Relief for the environment? The importance of an increasingly unimportant industrial sector.

Gassebner, Martin ; Gaston, Noel ; Lamla, Michael J. 等

I. INTRODUCTION

Among the more controversial views about economic growth and globalization is that both will eventually benefit the environment (Arrow et al. 1995). In part, this view is predicated on the nature of structural changes that are normally associated with trade liberalization and economic development. More specifically, economic growth and the shift of production away from polluting sectors and "dirty" technologies help to arrest the deterioration in the environment. In addition, environmental quality is a normal good, and wealthier economies will invest more heavily in environmental improvements and cleanup. According to this line of argument, another implication is that developing countries inevitably focus first on manufacturing production and basic forms of production while tolerating some degradation in the quality of the environment. Compounding this feature is the fact that the political pressures associated with industrialization are also likely to be influential. The factor owners employed in manufacturing industries lobby for less regulation of polluting activities. This accelerates the decay of the environment.

With the inevitable economic decline of basic manufacturing activities in more mature economies, the declining significance of basic manufacturing in industrialized countries may very well create social pressures that reduce the demand for pollution abatement. For instance, it has been argued that greater inequality of wealth and income could be bad news for the environment (see Boyce 1994; Torras and Boyce 1998). Other studies show that the pattern of sectoral resource ownership matters and that greater income inequality can yield either stricter or weaker environmental policies. For example, McAusland (2003) showed that the owners of clean factors of production may be less green voters because they may bear the burden of pollution taxes through adverse terms of trade effects on the production of "clean goods." However, in this paper, we propose the argument that associated with falling industrial wages may be declining political influence exercised by the factor owners in the polluting manufacturing industries of the economy. These latter features are likely to be manifested in the political process, that is, voting for change and a cleaner environment. In other words, structural change may not only involve less reliance being placed on the use of polluting inputs but also have the signal virtue of altering the demand for environmental policies.

More liberalized trade and the rapid onset of skill-biased technological change have been linked with the declining real incomes received by production workers in manufacturing industries. (1) Free trade raises national income which, in aggregate terms, increases the value placed on the environment. Political economic considerations are therefore likely to lead to a cleaner environment. Trade liberalization, which some authors continue to associate with increasing income inequality in Organisation for Economic Co-operation and Development (OECD) countries, may therefore be a "pro-environment" policy (see Bommer and Schulze 1999; Grossman and Krueger 1993, for instance).

Associated with this relatively sanguine view has been an empirical relationship--in the form of an inverted U-shaped curve-between per capita income and various measures of environmental degradation. The relationship, or the environmental Kuznets curve, has been investigated for a wide variety of environmental indicators (e.g., Dinda 2004, Grossman and Krueger 1995; Selden and Song 1994; Shafik 1994). For any country, the implication is that economic growth will be associated with environmental degradation until a "critical" level of per capita income is attained; from that point, there will be an improvement in environmental conditions.

Of course, the turning points in the relationship between economic growth and environmental quality can be affected by the policies implemented by decision makers (Grossman and Krueger 1995; Shafik 1994). Consequently, different political processes do not all imply that societies will grow their way out of environmental problems or that policies that promote economic growth can substitute for environmental policies.

This paper is also indirectly related to the political economy literature that deals with the effect of income inequality on redistributive policies and economic growth (e.g., Alesina and Rodrik 1994; Persson and Tabellini 1994; Saint Paul and Verdier 1996). A standard argument is that when income is more unequally distributed, the median voter is likely to be relatively less endowed with capital, either physical or human, and to thus favor a higher rate of capital taxation. A similar argument may well apply to pollution abatement policies. For instance, if the median voter is a low-income worker who receives their livelihood from supplying labor to the basic manufacturing or pollution-intensive sectors, then greater income inequality may be associated with damage to the environment because it reduces the demand for pollution abatement.

However, while environmental policies are shaped by the importance of potentially affected constituencies, the relative political importance of different constituencies is likely to change over time. The idea of an interaction between industry decline and endogenous policy formation is not a novel one, of course (e.g., Cassing and Hillman 1985). However, the perspective explored here is that the declining economic significance of polluting sectors in a developed economy is likely to be associated with greater income inequality. In turn, this is likely to reduce the "political clout" of the factor owners in the polluting sectors. In particular, as the workers in these sectors of the economy become less important economically, as reflected by their falling real incomes and falling employment levels, they also become less influential politically. (2) Consequently, a regulator, motivated by political considerations, will increase the stringency of environmental regulations. Of course, dynamic comparative advantages dictate that mature, developed economies shift resources away from basic manufacturing activities.

In the next section, we set out a simple model and derive some results that highlight the relationship between the sectoral decline of manufacturing and the stringency of environmental policies. In Section III, we present different types of empirical evidence to test the key findings of our model. First, we show that deindustrialization may have a "silver lining" in terms of reducing emissions from basic manufacturing activities. Specifically, we show that organic water pollution and industrial employment levels are close complements. Second, we investigate whether labor market institutions that have traditionally supported blue-collar interests and lowered the inequality of earnings affect the environmental regulation of industry. (3) We show that a greater degree of union coordination of wage bargaining is strongly linked to the observed pattern of environmental taxation of industry relative to households. We conclude Section III with a careful econometric study of panel data. In particular, we use extreme bounds analysis (EBA) to examine whether countries with greater income inequality and declining manufacturing employment have more stringent environment policies. The last section concludes.

II. THE MODEL

Consider an economy with two types of jobs: "blue collar" and "white collar," say. Further, assume that pollution creates blue-collar jobs (e.g., manufacturing) only. (4) All other jobs are white collar (e.g., services, high tech). Pollution afflicts all workers, however. A policymaker must reconcile the conflict between blue-collar jobs and environmental quality while also seeking the support of both groups of workers.

To make matters transparent, assume that the economy has a unit mass of each of two types of workers--blue-collar workers, indexed by b, and white-collar workers, indexed by w. For expositional purposes, we assume that white-collar workers are always employed. Blue-collar workers can be in one of two states at time t, employment (e) or unemployment (u). Workers receive income [y.sup.i.e.sub.t] if they work, i = b, w. If unemployed, blue-collar workers, receive income [y.sup.b,u.sub.t]. We assume that the {[y.sup.i,j.sub.t]} are deterministic processes beyond the decision maker's control. (5) If worker i supplies one unit of labor inelastically each period, [y.sup.i.e.sub.t] can be interpreted as the wage rate in period t for worker i.

At time t, manufacturing generates a residual called pollution, [s.sub.t]. Pollution and blue-collar labor are complementary inputs (see Cropper and Oates 1992). If the policymaker wants industry to create more blue-collar jobs, he must allow greater production--and pollution. The demand for blue-collar workers is given by [l.sup.b.sub.t] = f([y.sup.b,e.sub.t], [s.sub.t]), with

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The pollution stock, [p.sub.t], decays at rate [delta] [member of] [0, 1]. The transition equation is

[p.sub.t+1] = (1 - [delta])[p.sub.t] + [s.sub.t].

Equating [delta] to 1 gives the classic case of a pollutant that dissipates immediately; [delta] = 0 is the case for a pollutant that never dissipates. The utility for worker i is given by the concave function U([y.sup.i.sub.t], [p.sub.t]), for i = b, w. That is, all workers suffer from [p.sub.t].

In traditional political economy models, it is assumed that the policymaker maximizes a weighted average of the welfare of constituents over his career. The policymaker might be a politician who considers voter welfare to win elections, or he might be a regulator who considers constituent welfare to win promotions. In the current context, the political weights that a policymaker assigns to the welfare of blue- and white-collar workers may reflect the relative political influence of the two types of workers. Different weights may be attributed to interest groups according to the degree of organization or unionization or may simply vary with size of membership, for instance.

The common agency model developed by Bernheim and Whinston (1986), and applied by Grossman and Helpman (1994), provides microeconomic foundations for the political weights that are assigned to each interest group in a society. Grossman and Helpman showed that if policymakers, when choosing a policy (s, say) care about interest groups' welfare ([V.sup.j](s)) on one hand and about campaign contributions on the other hand, then they actually end up maximizing a weighted sum of the interest groups' objective functions. That is, the policymaker will choose a policy s to maximize

(1) [V.sup.g](s) = [[summation over (j[member of]I)]([I.sup.j] + [[alpha].sup.j])[V.sup.j](s),

where [V.sup.g](s) denotes the policymaker's welfare function, I is the set of all interest groups, the indicator function, [I.sup.j] equals 1 if the interest group is engaged in lobbying activities, and [I.sup.j] = 0 otherwise. (6)

Grossman and Helpman concentrated on studying the distortionary effects of lobbying and assumed that each group is originally given the same weight [[alpha].sup.j] = [alpha], j [member of] I. From Equation (1), it is clear that despite the presence of lobbying, the outcome will be equal to the efficient solution selected by the utilitarian social planner that would assign equal weights to everybody. The political system creates inefficiencies when some groups in the economy do not lobby. Naturally, the policymaker more heavily weights the policy preferences of the interest groups that do lobby (see Aidt 1998; Potters and Sloof 1996). In the spirit of Grossman and Helpman, we analyze cases in which both or either of the blue- and white-collar workers may organize lobbies to help attain their preferred environmental policy settings, perhaps via a trade union and an environmental lobby, respectively.

The welfare of all workers is assumed to be adversely affected by greater pollution-intensive production. However, blue-collar workers also benefit from higher pollution. While white-collar workers always prefer smaller production and less pollution, the interests and the lobbying stance of blue-collar workers depend on the elasticity of employment with respect to pollution emissions as well as the reservation utility if they were to be unemployed. Consequently, the relative position of the two interest groups is generally antithetical. (7)

Returning to the problem at hand, the decision maker's problem is

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

subject to [p.sub.t+1] = (1 - [delta])[p.sub.t] + [s.sub.t], [p.sub.o] given, where [rho] > 0 is the rate of time preference and [W.sup.i.sub.t] = [l.sup.i.sub.t]U([y.sup.i,e.sub.t],[p.sub.t]) + (1 - [l.sup.i.sub.t])U([y.sup.i,u.sub.t], [p.sub.t]), for i = b, w. (8)

Letting [p.sub.t] be the state and [s.sub.t] be the control, Bellman's equation is

(3) [V.sub.t]([p.sub.t]) = max {[[theta].sup.b.sub.t][W.sup.b.sub.t] + [[theta].sup.w.sub.t][W.sup.w.sub.t] + [beta][V.sub.t+1]([p.sub.t+1])},

where [beta] = [(1 + [rho]).sup.-1] and [[theta].sup.i] [member of] {[alpha], {1 + [alpha])}, i = b, w.

Standard solution techniques yield the Euler equation (see Appendix for details)

(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [[DELTA].sub.t] = U([y.sup.b,e.sub.t], [p.sub.t]) - U([y.sup.b,u.sub.t], [p.sub.t]).

In general, it is difficult to find closed-form solutions for the optimal [s.sub.t] (or [p.sub.t+1]) sequence. However, a simple perfect foresight example does illustrate some of the fundamental driving forces. For example, consider U([y.sup.i.sub.t], [p.sub.t]) = [y.sup.i.sub.t] - [gamma][p.sub.t], [gamma] > 0, i = b, w, and f([y.sup.b,e.sub.t], [s.sub.t])= ln(1 + [s.sub.t]) - [kappa] ln [y.sup.b,e.sub.t], [kappa] > 0. By appropriate substitutions into Equation (4) and solving the difference equation, we obtain

(5) [s.sub.t] = [((1 - (1 - [delta])[beta])[x.sub.t])/([beta][gamma](1 + [psi]))] - 1,

where [x.sub.t] = [y.sup.b,e.sub.t] - [y.sup.b,u.sub.t] and [psi] = ([[theta].sup.w]/[[theta].sup.b]).

To highlight the effects of changing real wages for blue-collar employees, we suppose that [x.sub.t] = [[omega]g.sup.-t], where g > (<) 1 if wages are shrinking (growing) exponentially. (9) From Equation (5), it follows that

(6) [s.sub.t] = [(([rho] + [delta])[omega])/([gamma](1 + [psi])[g.sup.t])] - 1.

The comparative dynamic effects are summarized in the following proposition.

PROPOSITION 1. Suppose that blue-collar income is given by [x.sub.t] = [[omega]g.sup.-t], then the stringency of environmental regulations falls in

(a) blue-collar income, [omega];

(b) the policymaker's discount rate, [rho];

(c) the pollution decay rate, [delta].

Environmental regulations are stricter

(d) the higher is the rate of diminution of blue-collar income, g;

(e) the higher is the marginal disutility of pollution, [gamma];

(f) when blue-collar workers do not lobby the policymaker.

Proof Differentiating Equation (6) yields Parts (a)-(e). As for Part (f), note that if only blue-collar workers lobby, then [[psi].sup.b] = [alpha]/(1 + [alpha]), if only white-collar workers lobby, then [[psi].sup.w] = (1 + [alpha])/[alpha], and if both groups lobby, then [[psi].sup.b,w] = 1. Clearly, [[psi].sup.b] < [[psi].sup.b,w] < [[psi].sup.w]. Finally, note that environmental quality improves in [psi].

There are some transparent implications. For example, if the policymaker discounts the future more heavily, then this is associated with deteriorating environmental quality. Congleton (1992) showed that autocratic countries are inclined to select less stringent environmental regulations. He argued that dictators tend to have shorter time horizons (i.e., higher [rho]) and are less likely to adopt pro-environment policies, since the benefits of doing so are likely to accrue only after they have left office, whereas the costs are incurred earlier. (10) A higher pollution decay rate, [delta], also increases pollution. This somewhat counterintuitive result occurs because worker utility depends on [p.sub.t], and not on [p.sub.t+1], so that the policymaker is likely to take a less conservative attitude with a pollutant that dissipates immediately, as opposed to the case for a pollutant that never dissipates. Consistent with this finding is that policymakers are more likely to be "policy-active" for the types of pollutants with short-term and local impacts (see Emerson and Pendleton 2004).

Higher income for workers in pollution-producing industries ([omega]) is associated with an increasing amount of economic importance attached to the polluting sector of the economy. However, if this economic "weight" falls, then because environmental quality is a normal good, the stringency of environmental regulations rises over time and, consequently, so too does the quality of the environment. Strictly speaking, it is not just the continued erosion of the earnings of blue-collar workers that beneficially impacts pollution emissions. More generally, it is the falling relative earnings of working in the polluting sector. For example, if the income while unemployed increases more rapidly than the income while employed in the polluting sector, then the same benefit to the environment results. Thus, some authors have argued that more generous unemployment benefits and changes to cash transfer and income tax systems have arisen to ensure worker acquiescence to potentially disruptive microeconomic reforms, such as trade liberalization (e.g., Rodrik 1998). Hence, while the earned income distribution may have widened in many OECD countries, the same is not true for the posttax and post-government transfer distribution of income (see, e.g., Smeeding and Gottschalk 1995).

Recall that when each interest group receives an equal political weight that this is equivalent to the utilitarian social planner's problem. However, if a lower political weight is attached to blue-collar worker interests ([psi]), then less importance is attached to the polluting sectors of the economy. As noted above, the increased likelihood of free riding in larger political constituencies poses problems for a straightforward interpretation of the political weights attached to interest groups by policymakers. In democratic countries, government officials may favor groups with more members. Larger groups are also likely to have greater electoral resources. However, groups with more members tend to be prone to free-riding problems. In addition, larger groups are likely to be costlier to organize, more difficult to develop a coherent and consistent platform for, and to involve greater difficulties in ensuring the political participation of all members. Potters and Sloof (1996) provided a fairly comprehensive survey of the empirical effects of group size on political outcomes. Overall, they concluded that free riding is, in fact, a serious problem for larger, unorganized groups. On the other hand, larger, organized groups, such as trade unions, for example, wield greater influence. In deindustrializing economies, the reduction in blue-collar power has in part been manifested by the declining influence of trade unionism (see Freeman 1993). Clearly, deunionization is likely to reinforce the declining political significance attached to the blue-collar interests in relaxing environmental standards. (11)

III. EMPIRICAL IMPLICATIONS AND EVIDENCE

In this section, we present evidence to examine the predictions of the model. More informally, in the next two subsections, we present some simple tabulations and correlations. First, we show that industrial employment and polluting activity go hand in hand. Second, we show that more unionized economies, which also tend to have more equitable earnings distributions, favor the imposition of eco-taxes on consumers rather than on industry. Finally, and more importantly, we present a formal econometric analysis using panel data to investigate the determinants of the stringency of environmental policy.

A. Deindustrialization and the Environment

A key feature of the model is that deindustrialization is associated with a cleaner environment. This occurs for two reasons--an economic one and a political one. The economic reason involves the trade-off between a cleaner environment and the production of basic manufacturing goods. In turn, higher employment in this sector is associated with greater pollution emissions.

Political pressures also imply a positive correlation between manufacturing employment levels and pollution emissions. For example, if underlying economic growth and dynamic comparative advantages reduce production and employment in basic manufacturing activities, then the remaining workers in these sectors are likely to receive smaller consideration in the political process. Consequently, policymakers weight more heavily the preferences of workers (and voters) involved in the production of "cleaner" goods. Doing so, of course, simply reinforces the decline in basic manufacturing industry.

To provide a visual perspective of the relationship between industrial employment and pollution emissions, we present plots for organic water pollutants and industrial employment for seven countries in Figure 1. The data are from a study of industrial emissions for a limited number of countries by Hettige, Mani, and Wheeler (1998). Data on water pollution are more readily available than other emissions data because most industrial pollution control programs start by regulating organic water emissions. Such data are also fairly reliable because sampling techniques for measuring water pollution are more widely understood and much less expensive than those for air pollution. The emissions estimates represent biochemical oxygen demand (BOD) in kilograms per day for each country and year. (12) The employment data are from the United Nations Industrial Development Organisation. The data series for each country is for 1975-1992 (or 1993). For ease of comparison, the emissions and employment data were converted to indices (with 1975 as the base year).

[FIGURE 1 OMITTED]

A couple of observations are immediate. First, changes in industrial employment and water pollution are strongly complementary. For instance, Canada suffered very steep reductions in its manufacturing employment in the late 1980s and early 1990s due to an unexpectedly severe recession and to the, somewhat more debatable, effects of the passage of the Canada-U.S. Free Trade Agreement (see Gaston and Trefler 1997). In countries in which manufacturing employment has grown rapidly (e.g., Singapore), there has been a corresponding increase in pollution emissions. (13)

In those countries that have experienced the most marked deindustrialization (see Baldwin and Martin 1999), for example, the United Kingdom, pollution emissions have steadily fallen as employment in manufacturing industry has declined over the entire time period. In the relatively few developed countries that have not deindustrialized as rapidly, emissions have remained relatively unchanged. (14)

B. Unions and the Environment

Next, we examine a subsidiary implication of the model developed in Section II. Specifically, as the political institutions that have traditionally supported blue-collar interests have declined in importance, associated environmental regulations have toughened.

Portney (1982) argued that increasing unemployment pressures policymakers to ease environmental standards. (15) By implication, political pressure brought to bear on environmental policymakers is greater when industrial employment levels fall. Fredriksson and Gaston (1999) have noted that unions lobbying on behalf of unemployed members may encourage policymakers to respond favorably to calls for easing environmental restrictions. Yandle (1983) found that state expenditures on environmental regulation in the United States were negatively related to the number of workers in polluting industries and positively related to the percentage of the manufacturing industry workforce that was unionized. He interpreted the former relationship as evidence that policymakers operate according to an environmental quality versus jobs trade-off and the latter relationship as evidence of union rent seeking. (16)

Overall, one expects that the decline of unionization in many countries has helped the passage of more stringent environmental regulations that affect industry. On the surface, the evidence on this point is somewhat mixed. According to Tobey's (1990) indices of environmental stringency, two of the three countries with the strictest environmental standards (the United States and Japan) have among the lowest rates of union membership in the world (as well as the lowest percentage of workers covered by collective bargaining agreements). However, the third, Sweden, has among the highest rates of unionization in the world. Fredriksson and Gaston (1999) explained this phenomenon by noting that the ambiguous stance of the trade union movement on environmental policies depends on the exposure to unemployment of their own members. It needs to be emphasized that it is the actual level of industrial employment, rather than the rates of unionization of a presumably smaller pool of manufacturing workers in deindustrializing economies, that may be of greater significance for policymakers.

In this paper, special interest groups, representing blue-collar and white-collar interests, help to determine the stringency of environmental policy. In most countries, trade unions are the most visible advocates of blue-collar interests. (17) If blue-collar workers perceive a trade-off between environmental regulations and jobs, unions are likely to oppose policies that threaten manufacturing employment.

Many OECD countries have recently introduced, or are considering implementing, fiscal instruments or "eco-taxes" for environmental management. Consider Table 1, which illustrates a specific example of an "eco-tax." The data in Columns (1) and (2) contain data on tax rates for household use and industrial use fuel for a number of OECD countries. A number of features are apparent. For example, the household use tax rate is highest in France and the industrial use tax rate is highest in Switzerland and both tax rates are lowest for the United States. The differences in tax rates across countries reflect a number of influences, including such disparate factors as the political importance of community environmental concerns as well as fiscal considerations.

Of more interest is the difference between the rates of taxation for industrial and household use. Column (3) indicates that the tax rate on industrial use fuel is 61% of the tax rate on household use fuel in Denmark; in the United States, the industrial use tax rate is 15% higher than the household use tax rate. Once again, there are likely to be a number of determinants of these cross-national differences. However, these differences are also likely to reflect the importance of industry concerns (i.e., shareholders and workers). Environmental and community concerns are likely to be reflected in the tax rate levels. On the other hand, national differences in the relative tax rates for industrial and household fuels are likely to reflect the relative influence of industry vis-a-vis households in the political process in which tax rates are determined. Moreover, institutional features of the labor market are important determinants of industry and union lobbying incentives and, consequently, the observed pattern of environmental policy.

Consider the last column of Table 1--"Union Coordination Index." Layard, Nickell, and Jackman (1994, 80-81) argued that when unions have a national focus (designated by an index of "3"), they take into account the common interests of the workforce in full employment "rather than bargaining as atomistic groups of insiders" (designated by an index of "1"). The data reveal that bargaining at the national level is negatively related to the tax rate disparity (the "Ratio" column). Of course, this correlation may be purely coincidental. On the other hand, it appears that a strong coordinated union movement is associated with relatively higher tax burdens on households (i.e., which comprise blue-collar and white-collar workers) as opposed to industry (which primarily employ the blue-collar workers). Overall, unionization does appear to be strongly linked to the observed pattern of environmental taxation of industry relative to households.

C. Inequality, Industrial Employment, and Environmental Regulation

To conclude the empirical section, we present a formal econometric analysis of the determinants of environmental policy. As much as possible, we follow the empirical specifications previously suggested in the literature. The major innovation, of course, is the introduction of variables suggested by our own model. Another major step forward is our use of EBA to isolate the most important determinants of environmental regulation.

Data and Variables. To proxy environmental stringency, we use the lead content of gasoline. This measure has been used in previous research (e.g., Damania, Fredriksson, and List 2003); its major advantage is that the data are available as a panel for the period from 1982 to 1992 and for up to 48 countries. We transform the series by taking the logarithm of it and multiplying it by -1. It is denoted by LREGS. (18)

Our primary focus is examining the importance of blue-collar workers for the political process that shapes the environmental policy. Damania, Fredriksson, and List (2003) used the percentage of the labor force employed in industry (INDSHEMP) to proxy political pressure by industrial workers. (19) This pressure is also central to our model's predictions. Since environmental regulations may increase employment uncertainty, industrial workers use their political power to prevent stricter regulations. The other variable important in our model is wage inequality. The stringency of environmental regulations is predicted to increase as blue-collar income declines. If wages fall exogenously (e.g., due to skilled labor-biased or sector-biased technological change that favors white-collar workers), then we predict a more stringent environmental policy (i.e., a lower lead content of gasoline).

In a recent paper, McAusland (2003) argued that greener pollution policies could be associated with either greater or smaller income inequality. In earlier research, inequality has often been associated with an intensification of polluting activities (e.g., Boyce 1994; Torras and Boyce 1998). Our model's predictions point in precisely the opposite direction. That is, as an economy deindustrializes, income inequality may increase as the wages paid to manufacturing workers in low-tech, pollution-intensive industries fall. As it does so, the influence of blue-collar workers in the policy-making process declines. (20) To measure income inequality, we use the Gini coefficient data recently updated and recalculated by Francois and Rojas-Romagosa (2005). (21)

Needless to say, a large number of other variables have been proposed as determinants of the level of environmental stringency. Cole, Elliot, and Fredriksson (2006) used the urban population share (URBAN) to test whether a greater exposure to industrial pollution by a larger number of citizens increases environmental stringency. Cole, Elliot, and Fredriksson (2006) also argued that the demand for environmental quality increases with per capita income (LGDPPC). On the other hand, Congleton (1992) emphasized that the effect of per capita income is theoretically indeterminate (even though he estimated a positive relationship in his study). He argued that despite the fact that the demand for environmental quality is likely to be increasing with personal wealth, voters and taxpayers also have to bear a higher share of the costs associated with environmental regulations. These costs reduce national income. A similar ambiguity is predicted for the effect of population density (LPOPDENS). Congleton argued that population also serves as a proxy for a country's human capital resources.

Damania, Fredriksson, and List (2003) contended that more open economies will have higher environmental standards. McAusland (2003) showed that trade openness and the pattern of factor ownership are important determinants of the preference for pollution standards. If an economy is small and open, then environmental policies have no effects on the terms of trade. Hence, if the poor have a larger relative stake in the production of dirty goods, then they may vote for weaker policies when the economy is open because there are no beneficial terms of trade effects associated with environmental regulations. We therefore use trade intensity (TRADE), measured by the ratio of trade flows to gross domestic product (GDP). Another commonly used openness measure is foreign direct investment, which we measure as the ratio of the net inflows of FDI to GDP (FDIGDP). (22) As a final proxy for openness, we employ the KOF Index of Globalization (GLOBAL) (see Dreher 2006). (23)

Congleton (1992) argued that autocratic countries have lower environmental standards because their rulers have shorter time horizons. Consequently, the incentives to invest in environmental protection are lower. Following Congleton, we also include a dictatorship dummy (DICT), which takes the value 1 if the executive index of electoral competitiveness is smaller than 3 (see Beck et al. 1999). In addition, we employ POLFREE, which we measure as the average of the Freedom House (2005) indices for civil liberties and political rights. Another included variable is LEFT, which measures whether the chief executive has a left-wing orientation or not. (24) Neumayer (2003) argued that a left-wing executive is traditionally more likely to care for the interests of blue-collar workers. As they work mostly in dirty sectors, this may reduce environmental stringency (see also Fredriksson and Gaston 1999). However, Neumayer noted that left-wing governments might also be more amenable to policies that protect the environment. (25)

Damania, Fredriksson, and List (2003), Fredriksson, List, and Millimet (2003), and Fredriksson and Svensson (2003) emphasized the role that corruption might play in affecting the political agenda. Accordingly, we include CORRUPT to measure the level of government corruption. This variable is the "Government Honesty" variable reported by the International Country Risk Grade. (26) For a summary of all variables, their sources, their descriptions, as well as the study that originally proposed them, see Table 2; Table 3 gives the descriptive statistics and correlations of the variables.

Extreme Bounds Analysis. Since there are several studies that investigate the effects on environmental stringency, there is a long list of potential explanatory variables. Studies often restrict their analysis to certain subsets of these variables and often ignore the effects of any omitted variable bias when other variables are not included. In addition to any model uncertainty, the limited number of observations often restricts the power of statistical tests that rule out irrelevant explanatory variables.

In order to address these issues, we use EBA, as proposed by Learner (1983) and Levine and Renelt (1992). EBA enables us to examine which explanatory variables are robustly related to our stringency measure and is a relatively neutral way of coping with the problem of selecting variables for an empirical model in situations where there are conflicting or inconclusive suggestions in the literature.

To conduct an EBA, equations of the following general form are estimated:

(7) Y = [[beta].sub.M]M + [[beta].sub.F]F + [[beta].sub.Z]Z + [upsilon],

where Y is the dependent variable, M is a vector of "commonly accepted" explanatory variables, and F is a vector containing the variables of interest. The vector Z contains up to three possible additional explanatory variables (as in Levine and Renelt 1992) which, according to the broader literature, are related to the dependent variable. The error term is [upsilon]. The EBA for a variable in [F.sub.k] states that if the lower extreme bound for [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]--that is, the lowest value for [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] minus two standard deviations--is negative, while the upper extreme bound for [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]--that is, the highest value for [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] plus two standard deviations--is positive, the variable [F.sub.k] is not robustly related to Y.

Sala-i-Martin (1997) argued that this testing criterion is far too strong for any variable to ever pass it. If the distribution of the parameter of interest has both positive and negative support, then a researcher is bound to find at least one regression model for which the estimated coefficient changes sign if enough regressions are run. Consequently, in what follows, we not only report the extreme bounds but also report the percentage of the regressions in which the coefficient of the variable [F.sub.k] is statistically different from zero. Moreover, instead of only analyzing the extreme bounds of the estimates of the coefficient of a particular variable, we follow Sala-i-Martin's (1997) recommended procedure and analyze the entire distribution. Accordingly, we also report the unweighted parameter estimate of [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] and its standard error, as well as the unweighted cumulative distribution function, CDF(0). The latter represents the proportion of the cumulative distribution function lying on each side of zero. CDF(0) indicates the larger of the areas under the density function either above or below zero, that is, whether this happens to be CDF(0) or 1--CDF(0). So CDF(0) always lies between 0.5 and 1.0. However, in contrast to Sala-i-Martin, we use the unweighted, instead of the weighted, CDF(0). (27)

Another objection to EBA is that the initial partition of variables in the M and in the Z vectors is likely to be arbitrary. However, as pointed out by Temple (2000), there is no reason why standard model selection procedures (such as testing down from a general specification) cannot be used in advance to identify variables that are particularly relevant. Furthermore, some variables are included in the large majority of studies and are by now common in this branch of the literature.

In our view, the inclusion of LGDPPC in the M vector is the only noncontentious inclusion as a regressor. In the literature on the environmental Kuznets curve, the relationship between GDP per capita and environmental quality has been widely discussed. Therefore, this variable may also play an important role in determining the stringency of environmental policy. While it is tempting to include our central variables (INDSHEMP and INEQUAL) in the M matrix, we are conscious of not prejudging the importance of our model and the outcome of the EBA.

The Results. As a preliminary to the EBA, we ran a first regression using LGDPPC as well as our central variables and conducted specification tests to test whether we have to correct for country- and/or time-specific effects in our panel setup. As a result of these tests, we include random country effects in all equations. (28)

Table 4 depicts the results of the EBA. (29) The criterion for considering a variable to be robustly related to stringency is the CDF(0) value. Sala-i-Martin (1997) suggested considering a variable to be robust if the CDF(0) criterion is greater than 0.90. Instead, we follow Sturm and de Haan's (2005) proposal to use a stricter threshold value of 0.95 due to the two-sided nature of the test.

Turning to the results of the EBA, we see that real GDP per capita (LGDPPC) is robustly and positively linked to the level of stringency. (30) This result is also found in the existing empirical literature. Therefore, to some extent, it resolves the potential theoretical ambiguity which Congleton (1992) highlighted. (31)

We now turn to the extended model. Here, each of the variables takes the role of the F vector once with the other 11 variables used in 175 combinations to test the robustness of this particular variable. The variable LEFT, representing a left-wing chief executive, is usually considered to be negatively related to stringency. A left-wing executive traditionally cares for the interest of industrial workers and might therefore be reluctant to increase environmental stringency. Our empirical results support this view. The share of the urbanized population (URBAN) has a negative relationship with stringency. Hence, citizens living in urban areas tolerate lower levels of environmental stringency.

From the viewpoint of our simple model, the most important findings of our analysis are the results for INDSHEMP and INEQUAL. (32) According to the EBA, the former variable is robustly negatively related to the measure of stringency. (33) Therefore, just as our theory suggests, a declining blue-collar labor force is associated with diminished blue-collar political power and leads to more stringent environmental regulations. Our result for inequality is the more novel finding, however. Greater dispersion in incomes is associated with a more, and not less, stringent environmental policy. While this finding stands in stark contrast to previous research, it is consistent with our model. All other variables that are proposed in the literature, as being an influencing factor for the stringency of environmental agenda setting, clearly fail to meet the robustness criterion of a CDF(0) value above 0.95. This is another major finding of the EBA although obviously a rather negative one with respect to the extant literature.

[FIGURE 2 OMITTED]

In order to evaluate the relevance of the variables, we also estimate the magnitude of the impact that all variables have on the policy stringency measure. We do this by calculating the effect that a shock of one standard deviation of each variable has on LREGS. We therefore multiply the average EBA coefficient with the standard deviation of the respective variable and rank them in descending order according to absolute value. (34) The resulting ranking is included in Table 4 in the column "Impact Rank." The five variables that exhibit a robust relation with the measure of stringency are among the six variables that have the biggest impact on the dependent variable. In addition, we report the histograms of the coefficients of our two central variables. Figure 2 reveals that the estimated coefficients of the key variables are distributed close to their respective means and that there are no major outliers.

Concerning the robustness of our results, we use the five variables that the EBA suggests are robustly linked to environmental stringency and estimate our final model. The results are shown in Table 5.

Based on purely statistical criteria, it is the preferred model. Again, the specification tests lead us to include country-specific random effects. (35) However, we also present the results when adding time-specific effects as a further robustness check. (36) Potentially worrisome is the relatively small number of observations (due to the list-wise deletion of missing observations on key variables). In order to test whether our results are driven by the small sample, we linearly interpolate our inequality measure to create more observations. The results of these estimations are summarized in Table 6. Except for URBAN, in the case of country- and time-specific fixed effects, all variables are robust to the estimation technique, the inclusion of time effects, and the sample size, that is, they are statistically significant at conventional levels. Overall, the results of the EBA are reinforced.

To summarize, there is obviously strong support for the argument forwarded in this paper; namely, that declining economic significance is associated with a decline in political significance. Both of these factors reinforce one another and lead to more stringent environmental regulation.

IV. CONCLUSIONS

Our paper has emphasized that political and economic considerations interact to help explain the observed relationship between measures of economic development and environmental quality. Deindustrialization, falling real incomes of production workers, and a greater dispersion of income are increasingly prominent features in many industrialized countries. From a political economy perspective, such features can also explain observed environmental policies.

When the social and economic consequences of either high unemployment or falling incomes in manufacturing industries are high, policymakers may be tempted to ease environmental regulations. Symmetrically, as deindustrialization proceeds, as reflected by declining industrial employment and the falling wages and incomes for workers in basic manufacturing and pollution-intensive industries, environmental stringency increases. That is, as these sectors of the economy become less important economically, they are also likely to carry less weight politically. Consequently, a regulator optimally increases the stringency of environmental regulations. The argument is simple and straightforward. Dynamic comparative advantages dictate that mature, developed economies shift resources away from basic manufacturing activities. Environmental policy simply reinforces this movement.

To some readers, the argument developed in this paper may seem overly optimistic from the point of view of the environment and overly cynical from a social equity perspective. The risk of being overcynical is particularly acute for those who believe that a sense of social justice should prevail during times of rapid deindustrialization and falling blue-collar worker incomes. In turn, the social and political pressures may be thought to help override the demand for increased regulatory stringency. If this were in fact the case, it would be expected that environmental policies are least stringent in those industrialized and democratic countries in which income inequality is greatest. The evidence presented in the paper is consistent with the exact opposite view. That is, countries with the strictest environmental standards tend to be those with the greatest dispersion in their incomes.

ABBREVIATIONS

BOD: Biochemical Oxygen Demand

EBA: Extreme Bounds Analysis

GDP: Gross Domestic Product

OECD: Organisation for Economic Co-operation and Development

APPENDIX: DERIVATION OF THE EULER EQUATION

The first-order condition for the maximization of Bellman's equation is

[[theta].sup.b]([partial derivative][W.sup.b.sub.t])/[partial derivative][s.sub.t]) + [[theta].sup.w]([partial derivative][W.sup.w.sub.t]/[partial derivative][s.sub.t]) + [beta][V'.sub.t+1]([p.sub.t+1]) = 0.

Rearranging and simplifying, we have

(A1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where [[DELTA].sub.t] = U([y.sup.b,e.sub.t], [p.sub.t]) - U([y.sup.b,u.sub.t], [p.sub.t]). Differentiating the value function yields

[V'.sub.t]([p.sub.t]) = [[theta].sup.b]([partial derivative][W.sup.b.sub.t]/[partial derivative][p.sub.t]) + [[theta].sup.w]([partial derivative][W.sup.w.sub.t]/[partial derivative][p.sub.t]) + (1 - [delta])[[beta][V'.sub.t+1]([p.sub.t+1]).

After simplifying, we have

(A2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Substituting Equation (A2) into Equation (A1) yields

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

or

(A3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Substituting Equation (A3) into Equation (A1) yields the Euler equation (i.e., Equation 4 in the text).

MARTIN GASSEBNER, NOEL GASTON and MICHAEL J. LAMLA *

* We would like to thank Rodney Beard, Per Fredriksson, Jeff Kline, Shravan Luckraz, Dan Sasaki, Jan-Egbert Sturm, Leon Taylor, and two anonymous referees as well as participants of the Annual Meeting of the German Economic Association (2006) and of the seminars at the University of Adelaide, Bond University, the University of Groningen, and the University of Tokyo for their helpful comments. Responsibility for errors and omissions rests with the authors.

REFERENCES

Aidt, T. S. "Political Internalization of Economic Externalities and Environmental Policy." Journal of Public Economics, 69, 1998, 1-16.

Alesina, A., and D. Rodrik. "Distributive Politics and Economic Growth." Quarterly Journal of Economics, 109, 1994, 465-90.

Arrow, K., B. Bolin. R. Costanza, P. Dasgupta, C. Folke, C. S. Holling, B.-O. Jansson, et al. "Economic Growth, Carrying Capacity, and the Environment." Ecological Economics, 15, 1995, 91-95.

Baldwin, R. E., and P. Martin. "Two Waves of Globalisation: Superficial Similarities, Fundamental Differences." NBER Working Paper No. 6904, 1999.

Barro, R. J. "Inequality, Growth, and Investment." NBER Working Paper No. 7038, 1999.

Beck, T., G. Clarke, A. Groff, and P. Keefer. The Database of Political Institutions. Washington, DC: World Bank Development Research Group, 1999.

Bernheim, B. D., and M. D. Whinston. "Menu Auctions, Resource Allocation, and Economic Influence." Quarterly Journal of Economics, 101, 1986, 1-31.

Bommer, R., and G. G. Schulze. "Environmental Improvement with Trade Liberalization." European Journal of Political Economy, 15, 1999, 639-61.

Boyce, J. K. "Inequality as a Cause of Environmental Degradation." Ecological Economics, 11, 1994, 169-78.

Cassing, J. H., and A. L. Hillman. "Political Influence Motives and the Choice between Tariffs and Quotas." Journal of International Economics, 19, 1985, 279-90.

Cole, M. A., R. J. R. Elliot, and P. G. Fredriksson. "Endogenous Pollution Havens: Does FDI Influence Environmental Regulations?" Scandinavian Journal of Economics, 108, 2006, 157-78.

Congleton, R. D. "Political Institutions and Pollution Control." Review of Economics and Statistics, 74, 1992, 412-21.

Cropper, M. L., and W. E. Oates. "Environmental Economics: A Survey." Journal of Economic Literature, 30, 1992, 675-740.

Damania, R., P. G. Fredriksson, and J. A. List. "Trade Liberalization, Corruption and Environmental Policy Formation: Theory and Evidence." Journal of Environmental Economics and Management, 46, 2003, 490-512.

Deininger, K., and L. Squire. "A New Data Set Measuring Income Inequality." World Bank Economic Review, 10, 1996, 565-91.

Dinda, S. "Environmental Kuznets Curve Hypothesis: A Survey." Ecological Economics, 49, 2004, 431-55.

Dreher, A. "Does Globalization Affect Growth? Evidence from a New Index of Globalization." Applied Economics, 38, 2006, 1091-110.

Emerson, T. N., and L. H. Pendleton. "Income, Environmental Disamenity, and Toxic Releases." Economic Inquiry, 42, 2004, 166-78.

Endersby, J. W., and M. C. Munger. "The Impact of Legislator Attributes on Union PAC Campaign Contributions." Journal of Labor Research, 13, 1992, 79-97.

Francois, J. F., and H. Rojas-Romagosa. "The Construction and Interpretation of Combined Cross-Section and Time-Series Inequality Datasets." World Bank Policy Research Working Paper No. 3748, 2005.

Fredriksson, P. G., and N. Gaston. "The 'Greening' of Trade Unions and the Demand for Eco-Taxes." European Journal of Political Economy, 15, 1999, 663-86.

--. "Environmental Governance in Federal Systems: The Effects of Capital Competition and Lobby Groups." Economic Inquiry, 38, 2000, 501-14.

Fredriksson, P. G., J. A. List, and D. L. Millimet. "Bureaucratic Corruption, Environmental Policy and Inbound US FDI: Theory and Evidence." Journal of Public Economics, 87, 2003, 1407-30.

Fredriksson, P. G., and J. Svensson. "Political Instability, Corruption and Policy Formation: The Case of Environmental Policy." Journal of Public Economics, 87, 2003, 1383-405.

Freedom House. Freedom in the Worm 2005: The Annual Survey of Political Rights and Civil Liberties. Washington DC: Rowman and Littlefield, 2005.

Freeman, R. B. "How Much Has De-Unionization Contributed to the Rise in Male Earnings Inequality?" in Uneven Tides: Rising Inequality in America, edited by S. Danziger and P. Gottschalk. New York: Russell Sage, 1993, 133-63.

--. "War of the Models: Which Labour Market Institutions for the 21st Century?" Labour Economics, 5, 1998, 1-24.

Gaston, N., and D. Trefler. "The Labour Market Consequences of the Canada-U.S. Free Trade Agreement." Canadian Journal of Economics, 30, 1997, 18-41.

Grossman, G. M., and E. Helpman. "Protection for Sale." American Economic Review, 84, 1994, 833-50.

Grossman, G. M., and A. B. Krueger. "Environmental Impacts of a North American Free Trade Agreement," in Mexican-U.S. Free Trade Agreement, edited by P. Garber. Cambridge, MA: MIT Press, 1993, 13-56.

--. "Economic Growth and the Environment." Quarterly Journal of Economics, 110, 1995, 353-77.

Hettige, H., M. Mani, and D. Wheeler. "Industrial Pollution in Economic Development: Kuznets Revisited." 1998. Available as "New Ideas in Pollution Regulation." http://www.worldbank.org/NIPR. [accessed June 1998]

Hillman, A. L. "Declining Industries and Political-Support Protectionist Motives." American Economic Review, 72, 1982, 1180-87.

Hilton, F., and A. Levinson. "Factoring the Environmental Kuznets Curve: Evidence from Automotive Lead Emissions." Journal of Environmental Economics and Management, 35, 1998, 126-41.

Johnson, G., and F. Stafford. "The Labor Market Implications of International Trade," in Handbook of Labor Economics, Vol. 3B, edited by O. Ashenfelter and D. Card. Amsterdam, Netherlands: Elsevier North-Holland, 1999, 2215-88.

Klepper, G. "The Political Economy of Trade and the Environment in Western Europe," in International Trade and the Environment. World Bank Discussion Papers No. 159, edited by P. Low. Washington DC: World Bank, 1992, 247-60.

Knack, S., and P. Keefer. "Institutions and Economic Performance: Cross-County Tests Using Alternative Institutional Measures." Economics and Politics, 7, 1995, 207-27.

Layard, R., S. Nickell, and R. Jackman. The Unemployment Crisis. Oxford, UK: Oxford University Press, 1994.

Learner, E. E. "Let's Take the Con Out of Econometrics." American Economic Review, 73, 1983, 31-43.

Levine, R., and D. Renelt. "A Sensitivity Analysis of Cross-County Growth Regressions." American Economic Review, 82, 1992, 942-63.

McAusland, C. "Voting for Pollution Policy: The Importance of Income Inequality and Openness to Trade." Journal of International Economics, 61, 2003, 425-51.

Neumayer, E. "Are Left-Wing Party Strength and Corporatism Good for the Environment? Evidence from Panel Analysis of Air Pollution in OECD Countries." Ecological Economics, 45, 2003, 203-20.

Octel (The Associated Octel Company Ltd.). Worldwide Gasoline Survey. London: Octel, 1982-1992.

Olson, M. "Dictatorship, Democracy, and Development." American Political Science Review, 87, 1993, 567-76.

Organisation for Economic Co-operation and Development. Environmental Taxes in OECD Countries. Paris: OECD, 1995.

--. Employment Outlook. Paris: OECD, 1997.

Persson, T., and G. Tabellini. "Is Inequality Harmful for Growth?" American Economic Review, 84, 1994, 600-21.

Portney, P. R. "How Not to Create a Job." Regulation, 6, 1982, 35-38.

Potters, J., and R. Sloof. "Interest Groups: A Survey of Empirical Models that Try to Assess their Influence." European Journal of Political Economy, 12, 1996, 403-42.

Rodrik, D. "Why Do More Open Economies Have Bigger Governments?" Journal of Political Economy, 106, 1998, 997-1032.

Rowthorn, R. E. "Centralisation, Employment and Wage Dispersion." Economic Journal, 102, 1992, 506-23.

Saint Paul, G., and T. Verdier. "Inequality, Redistribution and Growth: A Challenge to the Conventional Political Economy." European Economic Review, 40, 1996, 719-28.

Sala-i-Martin, X. "I Just Ran Two Million Regressions." American Economic Review, 87, 1997, 178-83.

Selden, T. M., and D. Song. "Environmental Quality and Development: Is There a Kuznets Curve for Air Pollution Emissions?" Journal of Environmental Economies and Management, 27, 1994, 147-62.

Shafik, N. "Economic Development and Environmental Quality: An Econometric Analysis." Oxford Economic Papers, 46, 1994, 757-73.

Smeeding, T., and P. Gottschalk. "The International Evidence on Income Distribution in Modern Economies: Where Do We Stand?" Luxembourg Income Study Working Paper No. 137, 1995.

Sturm, J.-E., and J. de Haan. "How Robust is Sala-i-Martin's Robustness Analysis?" mimeo, University of Konstanz, 2002.

--. "Determinants of Long-Term Growth: New Results Applying Robust Estimation and Extreme Bounds Analysis." Empirical Economies, 30, 2005, 597-617.

Taylor, L. "Do the States Try to Trade off Environmental Quality Tomorrow for Jobs Today?" Washington University--St. Louis Economics Working Paper No. 9810006, 1998.

Temple, J. "Growth Regressions and What the Textbooks Don't Tell You." Bulletin of Economic Research, 52, 2000, 597-617.

Tobey, J. A. "The Effects of Domestic Environmental Policies on Patterns of World Trade: An Empirical Test." Kyklos, 43, 1990, 191-209.

Torras, M., and J. K. Boyce. "Income, Inequality, and Pollution: A Reassessment of the Environmental Kuznets Curve." Ecological Economics, 25, 1998, 147-60.

World Bank. Worm Development Indicators. Washington DC: World Bank, 1999.

--. World Development Indicators. Washington DC: World Bank, 2003.

Yandle, B. "Economic Agents and the Level of Pollution Control." Public Choice, 40, 1983, 105-09.

Gassebner: Researcher, ETH Zurich, KOF Swiss Economic Institute, WEH D4, Weinbergstrasse 35, CH-8092 Zurich, Switzerland. E-mail gassebner@kof.ethz.ch

Gaston: Professor, Globalisation and Development Centre and School of Business, Bond University, Gold Coast, Queensland 4229, Australia. E-mail ngaston@bond.edu.au

Lamla: Researcher, ETH Zurich, KOF Swiss Economic Institute, WEH D4, Weinbergstrasse 35, CH-8092 Zurich, Switzerland. E-mail lamla@kof.ethz.ch

(1.) There are many excellent surveys of the enormous literature on international trade and labor market outcomes, for example, Johnson and Stafford (1999).

(2.) This implies that "economic power" and political power are both unequally distributed (see Barro 1999, 4).

(3.) As unionization has declined, there is some evidence that wage inequality has increased (e.g., Freeman 1998).

(4.) In many developed countries, the combination of advanced pollution abatement technologies, as well as the toxic waste generated from office-situated photocopiers, suggests that it may not be entirely appropriate to classify blue-collar work as polluting and white-collar work as not.

(5.) Apart from adding some largely irrelevant parameters to the model, allowing a dependence of income on emission flows adds very little.

(6.) If policymakers derive utility from a weighted sum of campaign contributions and aggregate social welfare, that is, [V.sup.G](s) = [[summation].sub.i = W, E, K] [[alpha][V.sup.i] ([alpha]) + [I.sup.i][C.sup.i](s)], by applying Lemma 2 in Bernheim and Whinston (1986), Grossman and Helpman (1994) showed that they actually end up maximizing a weighted sum of the interest groups' objective functions, that is, Equation (1). The application of the common agency framework by Grossman and Helpman to model the political decision-making process is not without its limitations. As a practical matter, political contributions by organized lobby groups are illegal in some countries. From a theoretical viewpoint, there is a two-sided moral hazard problem associated with either politicians reneging on their policy promises after contributions are received or lobbies reneging on promised contributions once preferred policies are locked in place. However, note that the objective described by Equation (1) is quite general and could be alternatively motivated by a linear, additive version of a political support function (Hillman 1982).

(7.) More specifically, since [y.sup.w] is deterministic and for a threshold [[bar.p].sup.w] [greater than or equal to] [p.sub.0], white-collar workers will always lobby when the stock of pollution reaches this threshold value. On the other hand, blue-collar workers lobby for laxer industry regulation until the marginal effect of higher pollution on the probability of being employed falls below the marginal disutility of greater pollution emissions.

(8.) By the law of large numbers [l.sup.p.sub.t] (1 - [l.sup.b.sub.t]) can be interpreted as the fraction of employed (unemployed) blue-collar workers. Recall that white-collar workers are always employed, that is, [l.sup.w.sub.t] = 1.

(9.) Alternatively, g may represent a direct measure of wage inequality between white- and blue-collar workers. For example, define [y.sup.w,e.sub.t]/[y.sup.b,e.sub.t] = [g.sub.t] and [y.sup.b,u.sub.t] = 0.

(10.) For reasons other than the expected shorter duration of dictatorships, Olson (1993) argued that dictators wish to maximize tax revenues and thus oppose any policies that would reduce revenue, for example, those that result from increased pollution abatement expenditures.

(11.) Fredriksson and Gaston (1999) showed that the stance taken by the trade union movement on environmental policies is far from unambiguous. Among other things, union "environmentalism" may depend on the risk of unemployment for their members as well as the presence of unemployed, nonunionized "outsiders."

(12.) Emissions of organic pollutants from industrial activities are a major source of water quality degradation. The Hettige, Mani, and Wheeler (1998) data are based on measurements of plant-level water pollution in a number of countries. The focus is on organic water pollution as indicated by the presence of organic matter, metals, minerals, sediment, bacteria, and toxic chemicals. The pollution is measured by BOD because it provides the most plentiful and reliable source of comparable cross-country emissions data. BOD measures the strength of an organic waste in terms of the amount of oxygen consumed in breaking it down. A sewage overload in natural waters exhausts the water's dissolved oxygen content. Wastewater treatment, by contrast, reduces BOD. (The previous discussion is drawn from Worm Development Indicators (World Bank 1999, 143).)

(13.) In the case of Ireland, the growing disparity between the two indices may reflect the rapid increase in the service sector industries associated with Ireland's "green tiger." We are grateful to an anonymous referee for this observation.

(14.) Baldwin and Martin (1999) noted that the "first wave" of globalization (pre-WW1), which generated rapid economic development for many countries, was characterized by rapid industrialization. In contrast, the "second wave" of globalization (since 1960), which generated rapid income growth for many developed countries, has been characterized by a process of deindustrialization and the associated steady decline in industrial employment.

(15.) In reference to the stance taken by European labor unions on environmental regulation, Klepper (1992, 253) noted that the primary objective of securing or increasing employment was thought to be threatened by environmental policies.

(16.) Endersby and Munger (1992) found that union contributions were given disproportionately to members of Congress who were members of committees with legislative and regulatory jurisdiction over activities that would affect labor.

(17.) Of significance, for the purpose of this paper at least, is that countries with encompassing labor market institutions (i.e., large unionized sectors with centralized bargaining) are characterized by lower wage inequality (see, e.g., Freeman 1993; Organisation for Economic Co-operation and Development 1997; Rowthorn 1992).

(18.) A rise in the transformed LREGS therefore represents a higher level of environmental stringency. Hilton and Levinson (1998) and Octel (The Associated Octel Company Ltd.) (1982-1992) provided a more detailed description of these data.

(19.) Unless stated otherwise, all data are taken from the World Development Indicators (World Bank 2003).

(20.) Consistent evidence is provided by Taylor (1998). He used data for State expenditures per capita in the United States for hazardous waste in the 1980s and for air pollution in the 1960s and rejected the hypothesis that there is a trade-off between future environmental quality and current manufacturing jobs.

(21.) These authors addressed measurement error problems in the well-known World Bank inequality data set of Deininger and Squire (1996) and produced a new data set of consistent inequality series.

(22.) Fredriksson and Gaston (2000) argued that greater capital mobility weakens the incentive for capital owners to lobby and, ceteris paribus, may lead to stricter environmental policies.

(23.) This index incorporates economic as well as the political and social dimensions of globalization.

(24.) This variable is taken from Beck et al. (1999).

(25.) Neumayer argued that blue-collar workers are likely to be among the first exposed to the effects of environmental degradation.

(26.) For details, see Knack and Keefer (1995).

(27.) Sala-i-Martin (1997) proposed using the integrated likelihood to construct a weighted CDF(0). However, missing observations for some of the variables poses a problem. Sturm and de Haan (2002) showed that the goodness-of-fit measure may not be a good indicator of the probability that a model is the true model and that the weights constructed in this way are not invariant to linear transformations of the dependent variable. Hence, changing scales could result in different outcomes and conclusions. We therefore employ the unweighted version.

(28.) For readability, the results of the specification test of the EBA are not shown. (However, they are available from the authors upon request.)

(29.) Since there are substantial differences in the number of observations for each variable, which potentially could influence our results, we opt to restrict our sample based on our inequality measure and hence ensure a more homogeneous sample.

(30.) This result is based on 231 regressions.

(31.) We also tested for a potential nonlinear relationship by including the squared term into the model. However, this is not supported by the data.

(32.) We also tested for potential nonlinear relationships by including the squares of both variables in the model. However, both added variables were insignificant.

(33.) In our regression model, the estimated coefficients represent semi-elasticities. Hence, a one-unit change in the level of both variables implies a beta percentage change in the level of the dependent variable. As INDSHEMP is measured in percentages, a 1% increase in the percentage employed in the industry sector leads to a beta percentage change in the dependent variable.

(34.) Since the estimation results include country-specific random effects, we de-mean all variables. Failing to do so could seriously bias the results since the country-specific effects that were already taken into account would again contribute to the result.

(35.) For comparison. Table 5 also contains the results for country-specific fixed effects.

(36.) We also estimated model specifications that included a time trend. There were no changes to the results.

TABLE 1
Total Taxes as Percent of End-User Price for Automotive
Fuels, 1994

 Household Industrial Ratio Coordination
Country Use (1) Use (2) (2) / (1) Index

Denmark 68.0 41.5 0.61 3
Sweden 76.5 48.3 0.63 3
Norway 67.3 46.0 0.68 3
Austria 63.9 49.1 0.77 3
Belgium 74.2 57.3 0.77 2
Netherlands 75.9 59.7 0.79 2
France 80.8 65.1 0.81 2
Portugal 73.5 59.4 0.81 2
Germany 76.9 62.5 0.81 2
Spain 68.6 56.9 0.83 2
Canada 50.0 41.6 0.83 1
Italy 76.1 65.1 0.86 2
United Kingdom 73.5 63.6 0.87 1
Switzerland 71.3 68.9 0.97 1
United States 34.4 39.6 1.15 1

Sources: Columns (1) and (2): Organisation for Economic
Co-operation and Development (1995, table 2, 48). Last
column: Layard, Nickell, and Jackman (1994, table 6, 78).
Values represent: 3 = high (national coordination); 2 =
inter-mediate; and 1 = low (firm level or uncoordinated).

TABLE 2 Variables--Definitions, Sources, and Hypotheses

Variable Source Description Sign Proposed By

LREGS Octel Log of lead Damania,
 (1982-1992) content of Fredriksson,
 gasoline, and List
 multiplied by (2003)
 (-1)
CORRUPT International "Government + Damania,
 Country Risk Honesty," Fredriksson,
 Guide higher values and List
 indicate less (2003)
 corruption
DICT Beck et al. Dummy variable - Congleton
 (1999) for (1992)
 dictatorship
 (executive
 index of
 electoral
 competitive-
 ness < 3)
FDIGDP World Bank Net inflows of + Cole, Elliot,
 (2003) foreign and
 direct Fredriksson
 investment (2006)
 (% of GDP)
GLOBAL Dreher (2006) KOF Index of + This paper
 Globalization
INDSHEMP World Bank Employment in - Damania,
 (2003) industry Fredriksson,
 sector (% of and List
 total (2003)
 employment)
INEQUAL Francois and Gini + This paper
 Rojas-Romagosa coefficient-
 (2005) household
 income
LEFT Beck et al. Dummy variable ? Neumayer
 (1999) for the chief (2003)
 executive's
 party being
 left wing
LGDPPC World Bank Log of GDP per ? Congleton
 (2003) capita (in (1992)
 constant 1995
 $US)
LPOPDENS World Bank Log of ? Congleton
 (2003) population (1992)
 per hectare
POLFREE FHI (2005) Average of + Congleton
 "civil (1992)
 liberties"
 and
 "political
 rights"

TRADE World Bank Trade intensity + Damania,
 (2003) ((imports + Fredriksson,
 exports)/GDP) and List
 (2003)
URBAN World Bank Urban + Damania,
 (2003) population Fredriksson,
 (% of total) and List
 (2003)

Notes: "Sign" refers to the expected sign of the variable
according to the literature "+/-" indicates a positive/negative
sign, while "?" represents an a priori indeterminate effect.
FHI, Freedom House.

TABLE 3
Variables-Descriptive Statistics and Correlation Matrix

 Standard
 Mean Deviation (1) (2)

 (1) LREGS -0.37 0.95 341 0.418
 (2) CORRUPT 3.49 1.55 328 328
 (3) DICT 0.22 0.42 341 328
 (4) FDIGDP 1.17 1.71 324 311
 (5) GLOBAL 1.91 0.74 310 310
 (6) INDSHEMP 26.18 8.03 194 194
 (7) INEQUAL 43.37 10.49 57 57
 (8) LEFT 0.36 0.48 332 319
 (9) LGDPPC 7.78 1.76 335 322
(10) LPOPDENS -0.45 1.36 330 319
(11) POLFREE 3.32 1.92 341 328
(12) TRADE 49.28 27.3 333 320
(13) URBAN 52.34 25.74 341 328

 (3) (4) (5) (6)

 (1) LREGS -0.225 -0.047 0.528 0.349
 (2) CORRUPT -0.360 0.190 0.769 0.533
 (3) DICT 341 -0.138 -0.357 -0.264
 (4) FDIGDP 324 324 0.250 0.127
 (5) GLOBAL 310 293 310 0.502
 (6) INDSHEMP 194 192 193 194
 (7) INEQUAL 57 57 57 52
 (8) LEFT 332 315 302 188
 (9) LGDPPC 335 324 304 194
(10) LPOPDENS 330 313 310 194
(11) POLFREE 341 324 310 194
(12) TRADE 333 322 302 194
(13) URBAN 341 324 310 194

 (7) (8) (9) (10)

 (1) LREGS -0.439 -0.110 0.490 0.285
 (2) CORRUPT -0.381 0.130 0.707 -0.208
 (3) DICT 0.471 -0.167 -0.537 -0.062
 (4) FDIGDP 0.087 0.068 0.131 -0.240
 (5) GLOBAL -0.582 0.053 0.865 -0.013
 (6) INDSHEMP -0.399 0.180 0.724 0.037
 (7) INEQUAL 57 -0.159 -0.428 -0.365
 (8) LEFT 55 332 -0.104 -0.193
 (9) LGDPPC 57 326 335 0.017
(10) LPOPDENS 57 322 324 330
(11) POLFREE 57 332 335 330
(12) TRADE 57 324 333 322
(13) URBAN 57 332 335 330

 (11) (12) (13)

 (1) LREGS -0.396 0.101 0.383
 (2) CORRUPT -0.540 0.340 0.608
 (3) DICT 0.635 -0.172 -0.372
 (4) FDIGDP -0.182 0.504 0.060
 (5) GLOBAL -0.706 0.438 0.771
 (6) INDSHEMP -0.527 0.096 0.565
 (7) INEQUAL 0.516 -0.429 -0.037
 (8) LEFT -0.040 -0.143 -0.212
 (9) LGDPPC -0.797 0.327 0.869
(10) LPOPDENS -0.110 0.187 -0.058
(11) POLFREE 341 -0.248 -0.670
(12) TRADE 333 333 0.235
(13) URBAN 341 333 341

Notes: The first two columns report the mean and the standard
deviation of each series; the upper right part of the remaining
table reports correlation coefficients, the main diagonal gives
the number of observations for each variable, while the lower
left shows the number of observations used to calculate the
correlation coefficients.

TABLE 4
EBA Results (Dependent Variable: LREGS)

 Lower Upper
Variable Bound Bound % Sign.

Base model
 LGDPPC -0.482 3.861 83.55
Extended model
 INDSHEMP -0.189 0.023 95.43
 LEFT -1.398 0.150 93.14
 URBAN -0.193 0.035 74.29
 INEQUAL -0.029 0.130 70.86
 LPOPDENS -1.115 3.579 34.29
 POLFREE -0.292 0.583 33.14
 FDIGDP -0.112 0.275 9.71
 GLOBAL -1.685 2.893 40.57
 TRADE -0.044 0.026 10.86
 DICT -1.272 1.157 14.29
 CORRUPT -0.648 0.438 2.86

 Unweighted Unweighted Standard Impact
Variable CDF(0) [beta] Error Rank

Base model
 LGDPPC 0.9878 1.513 0.445 3
Extended model
 INDSHEMP 0.9813 -0.079 0.034 4
 LEFT 0.9779 -0.626 0.285 1
 URBAN 0.9651 -0.069 0.035 2
 INEQUAL 0.9575 0.053 0.029 6
 LPOPDENS 0.9194 0.753 0.508 11
 POLFREE 0.8756 0.121 0.096 7
 FDIGDP 0.8625 0.077 0.065 8
 GLOBAL 0.8501 0.866 0.665 5
 TRADE 0.8066 -0.010 0.010 9
 DICT 0.7807 0.218 0.277 10
 CORRUPT 0.5707 -0.046 0.145 12

Notes: Results based on 231 (base model) and 175 (extended
model) regressions, respectively, using country-specific random
effects. "%Sign." refers to the percentage of the regressions
in which the respective variable is significant at the 10%
significance level. "Impact Rank" lists the variables in
descending order according to the impact of a one-standard
deviation shock. The standard deviation is calculated after
de-meaning each variable to correct for country-specific
effects. Variables are sorted according to the CDF(0)
criterion.

TABLE 5
Final Model (Dependent Variable: LREGS)

 Random Effects Fixed Effects

LGDPPC 3.652 *** (0.519) 4.363 *** (0.629)
INDSHEMP -0.098 *** (0.028) -0.132 *** (0.033)
LEFT -0.713 *** (0.232) -0.756 *** (0.253)
URBAN -0.107 *** (0.035) -0.080 * (0.049)
INEQUAL 0.062 *** (0.024) 0.063 ** (0.027)
Constant -24.221 *** (3.472) --
Observations 50 50
[R.sup.2] 0.944 0.925
Hausman test ([H.sub.0]: 2.863 --
 random effects,
 [H.sub.1]: fixed
 effects)
F test (significance of 4.573 *** --
 country-specific
 random effects)
LR test (Ho: pooled OLS, -- 107.7 ***
 [H.sub.1]:
 country-specific
 fixed effects)

Notes: Both regressions contain country-specific effects.
Standard errors are displayed in parentheses. LR, likelihood
ratio; OLS, ordinary least squares.

*, **, and *** indicate significance at the 10%, 5%, and
1% levels, respectively.

TABLE 6
Robustness Checks (Dependent Variable: LREGS)

 (a) (b)

LGDPPC 3.724 *** (0.533) 4.821 ** (2.141)
INDSHEMP -0.100 *** (0.030) -0.128 *** (0.046)
LEFT -0.702 *** (0.239) -0.970 *** (0.344)
URBAN -0.106 *** (0.036) -0.116 (0.076)
INEQUAL 0.063 ** (0.025) 0.057 * (0.031)
INEQUAL interpolated -- --
Constant -24.872 *** (3.595) --
Observations 50 50
[R.sup.2] 0.941 0.913
Hausman test ([H.sub.0]: 0.841 --
 random effects,
 [H.sub.1]: fixed
 effects)
F test (significance of 4.812 *** --
 country-specific
 random effects)
F test (significance of 1.329 --
 time-specific
 random effects)
LR test (Ho: pooled OLS, -- 120.0 ***
 [H.sub.1]: country and
 time-specific fixed
 effects)

 (c) (d)

LGDPPC 3.405 *** (0.479) 1.883 *** (0.466)
INDSHEMP -0.093 *** (0.030) -0.065 ** (0.028)
LEFT -0.378 * (0.232) -0.478 ** (0.234)
URBAN -0.100 *** (0.037) -0.064 ** (0.029)
INEQUAL - -
INEQUAL interpolated 0.060 ** (0.029) 0.041 * (0.026)
Constant -0.378 *** (0.232) 0.041 *** (0.026)
Observations 102 102
[R.sup.2] 0.921 0.934
Hausman test ([H.sub.0]: 2.965 0.891
 random effects,
 [H.sub.1]: fixed
 effects)
F test (significance of 12.960 *** 18.815 ***
 country-specific
 random effects)
F test (significance of -- 4.840 ***
 time-specific
 random effects)
LR test (Ho: pooled OLS, -- --
 [H.sub.1]: country and
 time-specific fixed
 effects)

Notes: Columns (a) and (d) contain country- and time-specific
random effects, Column (b) includes country- and time-specific
fixed effects, and Column (c) incorporates country-specific
random effects. Standard errors are displayed in parentheses.

*, **, and *** indicate significance at the 10%, 5%, and
1% levels, respectively.