged dependent variable resulting in autocorrelation. For example, the prefectures where many waste disposal facilities are located, with a large amount of landfilled waste originally, may be inclined to accept more industrial waste, even waste originated from outside those prefectures. This paper analyzes the effects of industrial waste taxation in Japan, considering these three econometric problems. In this study, panel data was collected at the prefecture level for Japan’s 47 prefectures from FY 2000 to FY 2009 to analyze the effect of local waste taxation on levels of landfilled industrial waste in Japan. The study has a short time dimension (T = 10) and a larger prefecture dimension (N = 47). Considering the possible econometrical problems noted above, the model specification is as follows:

lnðFDt Þ ¼ a þ b1 lnðFDt1 Þ þ b2 ln



 GPPt þ b3 ðTAX t Þ þ ci GPPt1

þ YEARt þ eit which is equivalent to the following equation5:

ln



FDt FDt1



¼ a þ ðb1  1Þ lnðFDt1 Þ þ b2 ln



GPP t GPPt1

 þ b3 ðTAX t Þ

þ ci þ YEARt þ eit 5

I referred to Mazzanti et al. (2012) and Presbitero (2005) for this transformation.

2242

T. Sasao / Waste Management 34 (2014) 2239–2250

In the above equation, FD is the amount of final disposal, namely landfilled waste; GPP is the gross prefectural product or output of each industry noted above; and TAX is the rate of industrial waste taxation. Furthermore, ci is the fixed effect of each prefecture; YEARt is the fixed effect of each year; and eit is the error term that includes all unobserved influences. A dependent variable with a one year lag is included as an explanatory variable to control the lagged dependent variable. All models in this paper are dynamic models. The explanatory variables are divided into two categories: economic determinants and type of industrial waste taxation. Since both amounts of landfilled industrial waste and economic outputs vary widely among prefectures, capturing accurate fluctuations is a challenge. Therefore, dependent variables and economic variables are considered in terms of their growth rates rather than their absolute amounts. However, the lagged dependent variables are not transformed to the growth rate because the past absolute amount rather than past growth rate of landfilled waste appears to affect on the present growth rate of landfilled waste. This is similar to the equation estimated in Mazzanti et al. (2012) to examine convergence in waste generation and landfill diversion trends. This study uses a Dynamic Ordinary Least Squares estimation (DOLS) and a dynamic Least Squares Dummy Variables estimation (LSDV) or standard fixed effects estimation as benchmark. In addition, the study applies the LSDVC estimator; difference GMM (Arellano and Bond, 1991); and system GMM (Arellano and Bover, 1995; Blundell and Bond, 1998) to consider the possible econometric problems. As noted by Bond (2002), the OLS levels estimator is biased upwards while the LSDV estimator is biased downwards, at least in large samples. Therefore, more reliable estimators would be located between the two estimators. Kiviet (1995) removed the bias from the LSDV estimator, proving that the LSDVC was more efficient than various instrumental variable type estimators such as IV and GMM. Judson and Owen (1999) showed that the LSDVC estimator is the best choice for a balanced panel while the GMM estimator is the best choice for an unbalanced panel when the time period is less than 20 using a Monte Carlo simulation. Bruno (2005) also showed that the LSDVC often outperformed the IV-GMM estimators in terms of bias and root mean squared error when the number of individuals is small. This paper applies the LSDVC estimator because the panel data in the study is a balanced panel. However, it should be noted that the LSDVC estimator is only applicable in the presence of strict exogeneity. On the other hand, Roodman (2009) summarizes the characteristics of the GMM approaches in general to six points: (1) ‘‘small T, large N’’ panels, implying few periods and many individuals; (2) a linear functional relationship; (3) one left-hand-side variable that is dynamic, depending on its own past realizations; (4) independent variables that are not strictly exogenous, that is, they are correlated with past and possibly current realizations of the error; (5) fixed individual effects; and (6) heteroskedasticity and autocorrelation within groups but not across them. While the standard GMM estimation begins by transforming all regressors, usually by differencing, and uses the difference GMM, the system GMM estimation makes an additional assumption that the instrument variables’ first differences are uncorrelated with the fixed effects, then builds a system of two equations, namely, the original and the transformed equation. Consequently, the system GMM allows for the introduction of more instruments and can dramatically improve efficiency (Baltagi, 2008; Roodman, 2009). For both, the difference GMM and system GMM estimators, we can select a one-step estimator or two-step estimator. The onestep estimator assumes that the weighting matrix is known. If the assumption is satisfied, it is efficient under the restrictive assumptions of homoscedasticity and not a correlation of the error

terms. On the other hand, the two-step estimator uses a first-step estimation to obtain the optimal weighting matrix at the second step. The standard covariance matrix by the two-step estimator is robust to panel-specific autocorrelation and heteroskedasticity, but the standard errors are downward biased (Mileva, 2007). Therefore, a better estimate of the standard errors are proposed by Windmeijer (2005). This study applies the both estimators for the difference GMM and system GMM. Both, the difference GMM and system GMM estimators, assume that a first-order serial correlation exists, but that a second-order serial correlation does not exist in the residuals. Therefore, we have to conduct autocorrelation tests (AR tests). We need to reject the null hypothesis that neither a first-order nor a second-order serial correlation exists in the first-order serial correlation, but not to reject it in the second-order. In addition, both the difference GMM and system GMM require the Hansen J-statistic tests to check the instruments’ validity. The null hypothesis is that overidentifying restrictions are valid and that the model specification is correct. Here, the rejection of the null hypothesis implies that either or both assumptions are incorrect. The study examines models that treat tax variables as endogenous in the difference GMM and system GMM estimations. Valid instruments require the following conditions: a lack of correlation with the error terms; a correlation with independent variables; and, if plural, mutual independence. I select the three variables referred by Sasao (2014b) as instruments: the fiscal capability index; the number of landfills per area; and the existence of industrial waste taxation in neighboring prefectures. Sasao (2014b) indicates that the fiscal capability index, the number of landfills per area, and the introduction of industrial waste taxation in neighboring prefectures significantly affect the introduction of taxation. The fiscal capability index represents basic financial revenues divided by basic financial requirements. The higher the index, the more abundant a prefecture’s financial resources. Here, this study adds a number of intermediate disposal facilities per area as the instrument variable, since it can also affect the introduction of taxation, in particular Types A and C taxation. However, as Bowsher (2002) highlighted, the use of numerous moment conditions causes the Hansen test of overidentifying restrictions to be undersized, and therefore the restrictions have extremely low power. Mileva (2007) advised to keep the number of instruments less than or equal to the number of groups (N = 47, in this study). Accordingly, two models that differ in which numbers should be considered are estimated in this paper. 4.2. Variables and data Approximately 30% of industrial waste transported to landfills originates from outside the prefecture, according to the Ministry of the Environment (MOE) in Japan. While the level of economic activity in a prefecture can affect the amount of waste generated, the level of economic activity in another prefecture cannot directly affect the amount of waste generated or dumped in landfills in the prefecture. A similar pattern emerges for local waste policy including industrial waste taxes. Consequently, it is probable that the effects of economic activity and waste taxes on waste flows differ in each case: exclusion and inclusion of waste originating from outside the prefecture. Therefore, this study uses data from MOE to examine two types of dependent variables for the amount of the industrial waste landfilled: exclusion and inclusion of waste originating from outside the prefecture. Waste disposed by generators and a part of the waste are excluded because they were not included in the original data. This study focuses on industrial waste categories accounting for a significant portion of landfilled waste: sludge; waste plastics; waste glass, waste concrete and ceramic; slag; debris; and soot

2243

T. Sasao / Waste Management 34 (2014) 2239–2250 Table 2 Ranking of waste categories of landfilled industrial waste. Source: Ministry of the Environment. 2007

1 2 3 4 5 6

*

2008

Category

Amounts of landfilled waste

Share in the total amount of landfilled waste (%)

Category

Amounts of landfilled waste

Share in the total amount of landfilled waste (%)

Category

Amounts of landfilled waste

Share in the total amount of landfilled waste (%)

Sludge Soot and dust Debris Waste plastics Glass* Slag Total of above six categories Total amounts of all categories

7887 2686 2350 1789 1663 1636 18,011

39.2 13.3 11.7 8.9 8.3 8.1 89.4

Sludge Debris Soot and dust Slag Glass* Waste plastics Total of above six categories Total amounts of all categories

6705 2249 2026 1498 1306 1305 15,089

40.1 13.5 12.1 9.0 7.8 7.8 90.3

Sludge Debris Soot and dust Glass* Waste plastics Slag Total of above six categories Total amounts of all categories

4991 1931 1809 1252 1230 1033 12,246

36.7 14.2 13.3 9.2 9.1 7.6 90.1

20,144

16,701

13,591

Grass represents waste glass, waste concrete, and ceramic.

Table 3 Expected categories of waste generated from each industry and finally landfilled.

*

2009

Industries

Waste categories

Agriculture Mining Ceramic and quarrying Other manufacturing industries Construction Electricity, gas and water supply

Waste plastics Sludge, slag, debris Glass* Sludge, waste plastics, soot and dust Sludge, waste plastics, glass*, debris Sludge, soot and dust

Glass represents waste glass, waste concrete, and ceramic.

and dust. Table 2 shows the recent composition of those six categories of industrial waste. It indicates that the six categories account for approximately 90% of the total amount of landfilled industrial waste. For the economic variables, the study focuses on the outputs from industries that generate large amounts of landfilled waste such as agriculture; mining; ceramic and quarrying; other manufacturing industries; construction; and electricity, gas, and water supply. Table 3 shows the expected categories of waste generated from each industry and finally landfilled. Moreover, economic

Table 4 Descriptive statistics of dependent variables. Category Sludge

Waste plastics

Waste glass, waste concrete and ceramic

Slag

Debris

Soot and dust

Abbreviation

Mean

p50

SD

Max

Min

17.5

195.68

1464

0

24

208.36

1466

0

8.9997 9.2592

8.4766 8.4766

36.44

272

0

57.51

412

0

10.1925 10.1962

9.6928 9.7643

24.55

228

0

29.24

255

0

7.6967 3.9318

7.0909 7.5501

36.76

429

0

56.95

507

0

7.5501 6.2166

8.8820 9.5325

119.88

1264

0

126.39

1437

0

7.6502 4.6151

7.0040 6.8035

32.35

311

0

45.57

350

0

9.1696 9.1696

8.6658 8.6658

Amount of waste disposed of in landfills excluding waste from outside the prefecture (ton) Amount of waste disposed of in landfills including waste from outside the prefecture (ton) Change rate of ‘fd’ Change rate of ‘fd2’

fd

88.14

fd2

107.54

d_fd d_fd2

0.1721 0.1440

0.0000 0.0000

Amount of waste disposed of in landfills excluding waste from outside the prefecture (ton) Amount of waste disposed of in landfills including waste from outside the prefecture (ton) Change rate of ‘fd’ Change rate of ‘fd2’

fd

29.40

18

fd2

47.54

28

d_fd d_fd2

0.0842 0.0422

0.0000 0.0000

Amount of waste disposed of in landfills excluding waste from outside the prefecture (ton) Amount of waste disposed of in landfills including waste from outside the prefecture (ton) Change rate of ‘fd’ Change rate of ‘fd2’

fd

20.23

13

fd2

25.52

17

d_fd d_fd2

0.0496 0.0440

0.0000 0.0000

Amount of waste disposed of in landfills excluding waste from outside the prefecture (ton) Amount of waste disposed of in landfills including waste from outside the prefecture (ton) Change rate of ‘fd’ Change rate of ‘fd2’

fd

15.11

1

fd2

21.85

2.25

d_fd d_fd2

0.1300 0.0839

0.0000 0.0000

Amount of waste disposed of in landfills excluding waste from outside the prefecture (ton) Amount of waste disposed of in landfills including waste from outside the prefecture (ton) Change rate of ‘fd’ Change rate of ‘fd2’

fd

67.02

33

fd2

74.44

39

d_fd d_fd2

0.1811 0.1812

0.0339 0.0458

Amount of waste disposed of in landfills excluding waste from outside the prefecture (ton) Amount of waste disposed of in landfills including waste from outside the prefecture (ton) Change rate of ‘fd’ Change rate of ‘fd2’

fd

12.32

0.25

fd2

17.46

1

d_fd d_fd2

0.0945 0.0002

0.0000 0.0000

1.5937 1.6314

1.4597 1.3825

1.1410 0.9953

1.6040 1.4737

1.1922 1.0344

2.2255 2.1109

2244

T. Sasao / Waste Management 34 (2014) 2239–2250

Table 5 Descriptive statistics of independent variables. Category

Abbreviation

Mean

P50

SD

Max

Min

Output (million yen) Agriculture Mining Ceramic and quarrying Other manufacturing industries Construction Electricity, gas and water supply

agr min cem oth con ele

107554.4 14587.89 77197.61 2540474 632900.6 463947.2

77219.5 10110 50488 1560430 385564.5 286058

86840.5 16715.64 75474.85 3156012 721590.4 428489.6

617247 154828 484197 4.41E+07 4791328 2432831

25226 143 2088 141338 110873 85644

Change rate of each output Agriculture Mining Ceramic and quarrying Other manufacturing industries Construction Electricity, gas and water supply

d_agr d_min d_cem d_oth d_con d_ele

0.0892 0.2721 0.1766 0.1914 0.0866 0.0673

0.2690 1.4201 0.5670 2.4170 0.1806 0.2259

Tax Tax rate of Industrial waste taxation type A Tax rate of Industrial waste taxation type B Tax rate of Industrial waste taxation type C

txra txrb txrc

1,000 1,000 1,000

0 0 0

0.0056 0.1352 0.0615 0.0090 0.0522 0.0123 29.78723 206.3574 63.82979

0.0018 0.0858 0.0410 0.0210 0.0519 0.0159

0.0018 0.0858 0.0410 0.0210 0.0519 0.0159

0 0 0

170.181 401.5938 244.7099

Table 6 Estimation results in case of excluding waste originating from other prefectures: Sludge.

L.l_fd

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.881 [29.57]***

0.4956 [6.58]*** 0.0008 [1.88]*

0.7585 [12.78]***

0.5599 [5.12]***

0.5671 [5.29]***

0.56 [5.25]***

0.5677 [5.16]***

0.8327 [12.26]***

0.8392 [13.25]***

0.8352 [11.70]***

0.8463 [13.41]***

txra txrb

0.0003 [1.82]*

txrc

0.0015 [3.23]***

d06

0.3892 [1.85]*

d07 _cons

0.0546 [0.50]

0.4805 [2.04]** 1.1537 [6.16]***

N

423

423

423

2.03

1.73

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

0.5277 [1.87]*

376 26.21 1.37 25 0.49 0.54

376 27.93 1.38 36 0.32 0.5

376 27.56 1.37 25 0.3 0.28

376 26.58 1.38 36 0.3 0.5

0.4788 [2.43]**

0.228 [1.37]

0.1019 [0.63]

0.1922 [1.02]

0.2151 [1.23]

423 151.51 1.55 42 0.45 0.43

423 194.82 1.55 53 0.36 0.71

423 136.83 1.55 42 0.36 0.43

423 196.09 1.56 53 0.47 0.65

#

‘‘GMM’’ represents the difference GMM, and ‘‘sysGMM’’ represents the system GMM. Values in square brackets represent t-statistics by robust estimations in DOLS and LSDV, z-statistics based on bootstrap standard errors in LSDVC, z-statistics by robust estimations in GMMs and sysGMMs, and z-statistics that are computed by Windmeijer (2005) in sysGMMs. ### Instruments in the GMMs and sysGMMs are as follows: Instruments for first differences equation; Standard: D.(fci idf ldf netx). GMM-type: L2.(L.d_fd txra txrb txrc) in GMM1-1, GMM2-1, sysGMM1-1, and sysGMM2-1. L(2/3).(L.d_fd txra txrb txrc) in GMM1-2, GMM2-2, sysGMM1-2, and sysGMM2-2. Instruments for level equation (sysGMM only); Standard: fci idf ldf netx cons; GMM-type: DL.(L.l_fd txra txrb txrc). Where, ‘‘D.X’’ represents the first differences, that is Xt–Xt1., and ‘‘L(2/3). X’’ represents that Xt2 and Xt3 are used as instruments. $ These notes are same in the following tables (Tables 7–17). * p < 0.1. ** p < 0.05. *** p < 0.01. ##

variables are examined with a one year lag to compensate for the time elapsed between production and final disposal. For taxation, the tax rates are examined considering the differences between the three taxation types noted in Section 2. These data are taken from the Ministry for Internal Affairs and Communications. Correlation coefficients indicate that the correlations between these explanatory variables are negligible. Table 4 shows the descriptive statistics of dependent variables and Table 5 shows those of independent variables.

5. Estimation results 5.1. Exclusion of industrial waste originating from other prefectures Tables 6–11 show the estimation results in cases of excluding industrial waste originating from other prefectures, for each category of waste. Although the original data covers 10 years, one or two years of data are dropped because of taking differences and lags in the analysis. Consequently, the total number of samples is 376 or 423, (47 prefectures multiplied by 8 or 9 years) for all models.

2245

T. Sasao / Waste Management 34 (2014) 2239–2250 Table 7 Estimation results in case of excluding waste originating from other prefectures: Waste plastics.

L.l_fd

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.7153 [10.33]⁄⁄⁄

0.3123 [4.36]⁄⁄⁄

0.4763 [8.10]⁄⁄⁄ 2.3786 [3.21]⁄⁄⁄

0.1805 [0.88]

0.1083 [0.77]

0.1709 [0.96]

0.1143 [0.82]

0.6217 [6.85]⁄⁄⁄

0.6313 [7.04]⁄⁄⁄

0.6582 [8.19]⁄⁄⁄

0.6625 [8.95]⁄⁄⁄

5.903 [2.14]⁄⁄

6.878 [1.97]⁄⁄

0.7902 [3.33]⁄⁄⁄

0.7542 [3.33]⁄⁄⁄

0.7676 [3.54]⁄⁄⁄

0.7598 [3.47]⁄⁄⁄

423 90.14 1.39 42 0.25 0.46

423 112.97 1.42 53 0.25 0.48

423 67.15 1.34 42 0.38 0.15

423 80.14 1.35 53 0.38 0.21

d_agr d_ele d07

0.4998 [2.47]⁄⁄

Constant

0.6286 [2.93]⁄⁄⁄

1.6378 [9.12]⁄⁄⁄

N

423

423

423

1.62

1.26

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 0.77 1.07 25 0.39 0.28

376 0.60 1.04 36 0.35 0.03

376 0.93 1.06 25 0.33 0.28

376 0.67 1.04 36 0.3 0.03

Table 8 Estimation results in case of excluding waste originating from other prefectures: waste glass, waste concrete and ceramic.

L.l_fd d_agr d_con d_ele

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.7908 [16.50]⁄⁄⁄ 1.37 [2.27]⁄⁄ 1.4807 [2.66]⁄⁄⁄ 1.1137 [1.96]⁄

0.3009 [3.12]⁄⁄⁄ 1.1315 [1.89]⁄ 1.2985 [2.53]⁄⁄ 1.3547 [2.14]⁄⁄

0.4622 [8.82]⁄⁄⁄

0.0344 [0.24]

0.0495 [0.38]

0.0995 [0.58]

0.0152 [0.08]

0.829 [10.94]⁄⁄⁄

0.8354 [13.88]⁄⁄⁄

0.833 [8.62]⁄⁄⁄

0.8398 [13.75]⁄⁄⁄

1.4194 [2.20]⁄⁄ 2.8399 [2.15]⁄⁄

L.d_con

2.8952 [1.95]⁄

2.9909 [2.64]⁄⁄⁄

txra constant

0.4753 [3.36]⁄⁄⁄

0.0006 [1.76]⁄ 1.5418 [7.33]⁄⁄⁄

N

423

423

423

1.41

0.97

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 0.06 0.78 25 0.6 0.5

376 0.14 0.81 36 0.31 0.43

376 0.33 0.77 25 0.93 0.5

376 0.01 0.80 36 0.57 0.43

0.1345 [0.64]

0.2722 [1.78]⁄

0.3303 [1.35]

0.2635 [1.69]⁄

376 119.83 1.21 42 0.25 0.79

423 210.52 1.21 53 0.13 0.63

423 74.23 1.21 42 0.14 0.46

423 201.81 1.21 53 0.14 0.63

Table 9 Estimation results in case of excluding waste originating from other prefectures: slag.

L.l_fd d_agr

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.8778 [37.57]⁄⁄⁄ 2.2774 [2.70]⁄⁄⁄

0.1461 [2.08]⁄⁄

0.2978 [5.33]⁄⁄⁄

0.2454 [1.93]⁄

0.103 [0.78]

0.2614 [1.94]⁄

0.1064 [0.74]

0.9086 [20.13]⁄⁄⁄

0.9367 [25.47]⁄⁄⁄

0.9349 [18.77]⁄⁄⁄

0.9372 [24.96]⁄⁄⁄

0.139 [2.55]⁄⁄

0.1362 [2.94]⁄⁄⁄

0.1529 [2.14]⁄⁄

0.1388 [3.02]⁄⁄⁄

423 405.32 1.80 42 0.24 0.79

423 648.52 1.82 53 0.18 0.69

423 352.26 1.82 42 0.18 0.61

423 623.15 1.82 53 0.18 0.69

txrc Constant

0.1292 [1.72]⁄

0.0011 [3.04]⁄⁄⁄ 0.1395 [5.63]⁄⁄⁄

N

423

423

423

2.85

1.30

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 3.71 1.07 25 0.57 0.8

376 0.62 1.11 36 0.91 0.74

GMM1-1 and GMM2-1 are models that limit the instruments to only two year lags, and GMM1-2 and GMM2-2 limit the instruments to two and three year lags. GMM1-1 and GMM1-2 are one-step estimations, and GMM2-1 and GMM2-2 are two-step estimations. A similar pattern emerges for the system GMMs,

376 3.77 1.06 25 0.51 0.8

376 0.54 1.11 36 0.9 0.74

‘‘sysGMM’’ in the tables. P-values in the second-order AR tests (AR(2)) are larger than 0.1 for all the waste categories other than debris, in both the difference GMM and system GMM models. Therefore, we do not reject the null hypothesis for all the waste categories other than for debris in GMM1-1, GMM1-2, system

2246

T. Sasao / Waste Management 34 (2014) 2239–2250

Table 10 Estimation results in case of excluding waste originating from other prefectures: Debris.

L.l_fd d_agr

DOLS

LSDV

LSDVC

GMM11

GMM12

GMM21

GMM22

sysGMM11

sysGMM12

sysGMM21

sysGMM22

0.8629 [19.90]⁄⁄⁄ 1.3903 [2.08]⁄⁄

0.3671 [4.48]⁄⁄⁄

0.4985 [9.33]⁄⁄⁄

0.306 [1.12]

0.2463 [1.34]

0.3013 [1.00]

0.32 [1.53]

0.9054 [9.98]⁄⁄⁄

0.9056 [10.90]⁄⁄⁄

0.9103 [9.11]⁄⁄⁄

0.8972 [9.65]⁄⁄⁄

0.0007 [1.87]⁄

3.4925 [2.08]⁄⁄ 0.0007 [1.91]⁄

0.0856 [0.31]

0.1047 [0.46]

0.1226 [0.37]

0.0098 [0.03]

423 142.33 1.40 42 0.06 0.7

376 192.39 1.39 53 0.15 0.75

423 82.97 1.40 42 0.09 0.65

376 105.82 1.38 53 0.18 0.79

L.d_min

1.5021 [2.41]⁄⁄

L.d_con txra d06

2.9399 [2.11]⁄⁄

0.3189 [1.78]⁄ 0.401 [2.36]⁄⁄

d07 d08

Constant

0.2448 [1.47]

0.1994 [2.03]⁄⁄ 0.4311 [3.43]⁄⁄⁄ 1.8191 [7.26]⁄⁄⁄

N

423

423

423

1.61

1.03

d09

v2 r

0.4613 [3.35]⁄⁄⁄

N of instruments AR2 (p-value) Hansen test (p-value)

376 1.26 1.00 25 0 0.21

329 18.97 1.00 36 0.07 0.53

376 1.01 1.00 25 0.24 0.21

376 2.33 1.01 36 0.18 0.55

Table 11 Estimation results in case of excluding waste originating from other prefectures: Soot and dust.

L.l_fd d06 constant N

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.7447 [21.92]⁄⁄⁄ 0.7396 [2.55]⁄⁄ 0.0186 [0.18]

0.1794 [3.44]⁄⁄⁄

0.2802 [5.97]⁄⁄⁄

0.0034 [0.02]

0.047 [0.29]

0.0034 [0.02]

0.0472 [0.29]

0.7419 [7.14]⁄⁄⁄

0.8343 [13.49]⁄⁄⁄

0.7904 [7.38]⁄⁄⁄

0.8348 [13.77]⁄⁄⁄

0.4135 [12.81]⁄⁄⁄

0.0653 [0.47]

0.0081 [0.08]

0.0191 [0.13]

0.006 [0.06]

423

423

423

2.78

1.67

423 50.93 2.06 42 0.42 0.48

423 181.98 2.17 53 0.42 0.67

423 54.49 2.11 42 0.41 0.72

423 189.50 2.17 53 0.48 0.97

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 0.00 1.37 25 0.19 0.57

376 0.08 1.40 36 0.17 0.53

376 0.00 1.37 25 0.11 0.57

376 0.08 1.40 36 0.12 0.53

Table 12 Estimation results in case of including waste originating from other prefectures: Sludge.

L.l_fd2

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.9198 [36.14]⁄⁄⁄

0.6686 [10.26]⁄⁄⁄

0.6096 [6.03]⁄⁄⁄

0.6445 [6.07]⁄⁄⁄

0.5959 [6.17]⁄⁄⁄

0.641 [6.18]⁄⁄⁄

0.8588 [14.17]⁄⁄⁄

0.8963 [16.54]⁄⁄⁄

0.864 [13.84]⁄⁄⁄

0.8981 [17.17]⁄⁄⁄

0.2051 [1.24]

0.1124 [0.76]

0.2253 [1.12]

0.1149 [0.73]

423 200.84 1.24 42 0.58 0.58

423 273.49 1.26 53 0.59 0.74

423 191.55 1.24 42 0.58 0.58

423 294.67 1.26 53 0.59 0.74

_cons

0.0545 [0.50]

0.5682 [9.07]⁄⁄⁄ 0.0007 [1.70]⁄ 0.0013 [2.55]⁄⁄ 0.4619 [2.28]⁄⁄ 1.0934 [6.22]⁄⁄⁄

N

423

423

423

1.75

1.67

txra txrc d07

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 36.37 1.12 25 0.48 0.54

376 36.82 1.13 36 0.5 0.49

376 38.12 1.11 25 0.47 0.54

376 38.24 1.13 36 0.49 0.49

2247

T. Sasao / Waste Management 34 (2014) 2239–2250 Table 13 Estimation results in case of including waste originating from other prefectures: Waste plastics.

L.l_fd2 d05 Constant N

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.7456 [10.89]⁄⁄⁄ 0.4816 [3.19]⁄⁄⁄ 0.767 [3.14]⁄⁄⁄

0.3282 [5.73]⁄⁄⁄ 0.3286 [2.18]⁄⁄ 1.9895 [12.02]⁄⁄⁄

0.5075 [8.40]⁄⁄⁄ 0.3588 [1.95]⁄

0.1502 [1.18]

0.0111 [0.12]

0.1866 [1.37]

0.0095 [0.10]

0.7286 [16.30]⁄⁄⁄

0.8078 [31.91]⁄⁄⁄

0.7254 [15.23]⁄⁄⁄

0.8093 [31.24]⁄⁄⁄

0.7638 [4.67]⁄⁄⁄

0.5286 [4.72]⁄⁄⁄

0.7748 [4.64]⁄⁄⁄

0.5245 [4.62]⁄⁄⁄

423

423

423

1.57

1.22

423 265.54 1.30 42 0.35 0.38

423 1018.37 1.35 53 0.35 0.67

423 232.01 1.29 42 0.35 0.38

423 975.85 1.35 53 0.35 0.67

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 1.39 0.98 25 0.26 0.16

376 0.01 0.94 36 0.2 0.46

376 1.88 0.99 25 0.24 0.16

376 0.01 0.94 36 0.16 0.46

Table 14 Estimation results in case of including waste originating from other prefectures: Waste glass, waste concrete and ceramic.

L.l_fd2 d_agr d_con d_ele

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.8408 [22.64]⁄⁄⁄ 1.2314 [2.21]⁄⁄ 1.577 [3.16]⁄⁄⁄ 0.8804 [1.77]⁄

0.3885 [4.37]⁄⁄⁄ 1.1491 [1.97]⁄ 1.4167 [2.86]⁄⁄⁄

0.5484 [9.67]⁄⁄⁄

0.0098 [0.03]

0.089 [0.67]

0.0841 [0.29]

0.0577 [0.38]

0.8601 [13.51]⁄⁄⁄

0.8627 [17.43]⁄⁄⁄

0.8498 [11.08]⁄⁄⁄

0.8633 [16.15]⁄⁄⁄

0.3471 [1.55]

0.312 [1.90]⁄

1.4181 [2.44]⁄⁄

1.4469 [1.71]⁄

txra 0.3672 [2.52]⁄⁄

0.2903 [2.45]⁄⁄

0.0001 [3.32]⁄⁄⁄ 0.2354 [1.75]⁄

Constant

0.3882 [3.19]⁄⁄⁄

1.5198 [7.14]⁄⁄⁄

0.2717 [1.45]

0.0001 [3.37]⁄⁄⁄ 0.3099 [2.29]⁄⁄ 0.4112 [1.91]⁄ 0.2116 [1.49]

N

423

423

423

1.25

0.86

376 3.67 0.67 36 0.88 0.5

423 202.51 1.07 42 0.01 0.46

423 449.09 1.07 53 0.02 0.6

423 122.69 1.06 42 0.01 0.44

423 260.86 1.07 53 0.01 0.75

d03 d06

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 0.00 0.69 25 0.66 0.36

376 0.45 0.68 36 0.81 0.43

376 0.08 0.68 25 0.89 0.36

Table 15 Estimation results in case of including waste originating from other prefectures: slag.

L.l_fd2 d_agr

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.9021 [41.47]⁄⁄⁄ 1.7707 [2.30]⁄⁄

0.3042 [3.17]⁄⁄⁄

0.5 [8.03]⁄⁄⁄

0.0561 [0.26]

0.0537 [0.26]

0.0613 [0.26]

0.0614 [0.31]

0.9058 [27.68]⁄⁄⁄

0.9441 [35.89]⁄⁄⁄

0.9225 [27.05]⁄⁄⁄

0.9449 [34.42]⁄⁄⁄

0.0641 [1.15]

0.0722 [1.53]

0.0374 [0.53]

0.0701 [1.68]⁄

423 766.36 1.57 42 0.83 0.65

423 1287.80 1.60 53 0.82 0.7

423 731.86 1.59 42 0.83 0.65

423 1184.97 1.60 53 0.83 0.7

txrc Constant

0.0534 [0.76]

0.0009 [2.88]⁄⁄⁄ 0.1255 [3.50]⁄⁄⁄

N

423

423

423

2.44

1.30

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 0.07 1.04 25 0.46 0.74

376 0.07 1.04 36 0.45 0.89

GMM1-1, and system GMM2-1. However, we reject the null hypothesis for debris because p-values in AR(2) are less than 0.1. In addition, for the Hansen J-statistic tests to check the instruments’ validity, p-values in the tests are larger than 0.1 for all the waste categories other than for waste plastics in the GMM12 and GMM2-2. Overidentifying restrictions are valid and the model specification is correct for all the waste categories except waste plastics in GMM1-2 and GMM2-2.

376 0.07 1.04 25 0.41 0.74

376 0.10 1.04 36 0.37 0.89

However, let us focus on the parameter of the lagged dependent variable (b in the equation in the previous section), ‘‘L.l_fd’’ in the tables, to check the robustness of each estimation model. For all waste categories, the order of the parameters by DOLS, LSDV, and LSDVC estimators is LSDV < LSDVC < DOLS. However, the parameters of the GMM and system GMM estimations are not stable, and different with each model. For sludge, the parameters by the GMM estimations are located between LSDV and DOLS, whereas

2248

T. Sasao / Waste Management 34 (2014) 2239–2250

Table 16 Estimation results in case of including waste originating from other prefectures: debris.

L.l_fd2 L.d_ele

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.8873 [23.46]⁄⁄⁄ 0.9993 [1.91]⁄

0.4129 [4.86]⁄⁄⁄

0.5685 [10.81]⁄⁄⁄

0.615 [2.48]⁄⁄

0.4045 [2.66]⁄⁄⁄

0.5822 [3.15]⁄⁄⁄

0.5161 [3.88]⁄⁄⁄

0.9867 [24.19]⁄⁄⁄

0.9985 [26.51]⁄⁄⁄

0.9981 [25.12]⁄⁄⁄

1.0066 [25.12]⁄⁄⁄

0.0005 [8.35]⁄⁄⁄

2.5611 [2.00]⁄⁄ 0.0006 [6.56]⁄⁄⁄

0.155 [1.15]

0.3059 [2.09]⁄⁄

0.1617 [1.16]

0.2009 [1.43]

423 1084.69 1.22 42 0.04 0.47

376 1447.93 1.19 53 0.33 0.66

423 631.00 1.23 42 0.04 0.49

423 630.90 1.24 53 0.04 0.67

L.d_min

1.6087 [2.56]⁄⁄

L.d_con txra

constant

0.2217 [1.49]

0.0007 [16.59]⁄⁄⁄ 0.3852 [3.87]⁄⁄⁄ 0.2093 [2.48]⁄⁄ 0.4846 [4.23]⁄⁄⁄ 1.8444 [6.69]⁄⁄⁄

N

376

423

423

1.44

0.91

d07 d08 d09

v2 r

0.3037 [2.04]⁄⁄

0.3881 [3.19]⁄⁄⁄

N of instruments AR2 (p-value) Hansen test (p-value)

376 6.15 1.00 25 0.04 0.54

329 42.27 0.91 36 0.08 0.73

376 9.94 0.98 25 0.08 0.54

376 15.06 0.95 36 0.06 0.71

Table 17 Estimation results in case of including waste originating from other prefectures: soot and dust.

L.l_fd2

DOLS

LSDV

LSDVC

GMM1-1

GMM1-2

GMM2-1

GMM2-2

sysGMM1-1

sysGMM1-2

sysGMM2-1

sysGMM2-2

0.8018 [24.94]⁄⁄⁄ 1.0646 [2.59]⁄⁄

0.2798 [5.49]⁄⁄⁄

0.2734 [1.21]

0.1757 [1.20]

0.262 [1.10]

0.1688 [1.14]

0.8044 [8.80]⁄⁄⁄

0.8931 [18.17]⁄⁄⁄

0.8434 [8.69]⁄⁄⁄

0.8907 [18.41]⁄⁄⁄

0.0212 [0.21]

0.0117 [0.16]

0.0079 [0.08]

0.0072 [0.09]

423 77.46 2.13 42 0.64 0.73

423 330.32 2.23 53 0.64 0.71

423 75.58 2.17 42 0.63 0.73

423 338.97 2.23 53 0.64 0.71

Constant

0.1297 [1.30]

0.1624 [2.45]⁄⁄ 0.8045 [2.04]⁄⁄ 0.4967 [2.39]⁄⁄ 0.1078 [2.21]⁄⁄

N

376

376

423

2.79

1.62

L.d_min d07

v2 r N of instruments AR2 (p-value) Hansen test (p-value)

376 1.47 1.28 25 0.13 0.84

376 1.45 1.30 36 0.09 0.85

those by the system GMM estimations are high and near DOLS. On the other hand, for waste plastics, the parameters by the GMM estimations are smaller than by DOLS, whereas those by the system GMM estimations are somewhat near the DOLS rather than between the LSDV and DOLS. For waste glass, waste concrete, and ceramic, debris, and soot and dust, the parameters by the GMM estimations are smaller than by the DOLS, while those by the system GMM estimations are larger than DOLS. Consequently, the study focuses on the estimation results by the LSDVC.6 On the whole, ‘‘r’’ in the tables that represents the square root of the error variance estimate by the LSDVC is smaller than those by the DOLS, LSDV, and system GMMs, although it is higher than those by the difference GMMs. Therefore, the estimation results by the LSDVC are more robust than other models, supporting Judson and Owen (1999) and Bruno (2005). The following are the main estimation results for each waste category. For sludge, no significant effects are observed for taxation in 6 In the LSDVC, we can select the level of the accuracy on possible bias approximations (Bruno, 2005). The estimation results in the paper show the results in case of an approximation up to 1/T, the default level by Stata (a software package for econometric analysis). Although I also examined other levels of the accuracy (N1T1and N1T2), the results were almost similar among the three approximation levels.

376 1.20 1.29 25 0.08 0.84

376 1.29 1.31 36 0.06 0.85

other than the LSDV, although negative significant effects are observed for Types A and C taxation in the LSDV. For waste plastics, no significant effects are observed for taxation in all the estimation models. For the economic determinants, only the output of agriculture is significantly positive in the LSDVC. This finding is similar to the study’s a priori expectation. For waste glass, waste concrete, and ceramic, no significant effects are observed for taxation besides in the LSDV, although a negative significant effect is observed for Type A taxation in the LSDV. For the economic determinants, only the output of construction is significantly positive in the LSDVC. This finding is similar to the study’s a priori expectation. For slag, no significant effects are observed for taxation other than in the LSDV although a negative significant effect is observed for Type C taxation in the LSDV. For debris, no significant effects are observed for taxation except in the system GMM, although positive significant effects are observed for Type A taxation in the system GMM despite a 10% significance level. For soot and dust, no significant effects are observed for taxation in any of the estimation models. 5.2. Inclusion of industrial waste originating from other prefectures Tables 12–17 show the estimation results in case of including industrial waste originating from other prefectures, for each

2249

T. Sasao / Waste Management 34 (2014) 2239–2250 Table 18 Annual effects after introducing industrial waste taxes. Excluding of waste originated from other prefecture Sludge Waste plastics Glass# Slag Debris Soot and dust * ** *** # ##

***

Including of waste originated from other prefecture Type C 3rd year (–)*** N.S. N.S. N.S. Type A 4th year (–)**, 6th year (+)* Type B 3rd year (+)*

Type C 3rd year (–) N.S. ## N.S. Type A 2nd year (+)**, Type C 4th year (–)* Type A 1st year (–)***, 2nd year (+)** Type B 3rd year (+)*

p < 0.1. p < 0.05. p < 0.01. Glass represents waste glass, waste concrete, and ceramic. N.S. represents not significant.

category of waste. P-values in AR(2) are less than 0.1 for waste glass, waste concrete, and ceramic in the system GMM models, debris in all the difference GMMs and system GMMs other than system GMM1-2, soot and dust in the difference GMMs other than the GMM1-1, although they are greater than 0.1 for other waste categories and models. For the Hansen J-statistic tests, p-values in the tests are greater than 0.1 for all the waste categories. Overidentifying restrictions are valid and the model specification is correct for all the waste categories in both the difference GMM and system GMM. However, the parameters of the lagged dependent variables, by both the difference GMM and system GMM estimations, are not stable and again differ across each model again. Therefore, this subsection also focuses on the estimation results by the LSDVC. For sludge, a negative significant effect is observed for Type C taxation in the LSDVC although negative significant effects are observed for Types A and C taxation in the LSDV. This is the only significant effect in the LSDVC. This result indicates that an increase of 1 yen in taxation decreases landfilled waste by 0.0014% over the previous year, that is, an increase of 1000 yen decreases landfilled wasted by 1.4%. In case of excluding waste originating from other prefectures, taxation was not significant in the LSDVC. This indicates that taxation may restrict waste inflow from other prefectures. For waste plastics, no significant effects are observed for taxation in any of the models. For waste glass, waste concrete, and ceramic, no significant effects are observed for taxation in other than the system GMM, although positive significant effects are observed for Type A taxation in the system GMM. For the economic determinants, only the output of construction is significantly positive in the LSDVC. This finding is similar to the study’s a priori expectation. For slag, no significant effects are observed for taxation in other than the LSDV, although a negative significant effect is observed for Type C taxation in the LSDV. For debris, no significant effects are observed for taxation in the DOLS, LSDVC, difference GMMs, and system GMMs 1-1 and 1-2, although positive significant effects are observed for Type A taxation in the LSDV and system GMMs 2-1 and 2-2. For soot and dust, no significant effects are observed for taxation in any of the estimation models. 5.3. Annual effects after taxation This subsection examines the robustness of the results in the previous subsections using yearly taxation dummy variables. Therefore, TAXs (the rate of industrial waste taxation) are replaced by yearly taxation dummy variables in the equation in Section 4.1. This estimation can clarify the timing of the effect’s emergence by taxation, or the annual effects after introducing industrial waste taxes. The main results estimated by LSDVC are presented in Table 18 because the parameters of the lagged dependent variable

by both the GMM and system GMM estimations are not stable, and again differ across each model. The table shows the taxation types and years that significant effects were observed after introducing industrial waste taxes. A positive/negative sign in the parentheses represents a significant/ negative positive effect. In case of excluding waste originating from other prefectures, significant negative effects are observed in Types A and C taxation for sludge, slag, and debris while no significant effects were observed in the previous subsection. However, the emergence of the significant effects is temporal, and a continuous reduction is not observed. Even some positive effects are observed in Types A and B taxation for slag, debris, and soot and dust. In case of including waste originating from other prefectures, significant negative effects are observed in Types A and C taxation for sludge and debris. The negative effects by Type C taxation for sludge are similar to the results in the previous subsection. However, a continuous reduction is not observed after taxation. 6. Conclusion The empirical results show that industrial waste taxes in Japan have minimal significant effects on the reduction of final disposal amounts, considering dynamic relationships between dependent variables and waste categories thus far. As Sasao (2014c) highlighted, some factors resulted in these minimal impacts. One possible reason is the low tax rate of 1000 yen (10 dollars) per ton on waste transported to landfills. For example, the fee for transporting ashes to a landfill in a prefecture is 24,000 yen (240 dollars) per ton and that for inorganic sludge is 22,000 yen (220 dollars) per ton. The ratios of taxes to fees seem to be smaller than those in most EU member countries (Bio Intelligence Service, 2012). On the other hand, the standard tax rate is higher than the rate of landfill tax for household MSW in some states in the United States, as noted in Kinnaman (2006). In contrast, it is lower than the average state tax on disposal of hazardous waste in the United States: around 15 dollars per ton (in 1995 dollars), as noted in Levinson (1999a). The properties of the industrial waste addressed in this paper are situated between MSW and hazardous waste in the properties. It is important to reconsider the tax rate from the standpoint of internalization of externality. Ley et al. (2002) employed an average externality of 20 dollars per ton to examine MSW shipments. Kinnaman (2006) suggested the external costs of transportation and disposal of MSW between 5.48 and 8.76 dollars per ton. If the waste disposal market is relatively competitive, then we would expect the taxes to be set equal to the marginal external cost, as shown by Buchanan (1969). The standard rate of industrial waste taxes is situated between the estimates by Kinnaman (2006) and those by Ley et al. (2002). If one regards the tax rate as consistent with compelling estimates of marginal external cost in reasonably competitive disposal market, the current tax rate may be set at just

2250

T. Sasao / Waste Management 34 (2014) 2239–2250

about economically efficient levels. In that case, the estimation results in this paper may indicate that the growth rates of landfilled industrial waste have not fallen more rapidly over time as policymakers might expect. However, Sasao (2004) pointed out that the public considers industrial waste more harmful than MSW using a choice experiment. Moreover, he showed that this is particularly true if the industrial waste to be accepted in a local landfill is imported from another region. If the marginal external costs are higher than the current tax rates, it could be argued that different tax rates be set according to waste categories, considering varied environmental impacts and the origin of waste, even when disposal methods are similar.7 Finally, this paper focused on the estimation results by the LSDVC estimation after checking econometric robustness. However, focusing now on the economic determinants, only some economic variables used in the analysis affected the dependent variables as significantly as anticipated. The selection of more relevant models or instrumental variables may improve the statistical reliability of other econometric models including the GMM estimations. This issue will be the subject of further research. Acknowledgements I am grateful for the helpful comments provided by three anonymous referees. This study was supported by a Grant-in-Aid for Young Scientists (B) (25740059) by the Ministry of Education, Culture, Sports, Science and Technology, Japan. References Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev. Econ. Stud. 58, 277–297. Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation of error-components models. J. Econ. 68 (1), 29–51. Baltagi, B.H., 2008. Econometric Analysis of Panel Data. fourth ed. John Wiley & Sons Ltd. Bartelings, H., van Beukering, P., Kuik, O., Linderhof V., Oosterhuis, F., 2005. Effectiveness of Landfill Taxation. Institute for Environmental Studies, Netherlands. BIO Intelligence Service, 2012. Use of economic instruments and waste management performances. Final Report 10 April 2012, Contract ENV.G.4/ FRA/2008/0112. Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. J. Econ. 87, 115–143. Bond, S.R., 2002. Dynamic panel data models: a guide to micro data methods and practice. Portuguese Econ. J. 1, 141–162. Bowsher, C.G., 2002. On testing overidentifying restrictions in dynamic panel data models. Econ. Lett. 77, 211–220.

7 From another perspective, Cossu and Masi (2013) points out that the applied landfill tax in weight is not an effective measure to reduce the use of landfills and could be replaced by another measure based on the volume.

Bruno, G.S.F., 2005. Estimation and inference in dynamic unbalanced panel-data models with a small number of individuals. Stata J. 5 (4), 473–500. Buchanan, J.M., 1969. External diseconomies, corrective taxes, and market structure. Am. Econ. Rev. 59 (1), 174–177. Cossu, R., Masi, S., 2013. Re-thinking incentives and penalties: economic aspects of waste management in Italy. Waste Manage. 33, 2541–2547. D’Amato, A., Massimiliano, M., Nicolli, F., 2014. Illegal Waste Disposal, Territorial Enforcement and Policy: Evidence from Regional Data. SEEDS Working Paper 03/2014. Judson, R.A., Owen, A.L., 1999. Estimating dynamic panel data models: a guide for macroeconomists. Econ. Lett. 65, 9–15. Kinnaman, T.C., 2006. Examining the justification for residential recycling. J. Econ. Perspect. 20 (4), 219–232. Kinnaman, T.C., 2009. The economics of municipal solid waste management. Waste Manage. 29, 2615–2617. Kiviet, J.F., 1995. On bias, inconsistency, and efficiency of various estimators in dynamic panel data models. J. Econ. 68, 53–78. Levinson, A., 1999a. State taxes and interstate hazardous waste shipments. Am. Econ. Rev. 89 (3), 666–677. Levinson, A., 1999b. NIMBY tax matter: the case of state hazardous waste disposal taxes. J. Public Econ. 74, 31–51. Ley, E., Macauley, M.K., Salant, S.W., 2002. Spatially and intertemporally efficient waste management: the costs of interstate trade restrictions. J. Environ. Econ. Manage. 43, 188–218. Mazzanti, M., Montini, A., Nicolli, F., 2012. Waste dynamics in economic and policy transitions: decoupling, convergence and spatial effects. J. Environ. Planning Manage. 55 (5), 563–581. Mileva, E., 2007. Using Arellano – Bond dynamic panel GMM estimators in Stata, Economics Department, Fordham University. Morotomi, T., 2003. Environmental management through industrial waste taxes: theoretical foundations and policy designs. Waste Manage. Res. 14 (4), 182–193 (in Japanese). Nicolli, F., Mazzanti, M., 2013. Landfill diversion in a decentralized setting: a dynamic assessment of landfill taxes. Resour. Conserv. Recycl. 81, 17–23. Presbitero, A.F., 2005. The Debt-Growth Nexus: a Dynamic Panel Data Estimation. Universita Politecnica delle Marche Economics Working Paper No. 243. Roodman, D., 2009. How to do xtabond2: an introduction to difference and system GMM in Stata. Stata J. 9 (1), 86–136. Sasao, T., 2004. An estimation of the social costs of landfill siting using a choice experiment. Waste Manage. 24, 753–762. Sasao, T., 2014a. Effects of local waste taxation and trade restrictions on industrial waste flow in Japan. In: Asano, K., Takada, M. (Eds.), Rural and Urban Sustainability Governance, United Nations University Press, pp. 59–80 (Chapter 4). Sasao, T., 2014b. Industrial waste shipments and trade restrictions. In: Kinnaman, T., Takeuchi, K. (Eds.), Handbook on Waste Management. Edward Elgar Publishing, pp. 186–215 (Chapter 7). Sasao, T., 2014c. Does industrial waste taxation contribute to waste reduction? Panel data analysis of the generation and final disposal of industrial waste in Japan. In: Environmental Taxation and Green Fiscal Reform: Theory and Impact (Critical Issues in Environmental Taxation Series). Edward Elgar Publishing, pp. 245–259 (Chapter 16) (forthcoming). Sigman, H., 1996. The effects of hazardous waste taxes on waste generation and disposal. J. Environ. Econ. Manage. 30, 199–217. Windmeijer, F., 2005. A finite sample correction for the variance of linear efficient two-step GMM estimators. J. Econ. 126, 25–51.