# Polluters and collective action: theory and evidence.

1. Introduction

The seminal theory by Olson (1965) predicts that industries with fewer firms have a greater ability to undertake collective action. They organize cooperative political action more easily because greater concentration lowers the cost of political action. (1) The empirical evidence is inconclusive, however. Andres (1985), Masters and Keim (1985), Heywood (1988), and Humphries (1991) find positive effects of industry concentration on the probability of making political action committee (PAC) contributions (see also McKeown 1994). Pittman (1976) finds that concentrated industries generate greater contribution levels. Grier, Munger, and Roberts (1991) find an inverted U-shaped relationship between the level of PAC formation and industry concentration, with a maximum political participation rate occurring at a four-firm concentration ratio around 0.45. Esty and Caves (1983) and Zardkoohi (1985) report ambiguous effects of concentration on PAC contributions. (2)

In this paper we suggest an alternative perspective on firms' ability to organize collective action which, to our knowledge, has been ignored so far. The novel argument is that industries that face multiple regulations (a greater number of policy issues) find it easier to overcome collective action problems and sustain lobbying. In particular, we focus on the difference between firms in polluting and clean industry sectors. Using a simple repeated game framework similar to Spagnolo (2000), we argue that firms in industries that are naturally polluting (because of their input requirements), and therefore incur pollution abatement costs, will face an additional policy battle compared with other industries, everything else equal. (3) This enables such industries to sustain greater cooperation and lobbying. This is because firms seeking to form a lobby group face a free-riding problem due to a limited amount of "enforcement power" available to punish deviation and ensure cooperation. Firms that face multiple areas of regulation have an advantage in the formation of lobby groups because they have a greater amount of enforcement power available to reallocate between policy issues. When joint lobbying gives large gains in environmental policy, this surplus can be reallocated to trade policy, for example. Free-riding behavior on trade policy lobbying may more easily be disciplined. The prediction that emerges from our theoretical model is that polluting industries are relatively less affected by the free-riding problems involved in organizing political action, and we thus expect the level of political contributions to be higher in these sectors. (4)

We evaluate this prediction using a cross-section data set of U.S. manufacturing industries. Our empirical model builds on a multiple-equation model by Gawande and Bandyopadhyay (2000), who test the well-established theory of Grossman and Helpman (1994) on the pattern of protection (their theory takes lobby group formation as given). We augment Gawande and Bandyopadhyay's model with an additional equation for environmental policy stringency. (5) The empirical results lend support to our theory. Industry PAC contributions, and thus the level of lobby group cooperation, are greater in industries with larger pollution abatement costs. This result is robust to several measures of lobby group formation and environmental policy.

The present paper contributes to the recent literature on the formation of lobby groups. In the area of pollution taxation, Damania and Fredriksson (2000) argue that collusive industries may more easily form lobby groups that oppose such taxes. Using a related setup, Damania and Fredriksson (2003) discuss the effect of (exogenous) trade liberalization on environmental policy formation when lobby group formation is endogenous. Pecorino (1998) and Mitra (1999) discuss the formation of trade lobby groups. (6) Neither of these papers explore the relationship between collective action and the number of policy instruments, however. (7)

Empirical work is severely lacking in this area, although some related work does exists. Our paper complements Grier, Munger, and Roberts (1994), who argue that industries that potentially may benefit from government assistance contribute more in corporate PAC contributions but are hindered by collective action problems. Pittman (1988) shows that the level of federal regulations (primarily measured as the level of capital expenditures on pollution abatement induced by Environmental Protection Agency regulations) significantly determines campaign contributions. To our knowledge, no study addresses the influence of the number of regulations on the degree of political action, however.

The paper is organized as follows. Section 2 sets up a stylized model of the lobbying game. Section 3 describes the econometric model, and section 4 discusses the data. Sections 5 and 6 report the empirical results and the sensitivity analysis, respectively, while section 7 concludes.

2. A Theoretical Example

In order to illustrate our argument, we outline a stylized infinite horizon model with complete information, which makes use of the framework developed by Spagnolo (2000) in his study of linkages of environmental and trade policies in international agreements. The model will underline the reasons why we may expect pollution intensive sectors to have an easier time to sustain lobbying.

We have two industry sectors, k = A, B, which are identical except that sector A's production is polluting whereas sector B's production is nonpolluting. Each industry sector has two identical firms, h = a, b. Thus in total we have four firms, [k.sup.h]. All firms face at a minimum n government policies each denoted by i, that is, trade policy, corporate and wage taxes, etc. Because sector A firms are polluting, they in addition encounter an environmental policy, and these firms face n + 1 government policies. For simplicity we set n = 1, where the common government policy is a trade policy. Thus, policy issue i = 1 is trade policy, and policy issue i = 2 is environmental policy. (8) Both sectors thus produce tradable goods.

The two firms in each industry sector wish to organize joint trade policy lobbying. However, they face a prisoner's dilemma since when one firm contributes to the lobbying effort its rival has an incentive to deviate. The one-period strategic interaction on policy issue i is represented as a prisoner's dilemma in which each firm may choose to either contribute ([C.sub.i]), or free ride (defect) ([D.sub.i]). In industry k, the one-shot policy prisoner's dilemma game is characterized by firm h's action space [[THETA}.sup.k,h.sub.i] = {[C.sub.i], [D.sub.i]},

k[member of]{A, B}, h[member of]{a, b}, and policy payoff functions [[pi].sup.kh].sub.i]:[[THETA].sup.k.sub.i] [right arrow] [R.sup.2], where [[THETA].sup.k.sub.i] = [[THETA].sup.ka.sub.i] x [[THETA].sup.kb.sub.i]. This generates the following policy payoff matrix for policy i = 1 (trade policy) for the sector k firms, where [Y.sub.i] > [X.sub.i] > [N.sub.i] > [Z.sub.i], and [Y.sub.i] + [Z.sub.i] < 2[X.sub.i]: The two firms in sector A in addition face a similar prisoner's dilemma on policy i = 2 (environmental policy). In sector A, the payoffs from the policy outcomes on both policy issues enter firm h's aggregate static profit functions [[PI].sup.h.sub.A] = [PI] ([[pi].sub.1][[pi].sub.2], h = a, b, which are continuous and twice differentiable, with [differential][[PI].sup.h.sub.A]/[differential][[pi].sub.i] > 0, I = 1,2. In sector B, the payoff from the policy outcome on trade policy alone enters each firm's aggregate static profit functions [[PI].sup.h.sub.B] = [PI]([[pi].sub.i]), h = a,b, with [differential][[PI].sup.h.sub.B]/[differential][[pi].sub.i] > 0. In each period t firm h in sector k maximizes the net present value of profits from policy outcomes NP[V.sup.k,h] = [[summation].sup.[infinity].sub.[tau]=t] [[delta].sup.[tau]-t][[PI].sup.[tau]], where [delta] < 1 is the firms' discount factor. Our theoretical analysis ignores the policy maker entirely, for simplicity our focus is on the lobby groups only. We also disregard the impact on firm behavior of redistributed tariff or pollution tax revenue income in this analysis.

The problem is analyzed in a supergame, that is, the firms are assumed to interact over an indefinite period of time. We let firms use trigger strategies as discussed by Friedman (1971), and we focus on symmetric stationary lobbying sustained by stationary punishment strategies. First, assume that the trigger strategies apply only to one separate policy issue at a time. Thus, defection on trade policy is punished by abandoning all cooperation on trade policy by reverting to the one-shot Nash equilibrium. In sector A, taking environmental policy (policy 2) as given, the one-period gain from unilateral defection on trade policy lobbying only is given by

[G.sup.A.sub.1] = [[PI].sup.A]([Y.sub.1], [[pi].sub.2]) - [[PI].sup.A]([X.sub.1], [[pi].sub.2]) (l)

and the cost of defection is given by

[C.sup.A.sub.1] = [delta]/1 - [delta] [[[PI].sup.A][X.sub.1], [[pi].sub.2]) - [[PI].sup.A] [N.sub.1], [[pi].sub.2]). (2)

In the polluting sector A, joint lobbying on policy issue 1 is sustainable if [G.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.1]. Suppose instead that the firms in sector A punish a deviator in both policy dimensions. Thus, if a firm deviates, the rival firm's retribution strategy is punishment in both trade and environmental policies. In this case the deviating firm would chose to deviate on both policies (see Bernheim and Whinston 1990). Then, the gain from deviation is given by

[G.sup.A.sub.1,2] = [PI]([Y.sub.1], [Y.sub.2]) - [PI]([X.sub.1], [X.sub.2]), (3)

and the cost of deviation is given by

[C.sup.A.sub.1,2] = [delta]/1 - [delta][PI]([X.sub.1], [X.sub.2]) - [PI]([N.sub.1], [N.sub.2])]. (4)

In this case, joint lobbying on both policy issues is sustainable if [G.sup.A.sub.1,2] [less than or equal to] [C.sup.A.sub.1,2] In sector B the optimal punishment strategy involves deviation on policy 1 (trade policy), and thus joint lobbying requires [G.sup.B.sub.1] [less than or equal to] [C.sup.B.sub.1].

The focus of our discussion is the fact that firms seeking to form a lobby group may face a problem of limited enforcement power. The expected short-term gains from deviation from lobbying may be greater than the long-term gains from cooperation. In this case, lobbying is not sustainable. However, firms that interact in multiple areas of regulation (policy issues) have an advantage in the formation of lobby groups due to the greater amount of available enforcement power that can be reallocated between policy issues. When cooperation gives large gains in one policy area such as environmental policy, this slack can be reallocated to enforce cooperation in another such as trade policy. The slack of expected gain from cooperation on environmental policy may then be used to discipline behavior on trade policy, or vice versa. The slack available in the environmental policy area will depend on the pollution intensity (pollution abatement costs) of the sector in question because this reflects the amount at stake (i.e., the costs of regulation). In this theoretical example we attempt to show this in the simplest possible manner. Thus, we follow Bernheim and Whinston (1990) by assuming that sector A firms' profit functions are linearly separable in policy issues. (9)

ASSUMPTION 1. In sector A, each firm has a profit function that is linearly separable in the two policies 1 and 2, such that [[PI].sup.A] = [[PI].sup.A.sub.1] ([[pi].sub.1]) + [[PI].sup.A.sub.2] ([[pi].sub.2]).

Thus, firm h in sector A will maximize the net present value of profits by maximizing NP[V.sup.A,h] = [[summation].sup.[infinity].sub.[tau]=1] [[delta].sup.[tau]-t] [[[PI].sup.A,h.sub.1]([[pi].sub.1]) + [[PI].sup.A,h.sub.2]([[pi].sub.2])]. Assumption 1 also implies that if the firms in sector A use an optimal punishment strategy by canceling cooperation on both policy issues in order to deter deviations, the two separate lobbying games played now collapse into one game. In industry A this game has a one-period static payoff function [[PI].sup.A.sub.1,2] = [[PI].sup.A.sub.1] ([[pi].sub.1]) + [[PI].sup.A.sub.2]([[pi].sub.2]). On the other hand, in industry B firms lobby on only one policy. The one-period payoff function is [[PI].sup.B.sub.1] = [[PI].sup.B.sub.1]([[pi].sub.1]). We can now state the following result.

RESULT 1. Under Assumption 1, the firms in the polluting sector A have an equal or greater ability to sustain cooperation on lobbying than the firms in the clean sector B.

PROOF. See Appendix 1.

Cooperation on trade and environmental policy lobbying is sustainable if

[G.sup.A.sub.1] + [G.sup.A.sub.2] [less than or equal to] [C.sup.A.sub.1] + [C.sup.A.sub.2], (5)

where subscript 1(2) denotes trade (environmental) policy. We identify four main cases.

CASE 1. [G.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.1] and [G.sup.A.sub.2] [less than or equal] [C.sup.A.sub.2]. Condition 5 holds, and being a polluting industry does not affect industry A's ability to sustain lobbying. It is sustainable in any case.

CASE 2. [G.sup.A.sub.1] > [C.sup.A.sub.1] and [G.sup.A.sub.2] > [C.sup.A.sub.2]. Then

[G.sup.A.sub.1] + [G.sup.A.sub.2] > [C.sup.A.sub.1] + [C.sup.A.sub.2] (6)

and cooperation is not sustainable in sector A on any policy issue. Interaction in the environmental policy arena confers no advantage to industry A.

CASE 3A. [G.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.1] and [G.sup.A.sub.2] > [C.sup.A.sub.2]. A polluting industry is able to cooperate on lobbying if [G.sup.A.sub.2] - [C.sup.A.sub.2] [less than or equal to] [C.sup.A.sub.1] - [G.sup.A.sub.1], that is, if the cost of deviation on the trade policy is greater than the gain from deviation on the environmental policy issue.

CASE 3B. [G.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.1] and [G.sup.A.sub.2] > [C.sup.A.sub.2], where [G.sup.A.sub.2] - [C.sup.A.sub.2] > [C.sup.A.sub.1] - [G.sup.A.sub.1]. The firms in sector A are able to cooperate on trade policy only, given that the two policies are treated as separate games.

CASE 4A. [G.sup.A.sub.1] > [C.sup.A.sub.1] and [G.sup.A.sub.2] [less than or equal] [C.sup.A.sub.2]. Industry A is able to cooperate on lobbying on both policy issues because of the existence of environmental regulations, if [G.sup.A.sub.1] - [C.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.2] - [G.sup.A.sub.2]. The benefit of cooperation on environmental policy is greater than the gain from deviation on trade policy. In this case, lobby group formation is feasible only in polluting sectors.

CASE 4B. [G.sup.A.sub.1] > [C.sup.A.sub.1] and [G.sup.A.sub.2] [less than or equal to] [C.sup.A.sub.2]. If [G.sup.A.sub.1] - [C.sup.A.sub.1] > [G.sup.A.sub.2] - [G.sup.A.sub.2]. Only environmental policy lobbying cooperation is possible, given that the two policies are treated as separate games.

In sum, cooperation on both policy dimensions is sustainable as long as the gains from deviation on policy i (that is [G.sup.A.sub.1] - [C.sup.A.sub.1], i = 1,2) are no greater than the gains from cooperation on policy j (that is [C.sup.A.sub.j] - [G.sup.A.sub.j], j [not equal to] i). Thus even if cooperation is not sustainable on policy issue i in isolation, it may be sustainable on both issues as long is there is some enforcement slack available on policy issue j.

For expositional ease these results have been derived on the assumption that trade and environmental policy are linearly independent in their effects on profits (Assumption 1). It is recognized that this assumption may in some cases not hold in practice. However, as shown by Spagnolo (2000) for the case of issue linkage in international agreements, the results are the same whether the policy instruments under consideration are linearly interdependent or strategic substitutes.

From the above discussion, it follows that a polluting industry never faces greater difficulties in organizing lobbying than it would have faced had it been a clean industry. On the other hand, there are circumstances where it is favorable for lobbying to be subject to environmental regulations, in particular when extra enforcement power (or slack in gains from cooperation) is available in the environmental policy area. This extra enforcement power can then be used, for example, in the trade policy area.

Note that although the discussion thus far has (for simplicity) been framed in terms of environmental and trade policies, it also applies to other policy areas such as taxation, subsidies, and other forms of regulations. Thus, the results are in reality more general, and we therefore formulate the following prediction in a more general fashion:

PREDICTION 1. Firms operating in sectors with multiple regulations sustain equal or greater levels of political contributions compared with firms operating in sectors with only one regulation, ceteris paribus.

This prediction is the subject of our empirical testing in the next section. (10) If, as is generally assumed in the political economy literature, the level of protection is increasing in political contributions, it follows that firms in polluting industries will obtain greater protection than those in clean industries. (11) For expositional ease we have ignored a number of other theoretical issues. Perhaps the most important of these is the assumption that firms sustain lobbying by the threat of reversion to the one-shot Nash equilibrium. Abreu (1986) has shown that in general the one-shot Nash equilibrium is not the harshest available credible punishment. In this paper we have followed most of the supergame literature and have ignored this issue for simplicity. However, the qualitative results are unlikely to be affected by this assumption, since the results hinge on the observation that for any given retribution strategy, enforcement slack in one domain can be used to enforce cooperation on some other issue. Another problem that we have ignored is that of the renegotiation proofness of the equilibrium. As noted by Shapiro (1989), the restriction that equilibria be renegotiation proof would shrink the set of outcomes where lobbying (i.e., cooperation) is sustainable. Thus, so long as renegotiation proof equilibria exist, this restriction would not alter the fundamental argument that interaction on numerous issues may facilitate greater cooperation. Finally, it is worth noting that the equilibria identified in this paper emerge from incentive compatible noncooperative actions of players. Hence the conclusions do not depend upon the existence of a formal industry lobby group, in much the same way as the tacitly collusive equilibria in repeated games do not presume the existence of a cartel. Our formal analysis simply reveals that when firms interact in multiple areas of regulation, their contributions to lobbying activity (which has public good characteristics) will be greater than if they interact on only one policy issue, irrespective of whether the lobbying occurs through an industry organization or by firms individually. (12)

3. Econometric Specification

Our primary focus in the empirical work is on the prediction generated by our theoretical example: We investigate whether environmental regulations and the amount of lobby group contributions are positively correlated across U.S. manufacturing sectors.

Our empirical model builds on the model estimated in Gawande and Bandyopadhyay (G-B) (2000), which provides structural estimations of the theoretical model of trade policy formation by Grossman and Helpman (G-H) (1994) (see Goldberg and Maggi 1999 for an alternative test of the G-H model). The G-H theory generates predictions of trade policy outcomes across industries, taking the existence of lobby groups across industries simply as given, and ignoring environmental policy. The G-H theory argues that given the existence of a lobby, the equilibrium lobby group contribution equals the welfare loss caused by the lobby's participation in the political equilibrium. In essence, seen in this light our theory suggests that polluting sectors will have an easier time generating sufficient funds to compensate the government for any induced trade policy distortions. In addition, these firms must also contribute the funds necessary to compensate the government for the environmental policy distortions associated with lobbying.

In G-B's empirical implementation of the G-H theory (i) lobby group contributions and (ii) the degree of import penetration are jointly determined with (iii) the level of protection, thus accounting for the endogeneity of three variables. Our empirical model extends the literature by adding a fourth endogenous variable: environmental policy stringency. Our empirical model thus consists of four equations. Since three of these equations are discussed in detail in G-B, we here restrict ourselves to a brief overview of each equation and its main features, highlighting the differences to the previous approach.

The first equation, the lobby group contributions (LGC) equation, explains the lobbying efforts by firms (CONTRIBUTIONS). (13) G-B builds on Vousden (1990) to derive an equation reflecting the lobby contributions necessary to compensate the government for the deadweight loss from protection. The equation includes the level of nontariff barriers (NTB/[1 + NTB]), the import elasticity (ELASTIC), and the import penetration ratio (IMPORTPENETRAT). The greater any of these variables, the larger the deadweight loss, and the greater the political spending needed to keep the government indifferent between free trade and protection. In addition, three variables account for the conflict of interest between downstream users, who oppose upstream protectionism and upstream firms (Olson 1965). These are the firm concentration in upstream industries (HERFIN) and downstream industries (DOWNSTREAMHERFIN), as well as the share of upstream producers' output sold downstream as intermediate inputs (DOWNSTREAMSHARE). The higher these measures are, the greater the bargaining power of downstream users, forcing upstream users to make greater political contributions.

The novel feature of our specification of the LGC equation is the inclusion of pollution abatement costs (ABATEMENTCOSTS1) as an additional regressor. (14) Our theoretical model suggests that pollution abatement costs and lobby group contributions (lobby group organization effort) by firms should be positively correlated. When the industry faces a greater number of policy battles, the overall loss from deviation in lobbying are larger, and potentially more enforcement power can be reallocated to trade policy. The higher the pollution intensity and thus cost of pollution abatement, ceteris paribus, the more is at stake for a polluting firm, the more enforcement power should be available. Thus, the generation of political funds by an industry lobby group becomes easier.

The NTB equation, determines the level of protection (NTB/[1 + NTB]). The equation is the empirical counterpart of the closed-form solution for the level of protection derived by G-H. (15) The advantage of a strictly theory-driven specification is that the determinants of protection as well as the interaction terms that should be included in the equation are determined a priori. (16)

The environmental policy stringency (EPS) equation determines the stringency of environmental regulation approximated by the level of pollution abatement cost (ABATEMENTCOSTS1). We include NTB protection as a right-hand side (RHS) variable in the EPS equation. If environmental and trade policies are determined by similar political forces (accounting for the same RHS variables as in Eqn. 8), the two variables will be negatively correlated. If, on the other hand, policy makers tend to compensate industries for low trade barriers (stringent environmental policy) with less stringent environmental policy (greater trade barriers), a positive coefficient is expected (Eliste and Fredriksson 2002). We also include political spending by firms as a regressor, since more intense lobbying should result in less stringent environmental standards. Finally, we include a measure of an industry's pollution intensity as an exogenous regressor (INTENSITY).

The final equation explains the level of import penetration (IMPORTPENETRAT). (17) Since the stringency of environmental regulation should affect the competitiveness of domestic firms, we include a measure of pollution abatement cost in our specification of the import penetration equation.

The system of equations that we estimate can be written as

ln CONTRIBUTIONS = [[alpha].sub.0] + [[alpha].sub.1] ln HERFIN + [[alpha].sub.2] In IMPORTPENETRAT + [[alpha].sub.3] ln (NTB/1 + NTB) + [[alpha].sub.4] ln ELASTIC + [[alpha].sub.5] ln DOWNSTREAMSHARE + [[alpha].sub.6] ln DOWNSTREAMHERFIN + [[alpha].sub.7] ln ABATEMENTCOSTS1 + [[epsilon].sub.1], (7)

NTB/1 + NTB = [[beta].sub.0] + [[beta].sub.1] [(IMP/CONS).sup.-1]/ELASTIC + [[beta].sub.2]ORGANIZED[[(IMPORTPENETRAT).sup-1]/ELASTIC] + [[beta].sub.3]INTERMTARIFF + [[beta].sub.4]INTERMNTB + [[epsilon].sub.2], (8)

ABATEMENTCOSTS1 = [[delta].sub.0] + [[delta].sub.1] NTB/1 + NTB + [[delta].sub.2] CONTRIBUTIONS + [[delta].sub.3]INTENSITY + [[epsilon].sub.3], (9)

IMP/CONS = [[gamma].sub.0] + [[gamma].sub.1] NTB/1 + NTB + [[gamma].sub.2] ABATEMENTCOSTS1 + X[GAMMA] + [[epsilon].sub.4], (10)

where X is a (row) vector of exogenous variables, and [GAMMA] is the corresponding (column) vector of coefficients. (18) The error terms are assumed to be normal and independently distributed across observations. A description of all variables used is given in Table A1 in the Appendix.

In addition to the endogenous left-hand side variables, in CONTRIBUTIONS, NTB/(1 + NTB), IMPORTPENETRAT, and ABATEMENTCOSTS1, the system of Equation 7 includes nonlinear functions of some of these variables on the right-hand side: in (NTB/[1 + NTB]), In IMPORTPENETRAT, and in ABATEMENTCOSTS1 in Equation 7, and ORGANIZED in Equation 8. (19) To account for these nonlinearities, we use a modified version of the two-stage least squares estimator proposed by Kelejian (1971) (see also G-B). In the first stage, we regress each of the endogenous variables and their transformations on the set of exogenous variables. In addition to the linear terms, we admit squared terms and a number of cross product terms. (20) In the second stage, we estimate Equation 7 but replace the endogenous right-hand side variables by their predicted values from the first-stage regressions.

4. Data

Our data are at the four-digit Standard Industrial Classification (SIC) level, and the year is 1983 unless noted differently. (21) Summary statistics are given in Table A2 in the Appendix. All data except for pollution abatement costs and pollution intensity are from G-B. The following data description is brief, since a detailed description of how each variable was constructed is found there.

Political action committee (PAC) contributions cover the four Congressional election cycles, 1977-78, 1979-80, 1981-82, and 1983-84, were constructed by Gawande (1998) from Federal Election Commission tapes. (22) While we acknowledge that individual firms may sometimes have different political objectives, we follow the extant empirical literature by assuming that the contributions made by corporate PACs may be assumed to be a public good. This is also consistent with a large share of the theoretical and empirical literature on the determination of trade and environmental policies.

Unfortunately, there are no direct measures of EPS (such as regulation dummies, etc.) available at detailed levels of disaggregation. Instead, we use two indirect measures of pollution abatement costs, scaled by added value: capital expenditures (ABATEMENTCOSTS1) and operating costs (ABATEMENTCOSTS2). Both measures are available from the Pollution Abatement Costs and Expenditures Survey (U.S. Department of Commerce 1983). With ABATEMENTCOSTS1 as the EPS measure, the size of the data set is 89, compared with 177 when ABATEMENTCOSTS2 is used. When using lagged instead of current pollution abatement measures, the sample sizes are 94 and 155, respectively. Our measure of an industry's pollution intensity (INTENSITY) is taken from List and Co (2000) who identify pollution intensity at the two-digit level. We replicate their results at the four-digit level.

The political organization dummy (ORGANIZED) is based on a regression of PAC contributions per firm (scaled by value added) on bilateral import penetration by partner country interacted with 20 two-digit SIC dummies. There are five partner countries: France, Germany, Italy, Japan, and the UK. Two-digit industries with positive coefficients are considered organized. The union of all organized industries from the five regressions constitutes the set of organized industries (at the four-digit SIC level). In the original data set used by G-B, 165 out of 242 industries are organized (68.2%).

As is typical in the empirical literature on endogenous protection, an aggregate NTB coverage ratio is used to measure the level of protection. The inverse import penetration ratio is measured as consumption divided by imports and then scaled by 10,000 to account for small values of imports in some industries. The own- and cross-price elasticity measures are taken from Sheills, Stem, and Deardorff (1986), purged of their inherent errors-in-variables problem, and reproduced at the four-digit SIC level. Industry characteristics, added value, and the Herfindahl index were constructed from the 1982 census of manufacturing and, if necessary, from various annual surveys of manufactures. Protection levels in intermediate industries as well as the concentration measure in downstream industries were constructed from U.S. input-output tables.

5. Empirical Results

Table 1 contains 2SLS estimates of the LGC and the EPS equations (Eqns. 7 and 9, respectively). (23) We use notation similar to G-B, as well as G-B's convention for reporting significance levels. In particular, estimates with a t statistic greater than unity are marked with an asterisk since their inclusion in the regression equation increases the value of the adjusted [R.sup.2]. The most striking result of Table 1 is that the coefficient estimate for ABATEMENTCOSTS1 has the expected positive sign in the LGC equation. This lends support to our theory. In addition, the precision of the estimated relationship is high--the estimated coefficient is statistically significant at the 1% level. (24) The remaining coefficient estimates in Equation 7 all have the same signs as in G-B. (25) In addition, the size of the estimated coefficients is similar as well, with two exceptions. The coefficient estimates for import penetration, In IMPORTPENETRAT, and downstream industry concentration, In DOWNSTREAMHERFIN, are smaller than in G-B and no longer significant. (26) The drop in sample size as a result of the inclusion of ABATEMENTCOSTS1 as a regressor in Equation 7 raises the issue of sample selection bias. Given the total number of four-digit SIC industries (N = 448), even the sample size used in G-B (N = 242) is not immune to this problem. More importantly, most of our coefficient estimates are very close in magnitude to those reported in G-B, while our t statistics are often lower. Thus, instead of sample selection bias, the main consequence of the smaller sample seems to be the efficiency loss of the estimation.

With regard to the EPS Equation 9, we find that trade barriers and environmental policy stringency are positively correlated. In addition, PAC contributions and environmental regulation are positively correlated, contrary to the expected sign. This may be explained in the following way. Polluting industries raise greater levels of PAC contributions but devote these funds primarily to other policy areas. Finally, industries that are more pollution intensive have higher levels of pollution abatement expenditures. Except for the intercept, all coefficient estimates are statistically significant at the 1% level.

To provide an alternative test of our theoretical prediction, we replace pollution abatement capital expenditures (ABATEMENTCOSTS1) by operating costs (ABATEMENTCOSTS2) as the measure of environmental policy stringency. One can argue that ABATEMENTCOSTS2 is a better measure of abatement spending, and thus policy stringency, than ABATEMENTCOSTS1 since it is easier for firms to identify the environmental portion of their total operating costs than the environmental fraction of their total capital expenditures, in particular when more and more abatement involves process changes or is an integral part of new technologies. In our case, using ABATEMENTCOSTS2 instead of ABATEMENTCOSTS1 does not lead to qualitatively different results with regard to our hypothesis (see Table 2). The operating cost measure is positive and significant around the 10% level in the LGC equation. The size of the estimated relation is different, though. The substantially smaller coefficient estimate for ABATEMENTCOSTS2 compared with ABATEMENTCOSTS1 indicates that the impact of environmental regulation on lobby group formation and protection may only be one-third as strong as suggested by the previous measure of policy stringency. The remaining coefficient estimates in both equations are in line with their corresponding estimates from Table 1. The overall fit of the LGC estimation regression (0.13) is lower than in Table 1 (0.36), despite the larger sample size (N = 177 in Table 2 compared with N = 89 in Table 1).

6. Sensitivity Analysis

We perform a number of sensitivity tests. First, we use lagged instead of current values for both ABATEMENTCOSTS1 and ABATEMENTCOSTS2. This reduces the system to three equations since the potential endogeneity of the environmental stringency variable is no longer an issue. We report the results for the LGC equation in Table 3.27 The results confirm our previous findings about the impact of environmental stringency on lobby contributions (formation). The positive coefficient estimate on lagged ABATEMENTCOSTS1 is significant at the 1% level. This not only confirms the prediction of our theoretical model but also provides indirect evidence for the goodness of our pollution intensity measure as an instrumental variable for environmental stringency. In contrast to lagged capital expenditures, lagged pollution abatement operating costs (ABATEMENTCOSTS2) have no effect on lobby group contributions. Combined with the findings from Table 2, the result implies that most of the impact of EPS (as measured by ABATEMENTCOSTS2) on lobby group contributions is contemporaneous.

Second, we investigate a potential endogeneity bias that may arise due to the use of added value as both a scale measure in Equation 10 and to standardize lobbying efforts by firm (CONTRIBUTIONS) and pollution abatement costs (abatement capital expenditures, as well as abatement operating costs). (28) We instead use the number of employees per firm (FIRMSCALE2) as an alternative measure of industry size in Equation 10. Reestimating the system of equations (Eqns. 7-10), we find no evidence of such a bias, as far as our main results are concerned. A comparison between Table 1 and the results presented in column 1 of Table 4 yields only minor changes in the point estimates in either equation. In particular, our previous finding of a positive and highly significant effect of pollution intensity (INTENSITY) on lobby group contributions remains unchanged. Similarly, when we contrast the results from column 3 in Table 4 with those reported in Table 2 we once again find no substantive changes in point estimates or t statistics. Most importantly, the coefficient estimate of the operating cost abatement measure (ABATEMENTCOSTS2) remains around 0.3, and the significance level falls only slightly.

Next, we investigate a potential bias of our environmental policy stringency estimates due to the special status of industries made up of large-scale firms. Owing to their size, larger firms may interact with government agencies in several dimensions. They may therefore be more likely to lobby the government for reasons such as tax breaks, subsidies, or changes in certain types of (nonenvironmental) regulations. To test the robustness of our results, we therefore remove the large-scale industries from each sample and reestimate Equations 7-10.29 The results reported in columns 2 and 4 of Table 4 show that the effect of environmental policy stringency on lobby group contributions remains positive and significant. In fact, omitting the largest industries from the sample strengthen the results. The point estimate is about 50% larger in both cases (1.22 and 0.55, compared with 0.82 and 0.32 in Tables 1 and 2, respectively), and the precision of the estimated relationship has improved markedly. In addition, the overall fit of the policy stringency equation reported in columns 2 and 4 of Table 4 (Eqn. 9) has improved, with an adjusted [R.sup.2] above 0.3 compared with 0.14 and 0.18 in Table 1 and 2, respectively. Note also that some coefficient estimates, in particular those controlling for concentration in downstream industries in the LGC equation, become smaller and statistically insignificant in column 2. We thus conclude that controlling for the potential additional interactions between large firms and the government does not alter the main conclusions emerging from our previous empirical analysis.

Our final robustness checks involve estimations of a number of different specifications for the system of equations such as semilog of Equation 7, additional regressors (i.e., the degree of unionization in Eqns. 7 and 8), and a discrete measure of CONTRIBUTIONS in Equation 9. None of these alternative specifications yield results that are qualitatively different from the results presented in Tables 1 to 3. In effect, if the two measures of downstream lobbying competition used as regressors in Equation 7 are used in Equation 9 instead, the coefficient estimate on In ABATEMENTCOSTS2 in Equation 7 is positive and significant at the 1% level (t value 2.66).

7. Conclusion

This paper proposes a new theory of lobby group formation. We argue that firms in industry sectors encountering a greater number of policy instruments more easily organize lobby groups because of a greater amount of available "enforcement power." When lobby group cooperation gives large gains in one policy area, this surplus can be reallocated to another policy area. This causes an advantage for pollution intensive firms. Any slack of expected gain from cooperation on environmental policy may be used to discipline lobbying behavior on, for example, trade policy. Firms in polluting sectors facing stiffer environmental policy should therefore be able to raise greater amounts of PAC contributions.

Our theoretical prediction finds support in the data. Sectors with greater pollution abatement expenditures and thus more stringent levels of environmental protection have significantly greater levels of PAC contributions. This result is robust to several different variable measures and specifications. To our knowledge, this is a novel finding in the literature.

Appendix

PROOF OF RESULT 1. Note that under Assumption 1, when the punishment strategy calls for deviation in only one policy instrument, sector A firms have [G.sup.A.sub.i] = [[PI].sup.A.sub.i]([Y.sub.i]) - [[PI].sup.A.sub.i]([X.sub.i]), and [C.sup.A.sub.i] = [delta]/1 - [delta][[[PI].sup.A.sub.i] ([X.sub.i]) - [[PI].sup.A.sub.i]([N.sub.i])], i = 1, 2. Thus, cooperation is sustainable if

[[PI].sup.A.sub.i]([Y.sub.i]) - [[PI].sup.A.sub.i]([X.sub.i]) [less than or equal to] [[delta]/[1 - [delta]]] [[PI].sup.A.sub.i]([X.sub.i]) - [[PI].sup.A.sub.i]([N.sub.i]) (A1)

Rearranging condition AI it follows that cooperation is feasible for all values of the discount factor such that

[delta] [greater than or equal to] [[[PI].sup.A.sub.i] ([Y.sub.i]) - [[PI].sup.A.sub.i]([X.sub.i])]/[[PI].sup.A.sub.i] ([Y.sub.i]) - [[PI].sup.A.sub.i]([N.sub.i]) [equivalent to] [[DELTA].sub.i] (A2)

On the other hand, firms in sector A have the option of using a punishment strategy that entails reverting to the noncooperative strategy on both policies after any form of deviation. Given these two alternatives, a sector A firm's optimal deviation is a deviation on both policy issues simultaneously. The gain from defecting unilaterally from lobbying cooperation is given by [G.sup.A.sub.1,2] = [G.sup.A.sub.1] + [G.sup.A.sub.2], and the cost of defection is given by [C.sup.A.sub.1,2] = [C.sup.A.sub.1] + [C.sup.A.sub.2]. Cooperation on trade and environmental policy lobbying is sustainable if

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (A3)

Rearranging condition A3 it follows that cooperation is feasible for all values of the discount factor such that

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (A4)

Without loss of generality, suppose that when there is cooperation with only one policy instrument this involves interaction over policy i = 1 (i.e., trade policy). Clearly cooperation is more easily sustained with interaction over two policies if [[DELTA].sub.1,2] < [[DELTA].sub.1]. Substituting from conditions A2 and A4 and rearranging this inequality yields

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (A5)

where [OMEGA] = [[PI].sup.A.sub.1] ([N.sub.1]) - [[PI].sup.A.sub.2]([X.sub.2]) + [[PI].sup.A.sub.2]([N.sub.2]) ([[PI].sup.A.sub.1]([X.sub.1]) - [[PI].sup.A.sub.1]([Y.sub.1]) < 0, and the sign of [OMEGA] follows from the assumption that ([[PI].sup.A.sub.i]([X.sub.i]) < [[PI].sup.A.sub.i] ([Y.sub.i]), [for all] i = 1,2. It follows that with interaction over two policy issues, cooperation is feasible over a larger range of values of the discount factor. Hence, lobbying is easier to sustain in sectors encountering two policies. (30) QED.

We are grateful to two helpful referees, Toke Aidt, Brian Copeland, Bouwe Dijkstra, Christian Hilber, Angeliki Kourelis, Daniel Millimet, Felix Oberholzer-Gee, Barkley Rosser Jr., Kevin Siqueira, and session participants at the Public Choice Society Meetings in San Antonio, at the European Public Choice Society Meetings in Paris, and at the EAERE Annual Conference in Southampton for useful discussions, comments, and suggestions, and to Kishore Gawande for sharing his data and providing additional comments. Earlier versions of the paper were circulated under the title "Trade Policy, Polluters, and Collective Action: Theory and Evidence." The usual disclaimers apply.

Received February 2003; accepted September 2004.

(1) A large body of literature has extended this theory in various directions. However, Grossman and Helpman (1994) argue that the literature suffers from a lack of attention to the issue of lobby group organization, and Persson and Tabellini (2000, p. 175) find that "The major problem is that we lack a precise model of the process whereby some groups get politically organized and others not. This is a difficult question to which there is still no satisfactory answer."

(2) Most previous studies of PACs examine the allocation and timing patterns of PACs (see, e.g., Snyder 1990 and Stratmann 1992, 1995), as well as the effects of PACs (see, e.g., Salamon and Siegfried 1977 and Stratmann 1991). A different strand of the literature argues that the importance of corporate PAC contributions in U.S. politics is relatively small and that PAC contributions are not equivalent to bribes or a proxy for lobbying activity (see Milyo, Primo, and Groseclose 2000; Ansolabehere, Snyder, and Tripathi 2002; Ansolabehere, de Figueiredo, and Snyder 2003). These authors argue that PAC contributions primarily buy access to legislators or are a form of consumption good with little effect on legislation.

(3) Spagnolo (2000) models (theoretically) issue linkage in international agreements. He does not discuss lobby group formation or the determination of trade protection or environmental policies. See also Bernheim and Whinston (1990) and Conconi and Perroni (2002).

(4) This conclusion is based on the assumption that the interests of firms in an industry coincide sufficiently. In the current context this requires that regulations are sufficiently severe to induce firms to lobby for less stringent regulations. Clearly in cases where either firms are highly heterogeneous or regulations have differing impacts on different firms there will be no incentive to lobby for the same policy changes.

(5) Gawande and Bandyopadhyay (2000), as well as Goldberg and Maggi (1999), find support for the predictions made by the model of Grossman and Helpman (1994). See also Gawande (1997, 1998). See Gawande and Krishna (2003) for a recent survey of this literature.

(6) Pecorino (1998) studies free riding in firms' lobbying for protectionism in a repeated game model. He finds that an increase in the number of firms in an industry does not necessarily imply that free-rider problems increase. Cooperation may be sustained even under perfect competition, that is, with an infinite number of firms. Mitra (1999) endogenizes the organization of lobby groups using the Grossman and Helpman (1994) model. He employs an industrial organization-endogenous market structure approach to find the equilibrium number of lobby groups. Mitra argues that geographically concentrated industries with large capital stocks, inelastic demand functions, and few capital owners are more likely to organize.

(7) Our work is a complement to the literature on collective action on the demand side (Hamilton 1993) and to the theory of regulatory "capture" originated by Stigler (1971) and extended by Peltzman (1976) and many others. The latter authors argue that regulations that erect barriers to entry function as outright transfers of wealth and are therefore demanded by industry. In our paper the existence of environmental regulations is due to industries being "naturally" polluting, and larger transfers may occur because of a greater ability to undertake collective action.

(8) An alternative would be to assume all sectors have positive levels of pollution, but some sectors produce only nontradable goods. However, in our data set all sectors are engaged in international trade.

(9) Spagnolo (2000) discusses the consequences of relaxing this assumption. We abstract from further complications in order to focus the example on our main point. Assumption 1 will most likely hold when output is relatively unaffected by the environmental policy instrument, such as when a technology standard is used to regulate pollution. This is the most commonly used environmental policy instrument in the United States Moreover, when a pollution standard and an import quota (assumed given to importers free of charge) are used, no revenues arise. This is consistent with the policies used in our empirical work.

(10) Given Result 1, a polluting industry lobby's problem is the allocation of the available political funds among policy lobbying areas. The marginal net return to lobbying must be equal across policy instruments. Thus, some positive share of these funds will likely be allocated to lobbying on policies other than environmental policies, for example, trade policy. In case the marginal return to lobbying is the greatest in the environmental policy area, the additional funds generated may potentially be fully used in the environmental policy area only, in which case trade policy lobbying would not be affected. This is ultimately an issue that is perhaps best resolved empirically.

(11) It is perhaps useful to note that if this condition is not satisfied in the Grossman and Helpman (1994) model, there is no equilibrium of the lobbying game (see Damania and Fredriksson 2003 for a discussion of this issue).

(12) In some cases the lobbying may be administered through industry groups, while in other circumstances firms may directly lobby policy makers. Irrespective of the organizational arrangements, the theory predicts that with interaction over multiple regulations there will be greater lobbying by firms.

(13) Lobbying is here viewed as directed toward lowering the stringency of environmental policy, rather than to change its type.

(14) Pollution abatement costs have been used as a proxy for the stringency of environmental policies in a number of empirical studies. See, for example, Levinson and Taylor (2001), who also provide an extensive survey of the literature.

(15) Our empirical specification of the NTB equation corresponds to the one estimated by G-B.

(16) First, the level of protection depends on the inverse import penetration ratio [(1MPORTPENETRAT).sup.-1], interacted with a political organization dummy variable (ORGANIZED), which takes the value of one if a sector is organized politically and zero otherwise. We expect that the inverse import penetration ratio has a positive effect on protection if the sector is organized since relatively larger domestic industries will make greater lobbying contributions and receive higher levels of protection provided they are organized. Second, the import demand elasticity has a negative effect on the level of protection due to the greater deadweight loss of taxation associated with a greater elasticity. Finally, our specification also includes two variables that measure the level of protection for intermediate goods used in that industry. The first is a measure of the average tariff level for intermediates (INTERMTARIFF), while the second measures the extent of intermediate goods protection through nontariff barriers (INTERMNTB). A higher level of protection for firms producing intermediate goods increases the cost of production for final goods producers. To stay competitive with foreign producers, final goods producers at home need higher levels of protection.

(17) Our specification of the import penetration equation follows specifications found in previous studies (G-B; Trefler 1993). Trade theory suggests that import penetration is a function of the level of protection, as well as factor endowments (as predicted by standard comparative advantage models), firm size, and industry concentration (as predicted by models with increasing returns), and own- and cross-price elasticities (to control for differing demand conditions). Trefler (1993) estimates an import penetration equation as part of a two-equation system. In addition to the protection measure, Trefler's import equation includes only variables that measure factor endowments. Our specification includes other variables as well and is therefore more general.

(18) The following variables are included in X: SCIENTIST, MANAGER, UNSKILLED, CONC4, FIRMSCALE, TARIFF, ELASTIC, CROSSELAST, REALELASTIC. For a description of these variables, see Table A1 in the Appendix.

(19) The dummy variable ORGANIZED is a transformation of CONTRIBUTIONS (see section 4 for details).

(20) The exact list of cross-product terms is available from the authors upon request. All first-stage estimation results including the results of the overidentification tests are available upon request.

(21) Like other empirical papers on endogenous tariff protection, we are confined to the use of 1983 data because of the lack of disaggregated data on nontariff barriers for later years.

(22) We thus follow, for example, Grier, Munger, and Roberts (1994) by using corporate PAC data in a study of the determinants of collective action (see also Goldberg and Maggi 1999 and G-B). The PACs in our data set are associated with individual firms. As pointed out by a helpful referee, PACs organized by corporations are "connected," as opposed to "unconnected." Whereas unconnected PACs must pay all of their operation costs from personal contributions from any U.S. citizen, connected PACs may have their operating expenses (staff salaries, lawyers' fees, fundraising costs, etc.) paid by their parent organizations. They can only contribute money to candidates that are solicited from and contributed by employees of the sponsoring organization. Connected PACs cannot legally give or spend corporate resources for a candidate for federal office. All contributions by a PAC to candidates for federal office must be drawn entirely from voluntary personal donations.

(23) The empirical findings with regard to the NTB Equation 8 and the import penetration equation 10 are similar to those reported in G-B and are available from the authors upon request. The coefficient for ABATEMENTCOSTS1 in Equation 10 is negative and significant at the 10% level.

(24) Estimation of the LGC equation with the pollution abatement measure in levels leads to qualitatively similar results. The coefficient for ABATEMENTCOSTS1 in the LGC equation is positive and statistically significant at the 1% level. Moreover, using a three-equation system instead, with INTENSITY as an exogenous variable in the LGC equation rather than the endogenous ABATEMENTCOSTS1, results in an insignificant INTENSIFY coefficient. This may be explained by the fact that the simple correlation between INTENSITY and ABATEMENTCOSTS1 (ABATEMENTCOSTS2) is a relatively low 0.24 (0.35). The results are available upon request.

(25) The negative sign of the NTB protection coefficient is contrary to what theory predicts. However, this is not due a potentially strong correlation between our two policy measures. In effect, import protection and ABATEMENTCOSTS1 are only weakly correlated, with a partial correlation coefficient of 0.14. In addition, leaving out the EPS measure does not change the sign on NTB protection, as the results in G-B show. Instead of focusing on the sign of this particular coefficient, it is important to note that the elasticity of corporate PAC spending with respect to deadweight loss is positive, with a value of 0.34 (the elasticity is equal to the sum of the coefficient estimates on IMPORTPENETRAT and ELASTIC plus one half the estimate on NTB/[1 + NTB]). This means that a 10% increase in the deadweight loss from protection would lead to an increase in PAC contributions per firm of 3.4%. Furthermore, the implication of the G-H model that lobbying competition with downstream industries will increase lobbying efforts by upstream firms is also confirmed by the data. Both a strong demand by downstream users and greater concentration of downstream industries are found to have a positive effect on the lobbying contributions per upstream firm. However, only the first effect is statistically significant at the 1% level.

(26) One potential explanation for this result is that ABATEMENTCOSTS1 may be positively correlated with firm concentration in downstream industries, DOWNSTREAMHERFIN. However, a partial correlation coefficient of-0.017 indicates that this is not the case. Instead, the changes in size and significance level of both coefficient estimates appear mostly driven by the change in sample size (N = 89 here vs. N = 242 in G-B). This can be seen from the fact that when Equation 7 is estimated without an EPS variable, both variables remain statistically insignificant. Note that the EPS measure is weakly correlated with industry concentration, HERFIN (r = 0.14), and somewhat stronger with output usage by downstream industries, DOWNSTREAMSHARE (r = 0.22). However, even with EPS as an additional regressor, both variables still matter for lobby group contributions--they are statistically significant at the 7% and 5% level, respectively.

(27) The empirical results for the remaining equations (Eqns. 8 and 10) are available from the authors upon request.

(28) We thank one of our referees for noticing this potential problem, as well as the one addressed in the subsequent paragraph.

(29) To be precise, we remove the largest 25% of all industries, as measured by added value per firm. This reduces the sample using ABATEMENTCOSTS1 from 89 to 63 observations and the sample using ABATEMENTCOSTS2 from 177 to 133 observations. Removal of large-scale industries using employment per firm as measure of size yields quantitatively similar results (available upon request).

(30) For a similar method of proving sustainable cooperative outcomes, see Friedman (1990).

References

Abreu, D. 1986. Extremal equilibria of oligopoly supergames. Journal of Economic Theory 39:191-225.

Andres, G. J. 1985. Business involvement in campaign finance: Factors influencing the decision to form a corporate PAC. PS 18:156-81.

Ansolabehere, S., J. M. de Figueiredo, and J. M. Snyder, Jr. 2003. Why is there so little money in U.S. politics? Journal of Economic Perspectives 17:105-30.

Ansolabehere, S., J. M. Snyder, Jr., and M. Tripathi. 2002. Are PAC contributions and lobbying linked? New evidence from the 1995 lobby disclosure act. Business and Politics 4:131-55.

Bernheim, B. D., and M. D. Whinston. 1990. Multimarket contact and collusive behavior. Rand Journal of Economies 21:1-26.

Conconi, P., and C. Perroni. 2002. Issue linkage and issue tie-in in multilateral negotiations. Journal of International Economics 57:423-47.

Damania, R., and P. G. Fredriksson. 2000. On the formation of industry lobby groups. Journal of Economic Behavior and Organization 41:315-35.

Damania, R., and P. G. Fredriksson. 2003. Trade policy reform, endogenous lobby group formation, and environmental policy. Journal of Economic Behavior and Organization 52:47-69.

Eliste, P., and P. G. Fredriksson. 2002. Environmental regulations, transfers, and trade: Theory and evidence. Journal of Environmental Economics and Management 43:234-50.

Esty, D. C., and R. E. Caves. 1983. Market structure and political influence: New data on political expenditures, activity and success. Economic Inquiry 21:24-38.

Friedman, J. 1971. A non-cooperative equilibrium for supergames. Review of Economic Studies 38:1-12.

Friedman, J. 1990. Game theory with applications to economics. Oxford: Oxford University Press.

Gawande, K. 1997. US non-tariff barriers as privately provided public goods. Journal of Public Economics 64:61-81.

Gawande, K. 1998. Stigler-Olson lobbying behavior and organization: Evidence from a lobbying function. Journal of Economic Behavior and Organization 35:477-99.

Gawande, K., and U. Bandyopadhyay. 2000. Is protection for sale? Evidence on the Grossman-Helpman theory of endogenous protection. Review of Economics and Statistics 82:139-52.

Gawande, K., and P. Krishna. 2003. The political economy of trade policy: Empirical approaches. In Handbook of international trade, edited by J. Harrigan. London: Basil Blackwell, pp. 213-250.

Goldberg, P. K., and G. Maggi. 1999. Protection for sale: An empirical investigation. American Economic Review 89:1135-55.

Grier, K. B., M. C. Munger, and B. E. Roberts. 1991. The industrial organization of corporate political participation. Southern Economic Journal 57:727-38.

Grier, K. B., M. C. Munger, and B. E. Roberts. 1994. The determinants of industrial political activity, 1978-1986. American Political Science Review 88:911-26.

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

Hamilton, J. T. 1993. Politics and social costs: Estimating the impact of collective action on hazardous waste facilities. Rand Journal of Economics 24:101-25.

Heywood, J. S. 1988. The structural determinants of corporate campaign activity. Quarterly Review of Economics and Business 28:39-48.

Humphries, C. 1991. Corporations, PACs and the strategic link between contributions and lobbying activities. Western Political Quarterly 44:353-72.

Kelejian, H. 1971. Two stage least squares and econometric systems linear in parameters but nonlinear in the endogenous variables. Journal of the American Statistical Association 66:373-82.

Levinson, A., and M. S. Taylor. 2001. Trade and the environment: Unmasking the pollution haven hypothesis. Unpublished paper, Georgetown University.

List, J. A., and C. Y. Co. 2000. The effects of environmental regulation on foreign direct investment. Journal of Environmental Economics and Management 40:1-20.

Masters, M. F., and G. D. Keim. 1985. Determinants of PAC participation among large corporations. Journal of Politics 47:1158-73.

McKeown, T. J. 1994. The epidemiology of corporate PAC formation, 1975-1984. Journal of Economic Behavior and Organization 24:43-57.

Milyo, J., D. Primo, and T. Groseclose. 2000. Corporate PAC campaign contributions in perspective. Business and Politics 2:75-88.

Mitra, D. 1999. Endogenous lobby formation and endogenous protection: A long-run model of trade policy determination. American Economic Review 89:1116-34.

Olson, M. 1965. The logic of collective action. Cambridge, MA: Harvard University Press.

Pecorino, P. 1998. Is there a free-rider problem in lobbying? Endogenous tariffs, trigger strategies, and the number of firms. American Economic Review 88:652-60.

Peltzman, S. 1976. Toward a more general theory of regulation. Journal of Law and Economics, 19:211-40.

Persson, T., and G. Tabellini. 2000. Political economics: Explaining economic policy. Cambridge, MA: MIT Press.

Pittman, R. 1976. The effects of industry concentration and regulation on contribution in three 1972 U.S. senate campaigns. Public Choice 27:71-80.

Pittman, R. 1988. Rent-seeking and market structure: Comment. Public Choice 58:173-85.

Salamon, L. M., and J. J. Siegfried. 1977. Economic power and political influence: The impact of industry structure on public policy. American Political Science Review 71: 1026-43.

Shapiro, C. 1989. Chapter 6: Theories of oligopoly behavior. In Handbook of industrial economics, vol. 1, edited by R. Schmalensee and R Willig. Amsterdam: Elsevier Science, pp. 329-414

Sheills, C. R., R. F. Stem, and A. V. Deardorff. 1986. Estimates of the elasticities of substitution between imports and home goods for the United States. Weltwirtschaftliches Archiv 122:497-519.

Snyder, J. 1990. Campaign contributions as investment. Journal of Political Economy 98:1195-227.

Spagnolo, G. 2000. Issue linkage, delegation, and international policy cooperation. Unpublished paper, Stockholm School of Economics.

Stigler, G. 1971. The theory of economic regulation. Bell Journal of Economics and Management Sciences 2:3-21.

Stratmann, T. 1991. What do campaign contributions buy? Southern Economic Journal 57:606-20.

Stratmann, T. 1992. Are contributors rational? Untangling strategies of political action committees. Journal of Political Economy 100:647-64.

Stratmann, T. 1995. Campaign contributions and congressional voting: Does the timing of contributions matter? Review of Economics and Statistics 77:127-36.

Trefler, D. 1993. Trade liberalization and the theory of endogenous protection: An econometric study of U.S. import policy. Journal of Political Economy 101:138-60.

U.S. Department of Commerce, Bureau of the Census. 1983. Pollution abatement costs and expenditures. Current Industrial Reports. Washington, DC: GPO.

Vousden, N. 1990. The economics of trade protection. Cambridge, UK: Cambridge University Press.

Zardkoohi, A. 1985. On the political participation of the firm in the political process. Southern Economic Journal 51:804-17.

Richard Damania, * Per G. Fredriksson, ([dagger]) and Thomas Osang ([double dagger])

* Department of Economics, Adelaide University, Adelaide, Australia 5005; E-mail Richard.damania@ adelaide.edu.au.

([dagger]) Department of Economics, Southern Methodist University, P.O. Box 750496, Dallas, TX 75275-0496, USA; E-mail pfredrik@mail.smu.edu.

([double dagger]) Department of Economics, Southern Methodist University, P.O. Box 750496, Dallas, TX 75275-0496, USA; E-mail tosang@mail.smu.edu; corresponding author.

The seminal theory by Olson (1965) predicts that industries with fewer firms have a greater ability to undertake collective action. They organize cooperative political action more easily because greater concentration lowers the cost of political action. (1) The empirical evidence is inconclusive, however. Andres (1985), Masters and Keim (1985), Heywood (1988), and Humphries (1991) find positive effects of industry concentration on the probability of making political action committee (PAC) contributions (see also McKeown 1994). Pittman (1976) finds that concentrated industries generate greater contribution levels. Grier, Munger, and Roberts (1991) find an inverted U-shaped relationship between the level of PAC formation and industry concentration, with a maximum political participation rate occurring at a four-firm concentration ratio around 0.45. Esty and Caves (1983) and Zardkoohi (1985) report ambiguous effects of concentration on PAC contributions. (2)

In this paper we suggest an alternative perspective on firms' ability to organize collective action which, to our knowledge, has been ignored so far. The novel argument is that industries that face multiple regulations (a greater number of policy issues) find it easier to overcome collective action problems and sustain lobbying. In particular, we focus on the difference between firms in polluting and clean industry sectors. Using a simple repeated game framework similar to Spagnolo (2000), we argue that firms in industries that are naturally polluting (because of their input requirements), and therefore incur pollution abatement costs, will face an additional policy battle compared with other industries, everything else equal. (3) This enables such industries to sustain greater cooperation and lobbying. This is because firms seeking to form a lobby group face a free-riding problem due to a limited amount of "enforcement power" available to punish deviation and ensure cooperation. Firms that face multiple areas of regulation have an advantage in the formation of lobby groups because they have a greater amount of enforcement power available to reallocate between policy issues. When joint lobbying gives large gains in environmental policy, this surplus can be reallocated to trade policy, for example. Free-riding behavior on trade policy lobbying may more easily be disciplined. The prediction that emerges from our theoretical model is that polluting industries are relatively less affected by the free-riding problems involved in organizing political action, and we thus expect the level of political contributions to be higher in these sectors. (4)

We evaluate this prediction using a cross-section data set of U.S. manufacturing industries. Our empirical model builds on a multiple-equation model by Gawande and Bandyopadhyay (2000), who test the well-established theory of Grossman and Helpman (1994) on the pattern of protection (their theory takes lobby group formation as given). We augment Gawande and Bandyopadhyay's model with an additional equation for environmental policy stringency. (5) The empirical results lend support to our theory. Industry PAC contributions, and thus the level of lobby group cooperation, are greater in industries with larger pollution abatement costs. This result is robust to several measures of lobby group formation and environmental policy.

The present paper contributes to the recent literature on the formation of lobby groups. In the area of pollution taxation, Damania and Fredriksson (2000) argue that collusive industries may more easily form lobby groups that oppose such taxes. Using a related setup, Damania and Fredriksson (2003) discuss the effect of (exogenous) trade liberalization on environmental policy formation when lobby group formation is endogenous. Pecorino (1998) and Mitra (1999) discuss the formation of trade lobby groups. (6) Neither of these papers explore the relationship between collective action and the number of policy instruments, however. (7)

Empirical work is severely lacking in this area, although some related work does exists. Our paper complements Grier, Munger, and Roberts (1994), who argue that industries that potentially may benefit from government assistance contribute more in corporate PAC contributions but are hindered by collective action problems. Pittman (1988) shows that the level of federal regulations (primarily measured as the level of capital expenditures on pollution abatement induced by Environmental Protection Agency regulations) significantly determines campaign contributions. To our knowledge, no study addresses the influence of the number of regulations on the degree of political action, however.

The paper is organized as follows. Section 2 sets up a stylized model of the lobbying game. Section 3 describes the econometric model, and section 4 discusses the data. Sections 5 and 6 report the empirical results and the sensitivity analysis, respectively, while section 7 concludes.

2. A Theoretical Example

In order to illustrate our argument, we outline a stylized infinite horizon model with complete information, which makes use of the framework developed by Spagnolo (2000) in his study of linkages of environmental and trade policies in international agreements. The model will underline the reasons why we may expect pollution intensive sectors to have an easier time to sustain lobbying.

We have two industry sectors, k = A, B, which are identical except that sector A's production is polluting whereas sector B's production is nonpolluting. Each industry sector has two identical firms, h = a, b. Thus in total we have four firms, [k.sup.h]. All firms face at a minimum n government policies each denoted by i, that is, trade policy, corporate and wage taxes, etc. Because sector A firms are polluting, they in addition encounter an environmental policy, and these firms face n + 1 government policies. For simplicity we set n = 1, where the common government policy is a trade policy. Thus, policy issue i = 1 is trade policy, and policy issue i = 2 is environmental policy. (8) Both sectors thus produce tradable goods.

The two firms in each industry sector wish to organize joint trade policy lobbying. However, they face a prisoner's dilemma since when one firm contributes to the lobbying effort its rival has an incentive to deviate. The one-period strategic interaction on policy issue i is represented as a prisoner's dilemma in which each firm may choose to either contribute ([C.sub.i]), or free ride (defect) ([D.sub.i]). In industry k, the one-shot policy prisoner's dilemma game is characterized by firm h's action space [[THETA}.sup.k,h.sub.i] = {[C.sub.i], [D.sub.i]},

k[member of]{A, B}, h[member of]{a, b}, and policy payoff functions [[pi].sup.kh].sub.i]:[[THETA].sup.k.sub.i] [right arrow] [R.sup.2], where [[THETA].sup.k.sub.i] = [[THETA].sup.ka.sub.i] x [[THETA].sup.kb.sub.i]. This generates the following policy payoff matrix for policy i = 1 (trade policy) for the sector k firms, where [Y.sub.i] > [X.sub.i] > [N.sub.i] > [Z.sub.i], and [Y.sub.i] + [Z.sub.i] < 2[X.sub.i]: The two firms in sector A in addition face a similar prisoner's dilemma on policy i = 2 (environmental policy). In sector A, the payoffs from the policy outcomes on both policy issues enter firm h's aggregate static profit functions [[PI].sup.h.sub.A] = [PI] ([[pi].sub.1][[pi].sub.2], h = a, b, which are continuous and twice differentiable, with [differential][[PI].sup.h.sub.A]/[differential][[pi].sub.i] > 0, I = 1,2. In sector B, the payoff from the policy outcome on trade policy alone enters each firm's aggregate static profit functions [[PI].sup.h.sub.B] = [PI]([[pi].sub.i]), h = a,b, with [differential][[PI].sup.h.sub.B]/[differential][[pi].sub.i] > 0. In each period t firm h in sector k maximizes the net present value of profits from policy outcomes NP[V.sup.k,h] = [[summation].sup.[infinity].sub.[tau]=t] [[delta].sup.[tau]-t][[PI].sup.[tau]], where [delta] < 1 is the firms' discount factor. Our theoretical analysis ignores the policy maker entirely, for simplicity our focus is on the lobby groups only. We also disregard the impact on firm behavior of redistributed tariff or pollution tax revenue income in this analysis.

The problem is analyzed in a supergame, that is, the firms are assumed to interact over an indefinite period of time. We let firms use trigger strategies as discussed by Friedman (1971), and we focus on symmetric stationary lobbying sustained by stationary punishment strategies. First, assume that the trigger strategies apply only to one separate policy issue at a time. Thus, defection on trade policy is punished by abandoning all cooperation on trade policy by reverting to the one-shot Nash equilibrium. In sector A, taking environmental policy (policy 2) as given, the one-period gain from unilateral defection on trade policy lobbying only is given by

[G.sup.A.sub.1] = [[PI].sup.A]([Y.sub.1], [[pi].sub.2]) - [[PI].sup.A]([X.sub.1], [[pi].sub.2]) (l)

and the cost of defection is given by

[C.sup.A.sub.1] = [delta]/1 - [delta] [[[PI].sup.A][X.sub.1], [[pi].sub.2]) - [[PI].sup.A] [N.sub.1], [[pi].sub.2]). (2)

In the polluting sector A, joint lobbying on policy issue 1 is sustainable if [G.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.1]. Suppose instead that the firms in sector A punish a deviator in both policy dimensions. Thus, if a firm deviates, the rival firm's retribution strategy is punishment in both trade and environmental policies. In this case the deviating firm would chose to deviate on both policies (see Bernheim and Whinston 1990). Then, the gain from deviation is given by

[G.sup.A.sub.1,2] = [PI]([Y.sub.1], [Y.sub.2]) - [PI]([X.sub.1], [X.sub.2]), (3)

and the cost of deviation is given by

[C.sup.A.sub.1,2] = [delta]/1 - [delta][PI]([X.sub.1], [X.sub.2]) - [PI]([N.sub.1], [N.sub.2])]. (4)

In this case, joint lobbying on both policy issues is sustainable if [G.sup.A.sub.1,2] [less than or equal to] [C.sup.A.sub.1,2] In sector B the optimal punishment strategy involves deviation on policy 1 (trade policy), and thus joint lobbying requires [G.sup.B.sub.1] [less than or equal to] [C.sup.B.sub.1].

The focus of our discussion is the fact that firms seeking to form a lobby group may face a problem of limited enforcement power. The expected short-term gains from deviation from lobbying may be greater than the long-term gains from cooperation. In this case, lobbying is not sustainable. However, firms that interact in multiple areas of regulation (policy issues) have an advantage in the formation of lobby groups due to the greater amount of available enforcement power that can be reallocated between policy issues. When cooperation gives large gains in one policy area such as environmental policy, this slack can be reallocated to enforce cooperation in another such as trade policy. The slack of expected gain from cooperation on environmental policy may then be used to discipline behavior on trade policy, or vice versa. The slack available in the environmental policy area will depend on the pollution intensity (pollution abatement costs) of the sector in question because this reflects the amount at stake (i.e., the costs of regulation). In this theoretical example we attempt to show this in the simplest possible manner. Thus, we follow Bernheim and Whinston (1990) by assuming that sector A firms' profit functions are linearly separable in policy issues. (9)

ASSUMPTION 1. In sector A, each firm has a profit function that is linearly separable in the two policies 1 and 2, such that [[PI].sup.A] = [[PI].sup.A.sub.1] ([[pi].sub.1]) + [[PI].sup.A.sub.2] ([[pi].sub.2]).

Thus, firm h in sector A will maximize the net present value of profits by maximizing NP[V.sup.A,h] = [[summation].sup.[infinity].sub.[tau]=1] [[delta].sup.[tau]-t] [[[PI].sup.A,h.sub.1]([[pi].sub.1]) + [[PI].sup.A,h.sub.2]([[pi].sub.2])]. Assumption 1 also implies that if the firms in sector A use an optimal punishment strategy by canceling cooperation on both policy issues in order to deter deviations, the two separate lobbying games played now collapse into one game. In industry A this game has a one-period static payoff function [[PI].sup.A.sub.1,2] = [[PI].sup.A.sub.1] ([[pi].sub.1]) + [[PI].sup.A.sub.2]([[pi].sub.2]). On the other hand, in industry B firms lobby on only one policy. The one-period payoff function is [[PI].sup.B.sub.1] = [[PI].sup.B.sub.1]([[pi].sub.1]). We can now state the following result.

RESULT 1. Under Assumption 1, the firms in the polluting sector A have an equal or greater ability to sustain cooperation on lobbying than the firms in the clean sector B.

PROOF. See Appendix 1.

Cooperation on trade and environmental policy lobbying is sustainable if

[G.sup.A.sub.1] + [G.sup.A.sub.2] [less than or equal to] [C.sup.A.sub.1] + [C.sup.A.sub.2], (5)

where subscript 1(2) denotes trade (environmental) policy. We identify four main cases.

CASE 1. [G.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.1] and [G.sup.A.sub.2] [less than or equal] [C.sup.A.sub.2]. Condition 5 holds, and being a polluting industry does not affect industry A's ability to sustain lobbying. It is sustainable in any case.

CASE 2. [G.sup.A.sub.1] > [C.sup.A.sub.1] and [G.sup.A.sub.2] > [C.sup.A.sub.2]. Then

[G.sup.A.sub.1] + [G.sup.A.sub.2] > [C.sup.A.sub.1] + [C.sup.A.sub.2] (6)

and cooperation is not sustainable in sector A on any policy issue. Interaction in the environmental policy arena confers no advantage to industry A.

CASE 3A. [G.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.1] and [G.sup.A.sub.2] > [C.sup.A.sub.2]. A polluting industry is able to cooperate on lobbying if [G.sup.A.sub.2] - [C.sup.A.sub.2] [less than or equal to] [C.sup.A.sub.1] - [G.sup.A.sub.1], that is, if the cost of deviation on the trade policy is greater than the gain from deviation on the environmental policy issue.

CASE 3B. [G.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.1] and [G.sup.A.sub.2] > [C.sup.A.sub.2], where [G.sup.A.sub.2] - [C.sup.A.sub.2] > [C.sup.A.sub.1] - [G.sup.A.sub.1]. The firms in sector A are able to cooperate on trade policy only, given that the two policies are treated as separate games.

CASE 4A. [G.sup.A.sub.1] > [C.sup.A.sub.1] and [G.sup.A.sub.2] [less than or equal] [C.sup.A.sub.2]. Industry A is able to cooperate on lobbying on both policy issues because of the existence of environmental regulations, if [G.sup.A.sub.1] - [C.sup.A.sub.1] [less than or equal to] [C.sup.A.sub.2] - [G.sup.A.sub.2]. The benefit of cooperation on environmental policy is greater than the gain from deviation on trade policy. In this case, lobby group formation is feasible only in polluting sectors.

CASE 4B. [G.sup.A.sub.1] > [C.sup.A.sub.1] and [G.sup.A.sub.2] [less than or equal to] [C.sup.A.sub.2]. If [G.sup.A.sub.1] - [C.sup.A.sub.1] > [G.sup.A.sub.2] - [G.sup.A.sub.2]. Only environmental policy lobbying cooperation is possible, given that the two policies are treated as separate games.

In sum, cooperation on both policy dimensions is sustainable as long as the gains from deviation on policy i (that is [G.sup.A.sub.1] - [C.sup.A.sub.1], i = 1,2) are no greater than the gains from cooperation on policy j (that is [C.sup.A.sub.j] - [G.sup.A.sub.j], j [not equal to] i). Thus even if cooperation is not sustainable on policy issue i in isolation, it may be sustainable on both issues as long is there is some enforcement slack available on policy issue j.

For expositional ease these results have been derived on the assumption that trade and environmental policy are linearly independent in their effects on profits (Assumption 1). It is recognized that this assumption may in some cases not hold in practice. However, as shown by Spagnolo (2000) for the case of issue linkage in international agreements, the results are the same whether the policy instruments under consideration are linearly interdependent or strategic substitutes.

From the above discussion, it follows that a polluting industry never faces greater difficulties in organizing lobbying than it would have faced had it been a clean industry. On the other hand, there are circumstances where it is favorable for lobbying to be subject to environmental regulations, in particular when extra enforcement power (or slack in gains from cooperation) is available in the environmental policy area. This extra enforcement power can then be used, for example, in the trade policy area.

Note that although the discussion thus far has (for simplicity) been framed in terms of environmental and trade policies, it also applies to other policy areas such as taxation, subsidies, and other forms of regulations. Thus, the results are in reality more general, and we therefore formulate the following prediction in a more general fashion:

PREDICTION 1. Firms operating in sectors with multiple regulations sustain equal or greater levels of political contributions compared with firms operating in sectors with only one regulation, ceteris paribus.

This prediction is the subject of our empirical testing in the next section. (10) If, as is generally assumed in the political economy literature, the level of protection is increasing in political contributions, it follows that firms in polluting industries will obtain greater protection than those in clean industries. (11) For expositional ease we have ignored a number of other theoretical issues. Perhaps the most important of these is the assumption that firms sustain lobbying by the threat of reversion to the one-shot Nash equilibrium. Abreu (1986) has shown that in general the one-shot Nash equilibrium is not the harshest available credible punishment. In this paper we have followed most of the supergame literature and have ignored this issue for simplicity. However, the qualitative results are unlikely to be affected by this assumption, since the results hinge on the observation that for any given retribution strategy, enforcement slack in one domain can be used to enforce cooperation on some other issue. Another problem that we have ignored is that of the renegotiation proofness of the equilibrium. As noted by Shapiro (1989), the restriction that equilibria be renegotiation proof would shrink the set of outcomes where lobbying (i.e., cooperation) is sustainable. Thus, so long as renegotiation proof equilibria exist, this restriction would not alter the fundamental argument that interaction on numerous issues may facilitate greater cooperation. Finally, it is worth noting that the equilibria identified in this paper emerge from incentive compatible noncooperative actions of players. Hence the conclusions do not depend upon the existence of a formal industry lobby group, in much the same way as the tacitly collusive equilibria in repeated games do not presume the existence of a cartel. Our formal analysis simply reveals that when firms interact in multiple areas of regulation, their contributions to lobbying activity (which has public good characteristics) will be greater than if they interact on only one policy issue, irrespective of whether the lobbying occurs through an industry organization or by firms individually. (12)

3. Econometric Specification

Our primary focus in the empirical work is on the prediction generated by our theoretical example: We investigate whether environmental regulations and the amount of lobby group contributions are positively correlated across U.S. manufacturing sectors.

Our empirical model builds on the model estimated in Gawande and Bandyopadhyay (G-B) (2000), which provides structural estimations of the theoretical model of trade policy formation by Grossman and Helpman (G-H) (1994) (see Goldberg and Maggi 1999 for an alternative test of the G-H model). The G-H theory generates predictions of trade policy outcomes across industries, taking the existence of lobby groups across industries simply as given, and ignoring environmental policy. The G-H theory argues that given the existence of a lobby, the equilibrium lobby group contribution equals the welfare loss caused by the lobby's participation in the political equilibrium. In essence, seen in this light our theory suggests that polluting sectors will have an easier time generating sufficient funds to compensate the government for any induced trade policy distortions. In addition, these firms must also contribute the funds necessary to compensate the government for the environmental policy distortions associated with lobbying.

In G-B's empirical implementation of the G-H theory (i) lobby group contributions and (ii) the degree of import penetration are jointly determined with (iii) the level of protection, thus accounting for the endogeneity of three variables. Our empirical model extends the literature by adding a fourth endogenous variable: environmental policy stringency. Our empirical model thus consists of four equations. Since three of these equations are discussed in detail in G-B, we here restrict ourselves to a brief overview of each equation and its main features, highlighting the differences to the previous approach.

The first equation, the lobby group contributions (LGC) equation, explains the lobbying efforts by firms (CONTRIBUTIONS). (13) G-B builds on Vousden (1990) to derive an equation reflecting the lobby contributions necessary to compensate the government for the deadweight loss from protection. The equation includes the level of nontariff barriers (NTB/[1 + NTB]), the import elasticity (ELASTIC), and the import penetration ratio (IMPORTPENETRAT). The greater any of these variables, the larger the deadweight loss, and the greater the political spending needed to keep the government indifferent between free trade and protection. In addition, three variables account for the conflict of interest between downstream users, who oppose upstream protectionism and upstream firms (Olson 1965). These are the firm concentration in upstream industries (HERFIN) and downstream industries (DOWNSTREAMHERFIN), as well as the share of upstream producers' output sold downstream as intermediate inputs (DOWNSTREAMSHARE). The higher these measures are, the greater the bargaining power of downstream users, forcing upstream users to make greater political contributions.

The novel feature of our specification of the LGC equation is the inclusion of pollution abatement costs (ABATEMENTCOSTS1) as an additional regressor. (14) Our theoretical model suggests that pollution abatement costs and lobby group contributions (lobby group organization effort) by firms should be positively correlated. When the industry faces a greater number of policy battles, the overall loss from deviation in lobbying are larger, and potentially more enforcement power can be reallocated to trade policy. The higher the pollution intensity and thus cost of pollution abatement, ceteris paribus, the more is at stake for a polluting firm, the more enforcement power should be available. Thus, the generation of political funds by an industry lobby group becomes easier.

The NTB equation, determines the level of protection (NTB/[1 + NTB]). The equation is the empirical counterpart of the closed-form solution for the level of protection derived by G-H. (15) The advantage of a strictly theory-driven specification is that the determinants of protection as well as the interaction terms that should be included in the equation are determined a priori. (16)

The environmental policy stringency (EPS) equation determines the stringency of environmental regulation approximated by the level of pollution abatement cost (ABATEMENTCOSTS1). We include NTB protection as a right-hand side (RHS) variable in the EPS equation. If environmental and trade policies are determined by similar political forces (accounting for the same RHS variables as in Eqn. 8), the two variables will be negatively correlated. If, on the other hand, policy makers tend to compensate industries for low trade barriers (stringent environmental policy) with less stringent environmental policy (greater trade barriers), a positive coefficient is expected (Eliste and Fredriksson 2002). We also include political spending by firms as a regressor, since more intense lobbying should result in less stringent environmental standards. Finally, we include a measure of an industry's pollution intensity as an exogenous regressor (INTENSITY).

The final equation explains the level of import penetration (IMPORTPENETRAT). (17) Since the stringency of environmental regulation should affect the competitiveness of domestic firms, we include a measure of pollution abatement cost in our specification of the import penetration equation.

The system of equations that we estimate can be written as

ln CONTRIBUTIONS = [[alpha].sub.0] + [[alpha].sub.1] ln HERFIN + [[alpha].sub.2] In IMPORTPENETRAT + [[alpha].sub.3] ln (NTB/1 + NTB) + [[alpha].sub.4] ln ELASTIC + [[alpha].sub.5] ln DOWNSTREAMSHARE + [[alpha].sub.6] ln DOWNSTREAMHERFIN + [[alpha].sub.7] ln ABATEMENTCOSTS1 + [[epsilon].sub.1], (7)

NTB/1 + NTB = [[beta].sub.0] + [[beta].sub.1] [(IMP/CONS).sup.-1]/ELASTIC + [[beta].sub.2]ORGANIZED[[(IMPORTPENETRAT).sup-1]/ELASTIC] + [[beta].sub.3]INTERMTARIFF + [[beta].sub.4]INTERMNTB + [[epsilon].sub.2], (8)

ABATEMENTCOSTS1 = [[delta].sub.0] + [[delta].sub.1] NTB/1 + NTB + [[delta].sub.2] CONTRIBUTIONS + [[delta].sub.3]INTENSITY + [[epsilon].sub.3], (9)

IMP/CONS = [[gamma].sub.0] + [[gamma].sub.1] NTB/1 + NTB + [[gamma].sub.2] ABATEMENTCOSTS1 + X[GAMMA] + [[epsilon].sub.4], (10)

where X is a (row) vector of exogenous variables, and [GAMMA] is the corresponding (column) vector of coefficients. (18) The error terms are assumed to be normal and independently distributed across observations. A description of all variables used is given in Table A1 in the Appendix.

In addition to the endogenous left-hand side variables, in CONTRIBUTIONS, NTB/(1 + NTB), IMPORTPENETRAT, and ABATEMENTCOSTS1, the system of Equation 7 includes nonlinear functions of some of these variables on the right-hand side: in (NTB/[1 + NTB]), In IMPORTPENETRAT, and in ABATEMENTCOSTS1 in Equation 7, and ORGANIZED in Equation 8. (19) To account for these nonlinearities, we use a modified version of the two-stage least squares estimator proposed by Kelejian (1971) (see also G-B). In the first stage, we regress each of the endogenous variables and their transformations on the set of exogenous variables. In addition to the linear terms, we admit squared terms and a number of cross product terms. (20) In the second stage, we estimate Equation 7 but replace the endogenous right-hand side variables by their predicted values from the first-stage regressions.

4. Data

Our data are at the four-digit Standard Industrial Classification (SIC) level, and the year is 1983 unless noted differently. (21) Summary statistics are given in Table A2 in the Appendix. All data except for pollution abatement costs and pollution intensity are from G-B. The following data description is brief, since a detailed description of how each variable was constructed is found there.

Political action committee (PAC) contributions cover the four Congressional election cycles, 1977-78, 1979-80, 1981-82, and 1983-84, were constructed by Gawande (1998) from Federal Election Commission tapes. (22) While we acknowledge that individual firms may sometimes have different political objectives, we follow the extant empirical literature by assuming that the contributions made by corporate PACs may be assumed to be a public good. This is also consistent with a large share of the theoretical and empirical literature on the determination of trade and environmental policies.

Unfortunately, there are no direct measures of EPS (such as regulation dummies, etc.) available at detailed levels of disaggregation. Instead, we use two indirect measures of pollution abatement costs, scaled by added value: capital expenditures (ABATEMENTCOSTS1) and operating costs (ABATEMENTCOSTS2). Both measures are available from the Pollution Abatement Costs and Expenditures Survey (U.S. Department of Commerce 1983). With ABATEMENTCOSTS1 as the EPS measure, the size of the data set is 89, compared with 177 when ABATEMENTCOSTS2 is used. When using lagged instead of current pollution abatement measures, the sample sizes are 94 and 155, respectively. Our measure of an industry's pollution intensity (INTENSITY) is taken from List and Co (2000) who identify pollution intensity at the two-digit level. We replicate their results at the four-digit level.

The political organization dummy (ORGANIZED) is based on a regression of PAC contributions per firm (scaled by value added) on bilateral import penetration by partner country interacted with 20 two-digit SIC dummies. There are five partner countries: France, Germany, Italy, Japan, and the UK. Two-digit industries with positive coefficients are considered organized. The union of all organized industries from the five regressions constitutes the set of organized industries (at the four-digit SIC level). In the original data set used by G-B, 165 out of 242 industries are organized (68.2%).

As is typical in the empirical literature on endogenous protection, an aggregate NTB coverage ratio is used to measure the level of protection. The inverse import penetration ratio is measured as consumption divided by imports and then scaled by 10,000 to account for small values of imports in some industries. The own- and cross-price elasticity measures are taken from Sheills, Stem, and Deardorff (1986), purged of their inherent errors-in-variables problem, and reproduced at the four-digit SIC level. Industry characteristics, added value, and the Herfindahl index were constructed from the 1982 census of manufacturing and, if necessary, from various annual surveys of manufactures. Protection levels in intermediate industries as well as the concentration measure in downstream industries were constructed from U.S. input-output tables.

5. Empirical Results

Table 1 contains 2SLS estimates of the LGC and the EPS equations (Eqns. 7 and 9, respectively). (23) We use notation similar to G-B, as well as G-B's convention for reporting significance levels. In particular, estimates with a t statistic greater than unity are marked with an asterisk since their inclusion in the regression equation increases the value of the adjusted [R.sup.2]. The most striking result of Table 1 is that the coefficient estimate for ABATEMENTCOSTS1 has the expected positive sign in the LGC equation. This lends support to our theory. In addition, the precision of the estimated relationship is high--the estimated coefficient is statistically significant at the 1% level. (24) The remaining coefficient estimates in Equation 7 all have the same signs as in G-B. (25) In addition, the size of the estimated coefficients is similar as well, with two exceptions. The coefficient estimates for import penetration, In IMPORTPENETRAT, and downstream industry concentration, In DOWNSTREAMHERFIN, are smaller than in G-B and no longer significant. (26) The drop in sample size as a result of the inclusion of ABATEMENTCOSTS1 as a regressor in Equation 7 raises the issue of sample selection bias. Given the total number of four-digit SIC industries (N = 448), even the sample size used in G-B (N = 242) is not immune to this problem. More importantly, most of our coefficient estimates are very close in magnitude to those reported in G-B, while our t statistics are often lower. Thus, instead of sample selection bias, the main consequence of the smaller sample seems to be the efficiency loss of the estimation.

With regard to the EPS Equation 9, we find that trade barriers and environmental policy stringency are positively correlated. In addition, PAC contributions and environmental regulation are positively correlated, contrary to the expected sign. This may be explained in the following way. Polluting industries raise greater levels of PAC contributions but devote these funds primarily to other policy areas. Finally, industries that are more pollution intensive have higher levels of pollution abatement expenditures. Except for the intercept, all coefficient estimates are statistically significant at the 1% level.

To provide an alternative test of our theoretical prediction, we replace pollution abatement capital expenditures (ABATEMENTCOSTS1) by operating costs (ABATEMENTCOSTS2) as the measure of environmental policy stringency. One can argue that ABATEMENTCOSTS2 is a better measure of abatement spending, and thus policy stringency, than ABATEMENTCOSTS1 since it is easier for firms to identify the environmental portion of their total operating costs than the environmental fraction of their total capital expenditures, in particular when more and more abatement involves process changes or is an integral part of new technologies. In our case, using ABATEMENTCOSTS2 instead of ABATEMENTCOSTS1 does not lead to qualitatively different results with regard to our hypothesis (see Table 2). The operating cost measure is positive and significant around the 10% level in the LGC equation. The size of the estimated relation is different, though. The substantially smaller coefficient estimate for ABATEMENTCOSTS2 compared with ABATEMENTCOSTS1 indicates that the impact of environmental regulation on lobby group formation and protection may only be one-third as strong as suggested by the previous measure of policy stringency. The remaining coefficient estimates in both equations are in line with their corresponding estimates from Table 1. The overall fit of the LGC estimation regression (0.13) is lower than in Table 1 (0.36), despite the larger sample size (N = 177 in Table 2 compared with N = 89 in Table 1).

6. Sensitivity Analysis

We perform a number of sensitivity tests. First, we use lagged instead of current values for both ABATEMENTCOSTS1 and ABATEMENTCOSTS2. This reduces the system to three equations since the potential endogeneity of the environmental stringency variable is no longer an issue. We report the results for the LGC equation in Table 3.27 The results confirm our previous findings about the impact of environmental stringency on lobby contributions (formation). The positive coefficient estimate on lagged ABATEMENTCOSTS1 is significant at the 1% level. This not only confirms the prediction of our theoretical model but also provides indirect evidence for the goodness of our pollution intensity measure as an instrumental variable for environmental stringency. In contrast to lagged capital expenditures, lagged pollution abatement operating costs (ABATEMENTCOSTS2) have no effect on lobby group contributions. Combined with the findings from Table 2, the result implies that most of the impact of EPS (as measured by ABATEMENTCOSTS2) on lobby group contributions is contemporaneous.

Second, we investigate a potential endogeneity bias that may arise due to the use of added value as both a scale measure in Equation 10 and to standardize lobbying efforts by firm (CONTRIBUTIONS) and pollution abatement costs (abatement capital expenditures, as well as abatement operating costs). (28) We instead use the number of employees per firm (FIRMSCALE2) as an alternative measure of industry size in Equation 10. Reestimating the system of equations (Eqns. 7-10), we find no evidence of such a bias, as far as our main results are concerned. A comparison between Table 1 and the results presented in column 1 of Table 4 yields only minor changes in the point estimates in either equation. In particular, our previous finding of a positive and highly significant effect of pollution intensity (INTENSITY) on lobby group contributions remains unchanged. Similarly, when we contrast the results from column 3 in Table 4 with those reported in Table 2 we once again find no substantive changes in point estimates or t statistics. Most importantly, the coefficient estimate of the operating cost abatement measure (ABATEMENTCOSTS2) remains around 0.3, and the significance level falls only slightly.

Next, we investigate a potential bias of our environmental policy stringency estimates due to the special status of industries made up of large-scale firms. Owing to their size, larger firms may interact with government agencies in several dimensions. They may therefore be more likely to lobby the government for reasons such as tax breaks, subsidies, or changes in certain types of (nonenvironmental) regulations. To test the robustness of our results, we therefore remove the large-scale industries from each sample and reestimate Equations 7-10.29 The results reported in columns 2 and 4 of Table 4 show that the effect of environmental policy stringency on lobby group contributions remains positive and significant. In fact, omitting the largest industries from the sample strengthen the results. The point estimate is about 50% larger in both cases (1.22 and 0.55, compared with 0.82 and 0.32 in Tables 1 and 2, respectively), and the precision of the estimated relationship has improved markedly. In addition, the overall fit of the policy stringency equation reported in columns 2 and 4 of Table 4 (Eqn. 9) has improved, with an adjusted [R.sup.2] above 0.3 compared with 0.14 and 0.18 in Table 1 and 2, respectively. Note also that some coefficient estimates, in particular those controlling for concentration in downstream industries in the LGC equation, become smaller and statistically insignificant in column 2. We thus conclude that controlling for the potential additional interactions between large firms and the government does not alter the main conclusions emerging from our previous empirical analysis.

Our final robustness checks involve estimations of a number of different specifications for the system of equations such as semilog of Equation 7, additional regressors (i.e., the degree of unionization in Eqns. 7 and 8), and a discrete measure of CONTRIBUTIONS in Equation 9. None of these alternative specifications yield results that are qualitatively different from the results presented in Tables 1 to 3. In effect, if the two measures of downstream lobbying competition used as regressors in Equation 7 are used in Equation 9 instead, the coefficient estimate on In ABATEMENTCOSTS2 in Equation 7 is positive and significant at the 1% level (t value 2.66).

7. Conclusion

This paper proposes a new theory of lobby group formation. We argue that firms in industry sectors encountering a greater number of policy instruments more easily organize lobby groups because of a greater amount of available "enforcement power." When lobby group cooperation gives large gains in one policy area, this surplus can be reallocated to another policy area. This causes an advantage for pollution intensive firms. Any slack of expected gain from cooperation on environmental policy may be used to discipline lobbying behavior on, for example, trade policy. Firms in polluting sectors facing stiffer environmental policy should therefore be able to raise greater amounts of PAC contributions.

Our theoretical prediction finds support in the data. Sectors with greater pollution abatement expenditures and thus more stringent levels of environmental protection have significantly greater levels of PAC contributions. This result is robust to several different variable measures and specifications. To our knowledge, this is a novel finding in the literature.

Appendix

PROOF OF RESULT 1. Note that under Assumption 1, when the punishment strategy calls for deviation in only one policy instrument, sector A firms have [G.sup.A.sub.i] = [[PI].sup.A.sub.i]([Y.sub.i]) - [[PI].sup.A.sub.i]([X.sub.i]), and [C.sup.A.sub.i] = [delta]/1 - [delta][[[PI].sup.A.sub.i] ([X.sub.i]) - [[PI].sup.A.sub.i]([N.sub.i])], i = 1, 2. Thus, cooperation is sustainable if

[[PI].sup.A.sub.i]([Y.sub.i]) - [[PI].sup.A.sub.i]([X.sub.i]) [less than or equal to] [[delta]/[1 - [delta]]] [[PI].sup.A.sub.i]([X.sub.i]) - [[PI].sup.A.sub.i]([N.sub.i]) (A1)

Rearranging condition AI it follows that cooperation is feasible for all values of the discount factor such that

[delta] [greater than or equal to] [[[PI].sup.A.sub.i] ([Y.sub.i]) - [[PI].sup.A.sub.i]([X.sub.i])]/[[PI].sup.A.sub.i] ([Y.sub.i]) - [[PI].sup.A.sub.i]([N.sub.i]) [equivalent to] [[DELTA].sub.i] (A2)

On the other hand, firms in sector A have the option of using a punishment strategy that entails reverting to the noncooperative strategy on both policies after any form of deviation. Given these two alternatives, a sector A firm's optimal deviation is a deviation on both policy issues simultaneously. The gain from defecting unilaterally from lobbying cooperation is given by [G.sup.A.sub.1,2] = [G.sup.A.sub.1] + [G.sup.A.sub.2], and the cost of defection is given by [C.sup.A.sub.1,2] = [C.sup.A.sub.1] + [C.sup.A.sub.2]. Cooperation on trade and environmental policy lobbying is sustainable if

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (A3)

Rearranging condition A3 it follows that cooperation is feasible for all values of the discount factor such that

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (A4)

Without loss of generality, suppose that when there is cooperation with only one policy instrument this involves interaction over policy i = 1 (i.e., trade policy). Clearly cooperation is more easily sustained with interaction over two policies if [[DELTA].sub.1,2] < [[DELTA].sub.1]. Substituting from conditions A2 and A4 and rearranging this inequality yields

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (A5)

where [OMEGA] = [[PI].sup.A.sub.1] ([N.sub.1]) - [[PI].sup.A.sub.2]([X.sub.2]) + [[PI].sup.A.sub.2]([N.sub.2]) ([[PI].sup.A.sub.1]([X.sub.1]) - [[PI].sup.A.sub.1]([Y.sub.1]) < 0, and the sign of [OMEGA] follows from the assumption that ([[PI].sup.A.sub.i]([X.sub.i]) < [[PI].sup.A.sub.i] ([Y.sub.i]), [for all] i = 1,2. It follows that with interaction over two policy issues, cooperation is feasible over a larger range of values of the discount factor. Hence, lobbying is easier to sustain in sectors encountering two policies. (30) QED.

Table A1. Descriptions of Variables used in the Econometric Analysis NTB NTB coverage ratio CONTRIBUTIONS PAC contributions per firm scaled by added value [(IMPORTPENETRAT).sup.-1] Inverse import penetration ratio, scaled by 10,000 (=[consumption/imports]/ 10,000) ELASTIC Own-price elasticity of imports, corrected for errors in variables ORGANIZED Political organization dummy (=1 if industry is organized, 0 otherwise) INTERMTARIFF Average tariff on intermediate goods used in an industry INTERMNTB Average NTB coverage ratio of intermediate goods used in an industry ABATEMENTCOSTS1 Pollution abatement capital expenditures, scaled by value added ABATEMENTCOSTS2 Pollution abatement operating costs, scaled by value added HERFIN Herfindahl index of firm concentration IMPORTPENETRAT Import penetration ratio: imports/ consumption DOWNSTREAMSHARE Percentage of an industry's shipments used as intermediate goods in downstream industries DOWNSTREAMHERFIN Intermediate-goods-output buyer concentration FIRMSCALE Measure of industry scale: value added per firm FIRMSCALE2 Measure of industry scale: employment per firm CONC4 Four-firm concentration ratio UNSKILLED Fraction of employees classified as unskilled SCIENTISTS Fraction of employees classified as scientists and engineers MANAGERS Fraction of employees classified as managerial REALELASTIC Real exchange rate elasticity of imports CROSSELAST Cross-price elasticity between domestic and imported goods, corrected for errors in variables TARIFF Post-Tokyo round ad valorem tariffs (ratio) [D.sub.g], g = 1, ..., 4 Dummies for four industry groups: food processing, resource intensive, general manufacturing, and capital intensive [(KIL).sub.g], g = 1, ..., 4 Capital labor ratio x [D.sub.g] INTENSITY Dummy for pollution intensity (=1 if industry is pollution intensive) Table A2. Summary Statistics (a) Standard Variable Mean Deviation Minimum Maximum NTB/(1 + NTB) 0.076 0.132 0 0.5 CONTRIBUTIONS 0.0275 0.0317 0.0017 0.2374 [(IMPORTPENETRAT).sup.-1] 0.0133 0.0772 0.0002 1.0 ABATEMENTCOSTS1 0.0057 0.011 0.0001 0.0781 ABATEMENTCOSTS2 0.0128 0.0207 0.0002 0.1714 ORGANIZED 0.6441 0.4802 0 1 ELASTIC 1.4865 0.3784 0.5491 2.1297 INTERMTARIFF 0.0523 0.0245 0.0116 0.1723 INTERMNTB 0.2165 0.1354 0.0226 0.6785 HERFIN 0.0753 0.069 0.0014 0.295 DOWNSTREAMSHARE 0.557 0.2898 0.0122 0.9641 DOWNSTREAMHERFIN 0.2325 0.1762 0.0448 0.7065 SCIENTISTS 0.042 0.0421 0 0.1667 MANAGERS 0.1009 0.0376 0 0.1807 UNSKILLED 0.0653 0.0475 0 0.3333 CONC4 0.3958 0.2114 0.0598 0.9883 FIRMSCALE 0.0115 0.0183 0.0002 0.1023 TARIFF 0.053 0.0451 0 0.3241 CROSSELAST -0.0602 0.8437 -1.7941 2.8864 REALELASTIC -0.9839 0.4795 -2.02 1.9 INTENSITY 0.3898 0.4891 0 1 (a) All data are from 1983. The sample size is 177 except for ABATEMENTCOSTS1, which for 1983 is available for only 89 four-digit SIC industries. Table 1. 2SLS Estimates: Lobby Group Contributions and Environmental Policy Stringency Equations (ABATEMENTCOSTS1) Lobby Group Environmental Contributions Policy Stringency Dependent Variable In CONTRIBUTIONS ABATEMENTCOSTS1 In HERF 0.18 * (1.85) In IMPORTPENETRAT 0.07 (0.89) In (NTB/[I + NTB]) -0.07 ** (2.85) In ELASTIC 0.31 * (1.03) In DOWNSTREAMSHARE 0.31 ** (2.03) In DOWNSTREAMHERFIN 0.16 * (1.25) In ABATEMENTCOSTS1 0.82 ** (4.33) NTB/(1 + NTB) 0.02 ** (2.05) CONTRIBUTIONS 0.10 ** (2.85) INTENSITY 0.01 ** (2.20) CONSTANT -1.55 ** (2.73) -0.001 (0.30) N 89 89 K 8 4 Adjusted [R.sup.2] 0.36 0.14 Model F 8.00 5.65 ln L -108.36 284.53 t statistics in parentheses. N, size of data set; K, number of regressors. * 2 > [absolute value of t] [greater than or equal to] 1. ** [absolute value of t] [greater than or equal to] 2. Table 2. 2SLS Estimates: Lobby Group Contributions and Environmental Policy Stringency Equations (ABATEMENTCOSTS2) Lobby Group Environ. Policy Contributions Stringency Dependent Variable ln CONTRIBUTIONS ABATEMENTCOSTS2 ln HERFIN 0.12 * (1.63) ln IMPORTPENETRAT 0.13 ** (2.09) ln (NTB/[l + NTB]) -0.07 ** (2.75) ln ELASTIC 0.45 * (1.80) ln DOWNSTREAMSHARE 0.25 ** (2.13) ln DOWNSTREAMHERFIN 0.09 (0.90) ln ABATEMENTCOSTS2 0.32 * (1.60) NTB/(1 + NTB) 0.02 * (1.88) CONTRIBUTIONS 0.21 ** (3.48) INTENSITY 0.01 ** (4.54) CONSTANT -2.90 ** (6.40) -0.000 (0.007) N 177 177 K 8 4 Adjusted [R.sup.2] 0.13 0.18 Model F 4.68 13.99 In L -238.39 455.20 t statistics in parentheses. N, size of data set; K, number of regressors. * 2 > [absolute value of t] [greater than or equal to] 1. ** [absolute value of t] [greater than or equal to] 2. Table 3. Sensitivity Analysis I: Lobby Group Contributions with Lagged Environmental Policy Stringency Dependent Variable ln CONTRIBUTIONS ln CONTRIBUTIONS ln HERFIN 0.13 * (1.24) 0.18 ** (2.56) ln IMPORTPENETRAT 0.09 * (1.24) 0.15 ** (2.47) ln (NTB/[l + NTB]) -0.09 ** (3.22) -0.08 ** (3.25) ln ELASTIC 0.73 ** (2.07) 0.57 ** (2.33) ln DOWNSTREAMSHARE 0.27 * (1.64) 0.26 ** (2.16) ln DOWNSTREAMHERFIN 0.04 (0.29) 0.13 * (1.28) ln [(ABATEMENTCOSTS1).sub.t-1] 0.58 ** (2.63) ln [(ABATEMENTCOSTS2).sub.t-1] 0.16 * (1.00) CONSTANT -2.68 ** (4.42) -3.09 ** (7.29) N 94 155 K 8 8 Adjusted [R.sup.2] 0.23 0.18 Model F 4.95 5.84 In L -123.35 -196.27 t statistics in parentheses. N, size of data set; K, number of regressors. * 2 > [absolute value of t] [greater than or equal to] 1. ** [absolute value of t] [greater than or equal to] 2. Table 4. Sensitivity Analysis II: Alternative Scale Measure and Adjustment for Industry Size Policy Stringency Measure ABATEMENTCOSTS1 Lobby Group Contributions Measure of Scale: Large Industries Equation Employment (1) Removed (2) Dependent variable ln CONTRIBUTIONS ln CONTRIBUTIONS ln HERFIN 0.19 * (1.88) 0.22 * (1.92) ln IMPORTPENETRAT 0.06 (0.80) 0.02 ln (NTB/[1 + NTB]) -0.07 ** (2.84) -0.06 * (1.98) ln ELASTIC 0.32 * (1.04) 0.35 ln DOWNSTREAMSHARE 0.31 ** (2.06) 0.17 ln DOWNSTREAMHERFIN 0.16 * (1.25) 0.05 ln ABATEMENTCOSTS1 0.80 ** (4.25) 1.22 ** (5.33) ln ABATEMENTCOSTS2 - -- CONSTANT -1.61 ** (2.83) -0.58 Adjusted [R.sup.2] 0.35 0.45 Model F 7.82 8.85 ln L -108.75 -75.35 Environmental policy Stringency equation Dependent variable ABATEMENTCOSTS 1 ABATEMENTCOSTS 1 NTB/(1 + NTB) 0.02 ** (2.06) 0.03 ** (4.36) CONTRIBUTIONS 0.10 ** (2.81) 0.08 ** (3.57) INTENSITY 0.005 ** (2.20) 0.003 * (1.89) CONSTANT -0.001 (0.29) -0.001 (0.73) Adjusted [R.sup.2] 0.14 0.32 Model F 5.58 11.42 In L 284.44 243.74 N 89 67 Policy Stringency Measure ABATEMENTCOSTS2 Lobby Group Contributions Measure of Scale: Large Industries Equation Employment (3) Removed (4) Dependent variable ln CONTRIBUTIONS ln CONTRIBUTIONS ln HERFIN 0.13 * (1.71) 0.16 ** (2.00) ln IMPORTPENETRAT 0.11 * (1.80) 0.12 * (1.73) ln (NTB/[1 + NTB]) -0.07 ** (2.60) -0.05 * (1.67) ln ELASTIC 0.45 * (1.77) 0.43 * (1.47) ln DOWNSTREAMSHARE 0.25 ** (2.11) 0.16 * (1.19) ln DOWNSTREAMHERFIN 0.08 (0.80) 0.10 (0.85) ln ABATEMENTCOSTS1 -- -- ln ABATEMENTCOSTS2 0.31 * (1.51) 0.55 ** (2.48) CONSTANT -2.95 ** (6.47) -2.01 ** (4.22) Adjusted [R.sup.2] 0.12 0.19 Model F 4.39 5.31 ln L -239.29 -170.40 Environmental policy Stringency equation Dependent variable ABATEMENTCOSTS2 ABATEMENTCOSTS2 NTB/(1 + NTB) 0.02 * (1.81) 0.02 * (1.51) CONTRIBUTIONS 0.20 ** (3.30) 0.22 ** (5.73) INTENSITY 0.01 ** (4.55) 0.01 ** (4.25) CONSTANT 0.0003 (0.10) -0.0003 (0.18) Adjusted [R.sup.2] 0.17 0.33 Model F 13.58 22.81 In L 454.69 395.07 N 177 133 t statistics in parentheses. N, size of data set. * 2 > [absolute value of t] [greater than or equal to] 1. ** [absolute value of t] [greater than or equal to] 2. Firm b [C.sub.i] [D.sub.i] Firm a [C.sub.i] [X.sub.i], [X.sub.i] [Z.sub.i], [Y.sub.i] [D.sub.i] [Y.sub.i], [Z.sub.i] [N.sub.i], [N.sub.i]

We are grateful to two helpful referees, Toke Aidt, Brian Copeland, Bouwe Dijkstra, Christian Hilber, Angeliki Kourelis, Daniel Millimet, Felix Oberholzer-Gee, Barkley Rosser Jr., Kevin Siqueira, and session participants at the Public Choice Society Meetings in San Antonio, at the European Public Choice Society Meetings in Paris, and at the EAERE Annual Conference in Southampton for useful discussions, comments, and suggestions, and to Kishore Gawande for sharing his data and providing additional comments. Earlier versions of the paper were circulated under the title "Trade Policy, Polluters, and Collective Action: Theory and Evidence." The usual disclaimers apply.

Received February 2003; accepted September 2004.

(1) A large body of literature has extended this theory in various directions. However, Grossman and Helpman (1994) argue that the literature suffers from a lack of attention to the issue of lobby group organization, and Persson and Tabellini (2000, p. 175) find that "The major problem is that we lack a precise model of the process whereby some groups get politically organized and others not. This is a difficult question to which there is still no satisfactory answer."

(2) Most previous studies of PACs examine the allocation and timing patterns of PACs (see, e.g., Snyder 1990 and Stratmann 1992, 1995), as well as the effects of PACs (see, e.g., Salamon and Siegfried 1977 and Stratmann 1991). A different strand of the literature argues that the importance of corporate PAC contributions in U.S. politics is relatively small and that PAC contributions are not equivalent to bribes or a proxy for lobbying activity (see Milyo, Primo, and Groseclose 2000; Ansolabehere, Snyder, and Tripathi 2002; Ansolabehere, de Figueiredo, and Snyder 2003). These authors argue that PAC contributions primarily buy access to legislators or are a form of consumption good with little effect on legislation.

(3) Spagnolo (2000) models (theoretically) issue linkage in international agreements. He does not discuss lobby group formation or the determination of trade protection or environmental policies. See also Bernheim and Whinston (1990) and Conconi and Perroni (2002).

(4) This conclusion is based on the assumption that the interests of firms in an industry coincide sufficiently. In the current context this requires that regulations are sufficiently severe to induce firms to lobby for less stringent regulations. Clearly in cases where either firms are highly heterogeneous or regulations have differing impacts on different firms there will be no incentive to lobby for the same policy changes.

(5) Gawande and Bandyopadhyay (2000), as well as Goldberg and Maggi (1999), find support for the predictions made by the model of Grossman and Helpman (1994). See also Gawande (1997, 1998). See Gawande and Krishna (2003) for a recent survey of this literature.

(6) Pecorino (1998) studies free riding in firms' lobbying for protectionism in a repeated game model. He finds that an increase in the number of firms in an industry does not necessarily imply that free-rider problems increase. Cooperation may be sustained even under perfect competition, that is, with an infinite number of firms. Mitra (1999) endogenizes the organization of lobby groups using the Grossman and Helpman (1994) model. He employs an industrial organization-endogenous market structure approach to find the equilibrium number of lobby groups. Mitra argues that geographically concentrated industries with large capital stocks, inelastic demand functions, and few capital owners are more likely to organize.

(7) Our work is a complement to the literature on collective action on the demand side (Hamilton 1993) and to the theory of regulatory "capture" originated by Stigler (1971) and extended by Peltzman (1976) and many others. The latter authors argue that regulations that erect barriers to entry function as outright transfers of wealth and are therefore demanded by industry. In our paper the existence of environmental regulations is due to industries being "naturally" polluting, and larger transfers may occur because of a greater ability to undertake collective action.

(8) An alternative would be to assume all sectors have positive levels of pollution, but some sectors produce only nontradable goods. However, in our data set all sectors are engaged in international trade.

(9) Spagnolo (2000) discusses the consequences of relaxing this assumption. We abstract from further complications in order to focus the example on our main point. Assumption 1 will most likely hold when output is relatively unaffected by the environmental policy instrument, such as when a technology standard is used to regulate pollution. This is the most commonly used environmental policy instrument in the United States Moreover, when a pollution standard and an import quota (assumed given to importers free of charge) are used, no revenues arise. This is consistent with the policies used in our empirical work.

(10) Given Result 1, a polluting industry lobby's problem is the allocation of the available political funds among policy lobbying areas. The marginal net return to lobbying must be equal across policy instruments. Thus, some positive share of these funds will likely be allocated to lobbying on policies other than environmental policies, for example, trade policy. In case the marginal return to lobbying is the greatest in the environmental policy area, the additional funds generated may potentially be fully used in the environmental policy area only, in which case trade policy lobbying would not be affected. This is ultimately an issue that is perhaps best resolved empirically.

(11) It is perhaps useful to note that if this condition is not satisfied in the Grossman and Helpman (1994) model, there is no equilibrium of the lobbying game (see Damania and Fredriksson 2003 for a discussion of this issue).

(12) In some cases the lobbying may be administered through industry groups, while in other circumstances firms may directly lobby policy makers. Irrespective of the organizational arrangements, the theory predicts that with interaction over multiple regulations there will be greater lobbying by firms.

(13) Lobbying is here viewed as directed toward lowering the stringency of environmental policy, rather than to change its type.

(14) Pollution abatement costs have been used as a proxy for the stringency of environmental policies in a number of empirical studies. See, for example, Levinson and Taylor (2001), who also provide an extensive survey of the literature.

(15) Our empirical specification of the NTB equation corresponds to the one estimated by G-B.

(16) First, the level of protection depends on the inverse import penetration ratio [(1MPORTPENETRAT).sup.-1], interacted with a political organization dummy variable (ORGANIZED), which takes the value of one if a sector is organized politically and zero otherwise. We expect that the inverse import penetration ratio has a positive effect on protection if the sector is organized since relatively larger domestic industries will make greater lobbying contributions and receive higher levels of protection provided they are organized. Second, the import demand elasticity has a negative effect on the level of protection due to the greater deadweight loss of taxation associated with a greater elasticity. Finally, our specification also includes two variables that measure the level of protection for intermediate goods used in that industry. The first is a measure of the average tariff level for intermediates (INTERMTARIFF), while the second measures the extent of intermediate goods protection through nontariff barriers (INTERMNTB). A higher level of protection for firms producing intermediate goods increases the cost of production for final goods producers. To stay competitive with foreign producers, final goods producers at home need higher levels of protection.

(17) Our specification of the import penetration equation follows specifications found in previous studies (G-B; Trefler 1993). Trade theory suggests that import penetration is a function of the level of protection, as well as factor endowments (as predicted by standard comparative advantage models), firm size, and industry concentration (as predicted by models with increasing returns), and own- and cross-price elasticities (to control for differing demand conditions). Trefler (1993) estimates an import penetration equation as part of a two-equation system. In addition to the protection measure, Trefler's import equation includes only variables that measure factor endowments. Our specification includes other variables as well and is therefore more general.

(18) The following variables are included in X: SCIENTIST, MANAGER, UNSKILLED, CONC4, FIRMSCALE, TARIFF, ELASTIC, CROSSELAST, REALELASTIC. For a description of these variables, see Table A1 in the Appendix.

(19) The dummy variable ORGANIZED is a transformation of CONTRIBUTIONS (see section 4 for details).

(20) The exact list of cross-product terms is available from the authors upon request. All first-stage estimation results including the results of the overidentification tests are available upon request.

(21) Like other empirical papers on endogenous tariff protection, we are confined to the use of 1983 data because of the lack of disaggregated data on nontariff barriers for later years.

(22) We thus follow, for example, Grier, Munger, and Roberts (1994) by using corporate PAC data in a study of the determinants of collective action (see also Goldberg and Maggi 1999 and G-B). The PACs in our data set are associated with individual firms. As pointed out by a helpful referee, PACs organized by corporations are "connected," as opposed to "unconnected." Whereas unconnected PACs must pay all of their operation costs from personal contributions from any U.S. citizen, connected PACs may have their operating expenses (staff salaries, lawyers' fees, fundraising costs, etc.) paid by their parent organizations. They can only contribute money to candidates that are solicited from and contributed by employees of the sponsoring organization. Connected PACs cannot legally give or spend corporate resources for a candidate for federal office. All contributions by a PAC to candidates for federal office must be drawn entirely from voluntary personal donations.

(23) The empirical findings with regard to the NTB Equation 8 and the import penetration equation 10 are similar to those reported in G-B and are available from the authors upon request. The coefficient for ABATEMENTCOSTS1 in Equation 10 is negative and significant at the 10% level.

(24) Estimation of the LGC equation with the pollution abatement measure in levels leads to qualitatively similar results. The coefficient for ABATEMENTCOSTS1 in the LGC equation is positive and statistically significant at the 1% level. Moreover, using a three-equation system instead, with INTENSITY as an exogenous variable in the LGC equation rather than the endogenous ABATEMENTCOSTS1, results in an insignificant INTENSIFY coefficient. This may be explained by the fact that the simple correlation between INTENSITY and ABATEMENTCOSTS1 (ABATEMENTCOSTS2) is a relatively low 0.24 (0.35). The results are available upon request.

(25) The negative sign of the NTB protection coefficient is contrary to what theory predicts. However, this is not due a potentially strong correlation between our two policy measures. In effect, import protection and ABATEMENTCOSTS1 are only weakly correlated, with a partial correlation coefficient of 0.14. In addition, leaving out the EPS measure does not change the sign on NTB protection, as the results in G-B show. Instead of focusing on the sign of this particular coefficient, it is important to note that the elasticity of corporate PAC spending with respect to deadweight loss is positive, with a value of 0.34 (the elasticity is equal to the sum of the coefficient estimates on IMPORTPENETRAT and ELASTIC plus one half the estimate on NTB/[1 + NTB]). This means that a 10% increase in the deadweight loss from protection would lead to an increase in PAC contributions per firm of 3.4%. Furthermore, the implication of the G-H model that lobbying competition with downstream industries will increase lobbying efforts by upstream firms is also confirmed by the data. Both a strong demand by downstream users and greater concentration of downstream industries are found to have a positive effect on the lobbying contributions per upstream firm. However, only the first effect is statistically significant at the 1% level.

(26) One potential explanation for this result is that ABATEMENTCOSTS1 may be positively correlated with firm concentration in downstream industries, DOWNSTREAMHERFIN. However, a partial correlation coefficient of-0.017 indicates that this is not the case. Instead, the changes in size and significance level of both coefficient estimates appear mostly driven by the change in sample size (N = 89 here vs. N = 242 in G-B). This can be seen from the fact that when Equation 7 is estimated without an EPS variable, both variables remain statistically insignificant. Note that the EPS measure is weakly correlated with industry concentration, HERFIN (r = 0.14), and somewhat stronger with output usage by downstream industries, DOWNSTREAMSHARE (r = 0.22). However, even with EPS as an additional regressor, both variables still matter for lobby group contributions--they are statistically significant at the 7% and 5% level, respectively.

(27) The empirical results for the remaining equations (Eqns. 8 and 10) are available from the authors upon request.

(28) We thank one of our referees for noticing this potential problem, as well as the one addressed in the subsequent paragraph.

(29) To be precise, we remove the largest 25% of all industries, as measured by added value per firm. This reduces the sample using ABATEMENTCOSTS1 from 89 to 63 observations and the sample using ABATEMENTCOSTS2 from 177 to 133 observations. Removal of large-scale industries using employment per firm as measure of size yields quantitatively similar results (available upon request).

(30) For a similar method of proving sustainable cooperative outcomes, see Friedman (1990).

References

Abreu, D. 1986. Extremal equilibria of oligopoly supergames. Journal of Economic Theory 39:191-225.

Andres, G. J. 1985. Business involvement in campaign finance: Factors influencing the decision to form a corporate PAC. PS 18:156-81.

Ansolabehere, S., J. M. de Figueiredo, and J. M. Snyder, Jr. 2003. Why is there so little money in U.S. politics? Journal of Economic Perspectives 17:105-30.

Ansolabehere, S., J. M. Snyder, Jr., and M. Tripathi. 2002. Are PAC contributions and lobbying linked? New evidence from the 1995 lobby disclosure act. Business and Politics 4:131-55.

Bernheim, B. D., and M. D. Whinston. 1990. Multimarket contact and collusive behavior. Rand Journal of Economies 21:1-26.

Conconi, P., and C. Perroni. 2002. Issue linkage and issue tie-in in multilateral negotiations. Journal of International Economics 57:423-47.

Damania, R., and P. G. Fredriksson. 2000. On the formation of industry lobby groups. Journal of Economic Behavior and Organization 41:315-35.

Damania, R., and P. G. Fredriksson. 2003. Trade policy reform, endogenous lobby group formation, and environmental policy. Journal of Economic Behavior and Organization 52:47-69.

Eliste, P., and P. G. Fredriksson. 2002. Environmental regulations, transfers, and trade: Theory and evidence. Journal of Environmental Economics and Management 43:234-50.

Esty, D. C., and R. E. Caves. 1983. Market structure and political influence: New data on political expenditures, activity and success. Economic Inquiry 21:24-38.

Friedman, J. 1971. A non-cooperative equilibrium for supergames. Review of Economic Studies 38:1-12.

Friedman, J. 1990. Game theory with applications to economics. Oxford: Oxford University Press.

Gawande, K. 1997. US non-tariff barriers as privately provided public goods. Journal of Public Economics 64:61-81.

Gawande, K. 1998. Stigler-Olson lobbying behavior and organization: Evidence from a lobbying function. Journal of Economic Behavior and Organization 35:477-99.

Gawande, K., and U. Bandyopadhyay. 2000. Is protection for sale? Evidence on the Grossman-Helpman theory of endogenous protection. Review of Economics and Statistics 82:139-52.

Gawande, K., and P. Krishna. 2003. The political economy of trade policy: Empirical approaches. In Handbook of international trade, edited by J. Harrigan. London: Basil Blackwell, pp. 213-250.

Goldberg, P. K., and G. Maggi. 1999. Protection for sale: An empirical investigation. American Economic Review 89:1135-55.

Grier, K. B., M. C. Munger, and B. E. Roberts. 1991. The industrial organization of corporate political participation. Southern Economic Journal 57:727-38.

Grier, K. B., M. C. Munger, and B. E. Roberts. 1994. The determinants of industrial political activity, 1978-1986. American Political Science Review 88:911-26.

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

Hamilton, J. T. 1993. Politics and social costs: Estimating the impact of collective action on hazardous waste facilities. Rand Journal of Economics 24:101-25.

Heywood, J. S. 1988. The structural determinants of corporate campaign activity. Quarterly Review of Economics and Business 28:39-48.

Humphries, C. 1991. Corporations, PACs and the strategic link between contributions and lobbying activities. Western Political Quarterly 44:353-72.

Kelejian, H. 1971. Two stage least squares and econometric systems linear in parameters but nonlinear in the endogenous variables. Journal of the American Statistical Association 66:373-82.

Levinson, A., and M. S. Taylor. 2001. Trade and the environment: Unmasking the pollution haven hypothesis. Unpublished paper, Georgetown University.

List, J. A., and C. Y. Co. 2000. The effects of environmental regulation on foreign direct investment. Journal of Environmental Economics and Management 40:1-20.

Masters, M. F., and G. D. Keim. 1985. Determinants of PAC participation among large corporations. Journal of Politics 47:1158-73.

McKeown, T. J. 1994. The epidemiology of corporate PAC formation, 1975-1984. Journal of Economic Behavior and Organization 24:43-57.

Milyo, J., D. Primo, and T. Groseclose. 2000. Corporate PAC campaign contributions in perspective. Business and Politics 2:75-88.

Mitra, D. 1999. Endogenous lobby formation and endogenous protection: A long-run model of trade policy determination. American Economic Review 89:1116-34.

Olson, M. 1965. The logic of collective action. Cambridge, MA: Harvard University Press.

Pecorino, P. 1998. Is there a free-rider problem in lobbying? Endogenous tariffs, trigger strategies, and the number of firms. American Economic Review 88:652-60.

Peltzman, S. 1976. Toward a more general theory of regulation. Journal of Law and Economics, 19:211-40.

Persson, T., and G. Tabellini. 2000. Political economics: Explaining economic policy. Cambridge, MA: MIT Press.

Pittman, R. 1976. The effects of industry concentration and regulation on contribution in three 1972 U.S. senate campaigns. Public Choice 27:71-80.

Pittman, R. 1988. Rent-seeking and market structure: Comment. Public Choice 58:173-85.

Salamon, L. M., and J. J. Siegfried. 1977. Economic power and political influence: The impact of industry structure on public policy. American Political Science Review 71: 1026-43.

Shapiro, C. 1989. Chapter 6: Theories of oligopoly behavior. In Handbook of industrial economics, vol. 1, edited by R. Schmalensee and R Willig. Amsterdam: Elsevier Science, pp. 329-414

Sheills, C. R., R. F. Stem, and A. V. Deardorff. 1986. Estimates of the elasticities of substitution between imports and home goods for the United States. Weltwirtschaftliches Archiv 122:497-519.

Snyder, J. 1990. Campaign contributions as investment. Journal of Political Economy 98:1195-227.

Spagnolo, G. 2000. Issue linkage, delegation, and international policy cooperation. Unpublished paper, Stockholm School of Economics.

Stigler, G. 1971. The theory of economic regulation. Bell Journal of Economics and Management Sciences 2:3-21.

Stratmann, T. 1991. What do campaign contributions buy? Southern Economic Journal 57:606-20.

Stratmann, T. 1992. Are contributors rational? Untangling strategies of political action committees. Journal of Political Economy 100:647-64.

Stratmann, T. 1995. Campaign contributions and congressional voting: Does the timing of contributions matter? Review of Economics and Statistics 77:127-36.

Trefler, D. 1993. Trade liberalization and the theory of endogenous protection: An econometric study of U.S. import policy. Journal of Political Economy 101:138-60.

U.S. Department of Commerce, Bureau of the Census. 1983. Pollution abatement costs and expenditures. Current Industrial Reports. Washington, DC: GPO.

Vousden, N. 1990. The economics of trade protection. Cambridge, UK: Cambridge University Press.

Zardkoohi, A. 1985. On the political participation of the firm in the political process. Southern Economic Journal 51:804-17.

Richard Damania, * Per G. Fredriksson, ([dagger]) and Thomas Osang ([double dagger])

* Department of Economics, Adelaide University, Adelaide, Australia 5005; E-mail Richard.damania@ adelaide.edu.au.

([dagger]) Department of Economics, Southern Methodist University, P.O. Box 750496, Dallas, TX 75275-0496, USA; E-mail pfredrik@mail.smu.edu.

([double dagger]) Department of Economics, Southern Methodist University, P.O. Box 750496, Dallas, TX 75275-0496, USA; E-mail tosang@mail.smu.edu; corresponding author.

Printer friendly Cite/link Email Feedback | |

Author: | Osang, Thomas |
---|---|

Publication: | Southern Economic Journal |

Geographic Code: | 1USA |

Date: | Jul 1, 2005 |

Words: | 11500 |

Previous Article: | Price and output stability under price-level targeting. |

Next Article: | A note on deficit, implicit debt, and interest rates. |

Topics: |