Per G. Fredriksson. Department of Economics, University of Louisville, Louisville, KY

Transcription

1 POLITIAL INSTITUTIONS, INTRST GROUPS, AND TH RATIIATION O INTRNATIONAL NIRONMNTAL AGRMNTS Per G. redriksson Department of conomics, University of Louisville, Louisville, KY 409 mail: per.fredriksson@louisville.edu Gergely Uelyi Department of conomics, Harvard University, ambridge, MA 038 mail: uelyi@fas.arvard.edu August 4, 006 Abstract We investigate ow domestic political institutions and interest group pressures ointly determine te probability tat a country ratifies an international environmental agreement. We build a multi-agent, multi-principal model of government decision-making were government veto players suc as te legislative cambers or te president are offered political contributions from environmental and industry lobby groups. Te model suggests an asymmetry in te impact of political institutions on lobby groups. Institutional arrangements wit a greater number of veto players reduce te positive impact of environmental lobbying on te ratification probability. Suc institutional features ave ambiguous effects on industry lobbying, owever. We test tese predictions using Logit and stratified proportional azard models, and panel data from 70 countries on te timing of Kyoto Protocol ratification. onsistent wit our teory, increased environmental lobby group pressure raises te probability of ratification, and te effect declines as te number of veto players increases. irm lobbying is unaffected by te number of veto players. We are grateful for useful comments and suggestions from Alberto Alesina, Scott Barrett, Mats Bergman, Kate mans, Peter Hartley, lanan Helpman, Dale Jorgenson, Patrick Keoe, Angeliki Kourelis, Leandro de Magalaes, David Martimort, ric Neumayer, rançois Salanié, Robert Stavins, Andrea Szabó, Herman olleberg, Martin Weitzman, and seminar participants at rasmus University Rotterdam, University of Gotenburg, Harvard, Minnesota, NBR, University of Toulouse, te 004 AR conference in Budapest, and a ZW worksop in Manneim. Uelyi acknowledges te U grant Arcimedes Prize HPAW-T wic supported part of tis researc. Te usual disclaimers apply.

2 I. INTRODUTION To gain legal force, te Kyoto Protocol on greenouse gases needed ratification by at least 55 countries representing 55% of OD and former astern Bloc countries 990 emissions. Tis tresold was reaced in November 004 wit te ratification of te Russian ederation, almost 7 years after te Protocol was opened for ratification on Marc 6, 998. ountries ave exibited widely different ratification beavior. Wile some island states like te Maldives or ii ratified after only a few monts, rance and Japan waited several years. A few countries, including Australia, roatia, Kazakstan, and te U.S., ave not yet ratified as of September, 006. Te slow entry into force of international environmental agreements IAs appears to be an important obstacle to addressing global environmental problems in a timely manner. Tus, an improved understanding of ratification beavior is important for improving te IA policy process. Tis paper seeks to uncover ow domestic political institutions sape te ability of political interest groups to influence ratification of international environmental agreements. We provide a teory of ratification, and test te predictions on Kyoto Protocol data. We study te game played by te various interested parties witin a country facing a ratification decision, and te interaction between domestic institutions and interest group lobbying. Any international strategic influences are taken as given. Wile various features of te domestic institutional environment may ave an impact on te ratification of international environmental agreements, our focus is on te number of domestic individual or collective veto players Ps participating in a ratification decision suc as te president, te prime minister, te cambers of parliament, te maority party, or te government coalition parties; see Tsebelis, 999, 00.,3 Ratification decisions, similarly to most maor policymaking, involve several Ps tat are all subected to lobby group pressures. Previous literature on te ratification of international environmental agreements IAs as treated te domestic political process in eac country as a black box, and as focused on te games of negotiation countries play against eac oter see, e.g., arraro and Siniscalco, 993; Barrett, 994. Wile suc strategic aspects migt be teoretically important for large countries, it is unclear to wat extent tey influence decision making in te many smaller nations. Moreover, te few empirical studies tat exist on tis aspect of IA ratification ave found no empirical support for strategic interactions see Beron et al., 003. Te literature as identified positive effects of broad measures of democracy on te degree of international cooperation on global environmental problems see, for example, ongleton, 99; Neumayer, 00a; and Beron et al., 003. Moreover, see, e.g., Persson et al. 997, 000, Bennedsen and eldmann 00, and redriksson and Millimet 004 for discussions of te role of domestic political institutions suc as presidential and parliamentary regimes for fiscal policy, lobbying, and environmental policy, respectively.

3 We first present a simple model building on te vote-buyer approac of Groseclose and Snyder 996 and Diermeier and Myerson 999. In te model, firm and environmental lobby groups seek to affect an IA ratification decision made by n identical and independent Ps by paying campaign contributions. Ratification occurs only if all Ps vote for it. Te teory identifies an asymmetry between environmental and firm lobby groups impact on te IA process. Wile te pro-reform environmental lobby must convince eac P to support ratification, te anti-reform firm lobby needs only to convince one single P to block ratification. Te required contribution to eac P is constant, and tus wile environmental lobbying raises te ratification probability, te effect is conditional on te number of Ps. In particular, an increase in te number of Ps reduces te impact of environmental lobbying on te ratification probability. On te oter and, te impact of firm lobbying is independent of te number of Ps. We proceed to generalize tese implications to a setting in wic te lobbies move simultaneously, and Ps' effect on te ratification outcome is probabilistic. Tis extension builds on te common agency tradition Berneim and Winston, 986; Grossman and Helpman, 994, but as multiple agents as well as multiple principals, wit externalities among te agents. 4 In tis model, contributions are made by budget-constrained industry and environmental lobby groups, 5 and eac of te n Ps determines te level of er political support for ratification, taking bot te contribution offers from te lobby groups and social welfare into account. Te probability tat te country ratifies te IA is a function of te Ps political support, wit te property tat any P can unilaterally decrease tis probability, but er ability to increase it depends on te oter Ps support. Tis feature generalizes te notion of veto power and allows for probabilistic ratification outcomes. 6 3 Te literature argues tat te number of collective or individual actors wose agreement is necessary for a policy cange veto players, and te ideological distance between tem, affects te stability of government policies, as well as teir ability to address problems suc as budget deficits and inflation Roubini and Sacs, 989; Tsebelis, 999, 00; and Keefer and Stasavage, See Segal 999 and Martimort and Stole 003 for different approaces to externalities in agency problems, and Prat and Rusticini 003 for a general multi-principal multi-agent model witout externalities. Te lobbying literature as largely ignored te implications of multiple veto players, and te externalities created by policy decisions.see Grossman and Helpman 996, 00 for multiple-lobby group, multiple-party models. 5 or example, budget constraints migt result from te fact tat lobby groups ability to generate funds is restricted by free-riding incentives see, e.g., Olson, Tis approac is consistent wit our empirical work, wic investigates te determinants of te probability of ratification. It is related to te contest success functions approac previously used to model environmental conflicts see, for example, Heyes, 997; Hurley and Sogren, 998.

4 In tis extended model, increasing te number of Ps increases te lobbying expenditures of te environmental lobby, and terefore reduces its positive impact on te ratification probability as in te simple model. At te same time, te number of Ps migt increase or decrease te lobbying expenditures of te firm lobby. Tus, te effect of Ps on te influence of te firm lobby is ambiguous, and terefore an empirical issue. Te teory's predictions are tested empirically using panel data on Kyoto Protocol ratification. We measure te number of Ps wit i a dummy for bicameral as opposed to unicameral parliamentary systems, as well as ii a measure on political cecks and balances from te World Bank s database of political institutions Beck et al., 00; Keefer and Stasavage, irst, a bencmark Logit model is specified. Second, we present a survival analysis using stratified proportional azard models. Wile most earlier studies of te probability of environmental treaty ratification focus on te event, rater tan te timing of ratification for example, ongleton, 99; Beron et al., 003, real world ratification processes take place over time. A duration model, wic captures wen a treaty is ratified, may capture a greater amount of information regarding ratification probabilities. Wile redriksson and Gaston 000 and Neumayer 00a also use tis approac, tey employ static datasets rater tan a panel. A stratified proportional azard model allows for different baseline azard functions. Tis appears particularly appropriate since Annex OD and former astern Bloc and non-annex countries ave widely different obligations under te Kyoto Protocol. Bot te Logit and azard models reveal tat environmental lobbying and political institutions are important for te probability of IA ratification. ountries wit a greater number of environmental lobby groups ave a iger probability of ratifying te Kyoto Protocol, tus ratification occurs earlier. Moreover, te effect is conditional on te number of Ps. Te impact of environmental lobbying diminises as te number of Ps increases. However, we find no evidence tat firm lobbying is affected by te number of Ps. We believe tis is te first empirical evidence on te asymmetric impact of domestic political institutions on environmental and industry lobbying, and on domestic political institutions implications for te provision of global public goods, ere reflected in te degree of cooperation on a global environmental 7 Te effect of bicameralism on policy outcomes as rarely been te focus of researc and, consequently, its effects on policy processes and outcomes are not well understood Diermeier et al. 00, p.. See also Tsebelis and Money 997 and ongleton

5 treaty. 8 We also believe our findings may ave more general applicability to oter policy areas suc as te ratification of WTO trade agreements and ILO conventions, were rival lobby groups seek to influence ratification in opposing directions. Te paper is organized as follows. Section II provides te teory, Section III reports te empirical results, and Section I concludes. A simple model II. THORY In tis section, we first set up a simple model tat sows te intuition and delivers te main empirical predictions in a particularly parsimonious manner. We ten sow tat te predictions generalize, wit minor differences, to a more general setting. onsider a political decision taken ointly by n veto players Ps on weter or not to ratify an international environmental agreement IA. Ps are te collective political units cambers of te legislature, government coalition members or individuals president, prime minister wose approval is necessary for te IA to be ratified by te country Tsebelis, 999, 00. In te political process, Ps are influenced in teir decision by an environmental and a firm lobby. Tese groups are assumed to ave overcome teir free-riding problems for collective action Olson, 965, and membersip in te firm- and environmental lobby group equals f and e, respectively. Since te status-quo policy is no ratification, te environmental lobby, wanting to cange te status quo, is assumed to move first in te political game, following Groseclose and Snyder 996 and Diermeier and Myerson 999 see also Banks, 000. It offers political contributions to one or more Ps in excange for a promise to vote for ratification. 9 Te firm lobby moves next to defend te status quo, by offering contributions to te Ps in excange for teir vote against ratification. In te tird stage, voting takes place, and ratification occurs if and only if it gets unanimous support from te Ps. 0 inally, lobbies make te political 8 redriksson et al. 005 study te effect of government corruptibility on Kyoto Protocol ratification beavior but ignore te effect of multiple Ps. Tey find tat an increase in corruptibility raises te environmental lobby groups, but not te firm lobbies', impact on te ratification outcome. 9 See Riddel 003 for recent empirical evidence of environmental lobby groups influencing political decisions made in te U.S. Senate wit te elp of political contributions PAs. 0 Te precise manner in wic te political decision is made is not important. or example it could be some combination of voting and political bargaining. Wat is important is tat te Ps ave veto power, and te lobbies do not make furter offers during te process. 4

6 contributions based on te individual Ps voting beavior, all te players collect payoffs depending on weter ratification occurred, and te game ends. If IA ratification occurs, eac environmentalist gains group receives a payoff equal to ev c, were v, and tus te environmental lobby c denotes total contributions. Te payoff reflects te value of improved environmental quality resulting from stricter environmental policies following ratification. If ratification does not occur, te environmental lobby receives a payoff equal to zero. If ratification does not occur, individual members of te firm group eac gain v v, for a total payoff equal to fv c, were c is total firm contributions. Tis payoff reflects te greater factor rewards associated wit weaker environmental policies in te status quo. If ratification occurs, te firm lobby receives a payoff of zero. Te value of ratification to eac of te n Ps is given by W. We assume tat fv > W. Proposition. In a Subgame Perfect Nas quilibrium, ratification occurs if and only if e W v fv 0. n Proof. See te Appendix. Te condition in Proposition implies tat a country will be more likely to ratify if, ceteris paribus, te number of environmentalists, e, is larger, or te number of firm owners, f, is smaller. Proposition suggests tat te former effect is conditional on n, te number of Ps. Because wit more Ps, a larger contribution is needed from te environmental lobby to acieve ratification, a rising n will reduce te positive effect of e on te likeliood of ratification. Note tat since te firm lobby only needs to contribute to a single P to prevent ratification, n as no effect on te negative impact of f. Tus, te impact of n on environmental pro-reform and industry anti-reform lobbying is asymmetric. In Section II, we test tese predictions of te model. or simplicity, we assume tat all Ps ave identical preferences. Tis is not crucial for te teoretical prediction. Tis result is complementary to te status-quo bias identified by ernandez and Rodrik 99 against trade policy reform. Teir argument builds on te uncertainty associated wit reform. 5

7 A more general model Wile te above model is particularly transparent, one migt find some of its assumptions unsatisfactory. irst, te timing of moves migt seem artificial Wat guarantees tat te environmentalist lobby moves first? If it does, wat prevents it from responding to te industry's move?. In many respects, a simultaneous game seems more appropriate. Second, it migt seem tat te veto power of te Ps is taken too literally: in reality, a politician may ave an impact on te probability of ratification troug public statements, legislative negotiations, and, ultimately, voting. Below, we provide a model incorporating tese features, and sow tat it delivers predictions consistent wit tose obtained from te simple model above. Suppose it is known tat, at a future date, n Ps, indexed by H will decide weter or not to ratify an international environmental agreement tat as recently been negotiated. As suggested in te Introduction, te ratification process unfolds over time, implying tat some economic decisions ave to be taken based on te proected outcome of te Ps' decision. We terefore assume tat today eac P announces its level of political support for ratification, represented by a number [0,]. We interpret = as a credible commitment to vote for ratification, = 0 as a commitment to veto, and 0, as te likeliood tat P will approve ratification. Togeter, tese political supports determine te obective probability tat te country will ratify te IA in te future. We assume n = Π. Tis simple specification sares and generalizes two defining caracteristics of any model wit veto powers: i a player can always reduce te likeliood of ratification, irrespective of te oter players political support by coosing < ; ii a player s ability to increase te likeliood of ratification depends on te oter players support no P can increase above any oter P s. 3 = 3 Tis extension of unanimity voting as te advantage of allowing for probabilistic outcomes, providing a direct link between our model and te empirical tests, wic focus on te azard of ratification. Our approac is also related to te rent-seeking literature wic models contests in wic participants exert effort to increase teir probability of winning a prize. Tere, eac player's probability of winning is a function of all players' efforts see, e.g., Heyes, 997; Hurley and Sogren, 998. Alternative specifications saring tese features would be to i ave a country ratify wit certainty if larger tan some tresold, and not ratify oterwise; or ii specify te outcome to be min rater tan te multiplicative form we use. iter alternative would give similar results to H our specification, but would be discontinuous in, making te analysis less tractable. 6

8 Te probability tat te IA is ratified affects te expected environmental policy in te country. 4 To be concrete, assume tat ratification means te introduction of an emission tax τ, were te value of τ is exogenously fixed by te IA. Te expected tax rate is ten t τ = τ. Tree separate groups - consumers, factor owners firms, and environmentalists ave to make economic decisions before te uncertainty regarding te ratification outcome is resolved. Tese decisions, and te resulting payoffs, are based on te expected tax rate t. onsumers payoffs including consumer surplus and redistributed tax revenues are given by U t, wit U '> 0 for low t driven by increasing tax revenues, U '< 0 for ig t resulting from reduced production in response to iger expected taxes, and U '' < 0. actor owners payoff profit is v t, wit v ' < 0 and v '' < 0. inally, environmentalists payoff disutility from pollution is v t, wit v ' > 0 and v '' < 0. 5 Assuming tat te size of te tree groups is, f and e, respectively, social welfare is given by W t U t t t, were t fv t and t ev t. learly, W '' < 0, and we assume te existence of a unique ratification probability o 0, suc tat o t o = τ maximizes social welfare. Since τ is constant, and te quantity of interest is te probability of ratification, we encefort write all payoffs as functions of. As before, firms and environmentalists are assumed able to organize into separate lobby groups. Tese groups coordinate separate prospective political contribution offers to te Ps, wit an aim to influence te likeliood of ratification. Te contribution offered by lobby group to P is a continuous function dependent on te P s political support for ratification. Tus, lobby s payoff is given by n, =,. = Bot lobbies are assumed limited in teir fund-raising abilities by borrowing constraints and free-riding incentives see Olson, 965. Tus, teir budgets ave ard constraints: 4 Te following specification of te payoffs is te reduced form of a general equilibrium model of a small open economy were te expected environmental tax determines production decisions and ence pollution, consumers derive utility from consumption, firms care about increasing profits, and environmentalists care about reducing pollution cf. Aidt, Strict concavity of te individual payoff functions is not crucial, as long as weigted averages of te payoffs as described below remain concave. 7

9 n = M, so tat te sum of contributions cannot exceed M. 6 onsumers are assumed to face sufficiently severe free-riding problems to be unable to organize political action. P as obective function GU, wic is a weigted sum of te expected aggregate social welfare and te contributions associated wit P s support for ratification. Te exogenous parameter a is te P s affinity for delivering welfare relative to political contributions. Te model defines a two-stage game between te Ps and te lobby groups. Te timing assumptions are as follows: Stage. Te two lobby groups, =,, simultaneously and non-cooperatively offer eac identical P a political contribution scedule. Stage. ac P determines and announces er optimal level of support,. IA ratification occurs wit probability n = Π. At te end of te game, te payoffs are realized, in = particular, Ps receive te contributions associated wit te cosen level of support. 7 Te Political quilibrium A political equilibrium is a subgame perfect Nas equilibrium of our two-stage game in wic bot lobby groups and all Ps use pure strategies. or tis game, a set of political contribution scedules, {,, and a vector of levels of support, { } H, is te outcome of a political } H equilibrium if te following tree conditions old were : 8 q q for every H, maximizes on 0, 6 Our model focuses on only one policy and tus te lobby groups do not face a coice in allocating teir budget among policy issues. We believe tis may be a weak assumption since many lobbies bot firms and environmental groups believe global warming policy to be a priority, and even a matter of survival. 7 We assume bilateral contracting wit public offers Segal, 999, i.e. tat te lobbies can commit to publicly observable compensation scedules, and in te second stage all Ps coose teir action simultaneously, taking as given te offered scedules. 8 Te caracterization of te equilibrium is standard. See, for example, Prat and Rusticini 003, Teorem. 8

10 9 H { } maximizes bot i = n on 0, n, subect to n M =, and ii = n on 0, n, subect to n M = 3 for every H, max =, and max = ondition requires tat eac P selects te political support tat maximizes er utility, given te offered contribution scedules and te coices of te oter Ps. ondition states te utility maximization problem of te lobbies: for a given equilibrium contribution scedule, tey coose te vector H { } to be implemented so as to maximize teir payoff minus te sum of contributions. ondition 3 establises tat eac lobby group minimizes te lobbying expenditures necessary to obtain te equilibrium levels of support. To acieve tis, for eac P, lobby lowers its contribution offer until is indifferent between te equilibrium level of support,, and some alternative support associated wit a zero contribution from lobby. quilibrium Levels of Support Assuming tat te lobby groups use differentiable contribution scedules and te Ps pick teir levels of support in te interior of te 0, interval, te O of eac s maximization of condition equals 0 = W a. Te equilibrium levels of support also satisfy condition, implying, for all, 0 = µ, 3 and 0 = µ, 4 were µ and µ are te non-negative sadow prices of te environmental and firm lobby s budget, respectively. Te sadow price 0 = µ if lobby s budget constraint does not bind. Substituting 3 and 4 into and rearranging yields te equilibrium caracterization

11 W a β β = 0, 5 were β / µ, =,. Te equilibrium ratification probability trades off social welfare effects adusted by a and te welfare implications for te lobbies adusted by β. 9 As expression 5 demonstrates, budget constraints provide a natural way to represent te strengt of lobbies influence in te political process, as determined by teir available resources. If a lobby s budget M goes to zero e.g., due to fund raising difficulties, so tat te sadow price of its budget constraint becomes ig, its weigt β in te political process diminises. Lobbies wit tigter resources are less able to lobby te Ps to obtain favorable policies. At te oter extreme, as a lobby s budget becomes sufficiently large so tat its constraint does not bind, β reaces its maximum value of. 0 Intuitively, as te environmental lobby s impact on te equilibrium probability increases te second term, we expect to increase and conversely for te tird term, wic reflects te firm lobby s influence. To guarantee tat tis is indeed te case, we make a tecnical assumption: a Assumption. or every, S < 0. a Assumption requires tat in equilibrium te weigt of consumer surplus be sufficiently ig, relative to te weigt of te firm lobby. It is satisfied, for example, if te firm lobby s payoffs are linear, or if te Ps put enoug weigt a on social welfare relative to political contributions. Tis assumption rules out a substitution effect tat could cause te Ps to reduce in response to an increase in e. To see tis, note tat from 5, we ave a β v =, 6 e D β, β were D. Assumption ensures tat D β, β β, β as a β a β is negative as β. Hence, 6 is positive; an increase in te number of environmentalists i.e. in te environmentalists political pressure unambiguously raises te ratification 9 or example, an increase in te probability of delivering an environmental policy troug ratification tat yields greater welfare and a iger political contribution from te environmental lobby is weigted against a lower political contribution from te firm lobby. 0 Note tat te number of Ps enters 5 only troug te β variables; terefore, if neiter budget constraint binds so tat β = β =, te number of Ps does not affect te overall ratification probability. 0

12 probability. Similarly, Assumption implies tat an increase in f unambiguously lowers te ratification probability: a β v = < 0. f D β, β inally, Assumption also ensures uniqueness of te equilibrium in 5, for given β. 7 quilibrium ontributions Te caracterization of te equilibrium contributions is complicated by te fact tat tese depend on te sape of te oter lobby s contribution offers away from te equilibrium see condition 3. To put some structure on tese off-equilibrium contributions, and tus to restrict te range of possible equilibria, te literature commonly assumes some form of trutfulness Berneim and Winston, 986, i.e. tat te contribution scedules matc te lobbies gross payoff functions. We adopt a similar approac and sow tat in a large class of equilibria wic contains trutful equilibria, it cannot be te case tat bot lobbies budget constraint binds. Using tis result, in te next section we turn to cases were only one budget constraint binds, and derive te effect of te number of Ps, n, on te equilibrium ratification probability. Lemma. Suppose tat, olding oter Ps political support constant at its equilibrium level, lobby s contribution to P is weakly increasing in lobby s gross payoff, i.e., tat for some weakly increasing functions T, T, =,. Ten, te following olds: i If te budget constraint of te firm lobby group binds, ten in equilibrium te environmental lobby group offers positive contributions iff o >. ii If te budget constraint of te environmental lobby group binds, ten in equilibrium te firm o lobby group offers positive contributions iff <. Proof. See te Appendix. Lemma establises te following intuitive result. Suppose tat in equilibrium lobby i cannot treaten to increase its contribution in order to induce a more favorable outcome because or analyses of trutful contributions, see Berneim and Winston 986, Grossman and Helpman 994, and Dixit et al See Prat and Rusticini 003 for te concept of weak trutfulness.

13 its budget constraint binds. Ten te rival lobby i will only be willing to pay for tis equilibrium if it involves te Ps pusing te ratification probability in te direction favored by lobby. In oter words, te environmental firm lobby will only pay for a probability iger o lower tan te Ps preferred probability,. Tus, we ave: Lemma. Under te conditions of Lemma, te equilibrium cannot ave bot lobbies budget constraints binding simultaneously. Proof. rom Lemma, if bot budget constraints were binding, at least one lobby would not offer positive contributions. But ten te budget constraint of tis lobby cannot bind. Q..D. Lemma establises tat tere cannot be an equilibrium in wic bot lobbies budget constraints bind. In wat follows, we terefore restrict our attention to cases wen only one of te budget constraints binds. In te next section, we investigate te effect of a canging number of Ps on te equilibrium probability of ratification. Te effect of eto players In tis section, we investigate te effect of canging te number of Ps, n, on te equilibrium ratification probability. As Lemma sows, te two relevant cases are wen one lobby is budget constrained, wile te oter one is not. We assume tat te lobby wose budget constraint does not bind uses trutful contribution scedules, i.e., tat for all H, g g = b, 8 g were b is a constant cosen by lobby see Berneim and Winston, 986; Grossman and Helpman, or future reference, let us use to denote te equilibrium probability tat rom Lemma, we ave te natural result tat in tese cases competition between a constrained unconstrained environmental lobby and an unconstrained constrained firm lobby implies tat in equilibrium te Ps induce a o ratification probability tat is lower iger tan te welfare maximizing level. 3 Trutful contributions are commonly used to ensure uniqueness of te equilibrium. or our multi-principal multiagent model wit externalities, te definition in 8 effectively introduces outcome-contingent compensation scedules as it implies tat will be a function of, rater tan. Our qualitative results below old for any contribution scedule for wic it is true tat te sum of a lobby s contributions weakly increases in tat lobby s gross payoff see Lemma.

14 te unconstrained lobby would acieve if it were te only lobby exerting political influence: arg max o, =,. learly, < <, and te equilibrium probability wit bot lobbies active may lie anywere in te interval,. Our general results below establis an asymmetry in ow te number of Ps affects te influence of environmental and firm lobby groups. To provide some intuition for te nature of tis asymmetry, consider te following simple example. o xample onsider two countries, One and Two. Wile = 0. 3, = 0. 5, and = 0. 7 in bot countries, in country One, n =, and in country Two, n =. Suppose first tat = 0. 4 in bot countries. In tis case, from Lemma, te environmental lobby spends te full budget in equilibrium. How muc does it ave to spend? In country One, tis lobby as to pay te amount necessary to make te single P coose 0.4 rater tan te 0.3 favored by te firm lobby. all tis amount x. In country Two, since, > necessarily, bot Ps can treaten to > reduce te overall probability to 0.3, and ence tey bot ave to receive x. As sown below, a P's ability to favor te firm lobby by reducing te ratification probability never depends on te number of Ps. Tus, te environmental lobby s expenditure necessary to compensate te Ps always increases linearly in te number of Ps. Suppose now tat = in bot countries. In tis case, from Lemma, te firm lobby spends its budget completely. In particular, in country One te firm lobby as to pay te amount necessary to make te single P coose 0.64 rater tan te 0.7 favored by te environmental lobby. all tis amount y, and consider country Two. If te equilibrium in Two is suc tat 0.64 and, ten P as te same bargaining power as te single P in country = = One, and terefore as to receive y. P, on te oter and, cannot treaten to increase te probability any furter, and terefore receives zero from te firm lobby. In tis case, Ps' ability to favor te environmental lobby by increasing te ratification probability is te same regardless of te number of Ps. Terefore, te firm lobby s expenditures are identical in countries One and Two. If, owever, = 0. 8 in country Two, ten eiter P can unilaterally increase = from 0.64 to 0.7 by coosing = 0.7 / 0.8 = , terefore bot ave to be paid y. In tis case, te firm lobby s expenditure will increase linearly in te number 3

15 of Ps. As sown below, te firm lobby s expenditures may increase, stay constant, or even decrease wit te number of Ps. As sown in 5, te budget-constrained lobby group's weigt in determining te equilibrium probability is affected by its lobbying expenditures. 4 Tus, unless lobbies resources vary at least proportionately wit te number of Ps across countries controlling for observables, tese differences in lobbying costs translate into testable differences in ratification probabilities. Main Results We first consider te case wen te environmentalists budget constraint binds. Proposition. Assume tat te environmental lobby s budget constraint binds, wile te firm lobby s constraint does not bind. Assume furter tat te firm lobby uses trutful contribution scedules. Ten, te contribution expenditures from te environmental lobby necessary for implementing any given ratification probability players. increase linearly wit te number of veto Proof. See te Appendix. As suggested above, te result in Proposition derives from te fact tat is always greater tan. Hence, in any equilibrium, it is always true tat any given P can reduce te overall probability to, given te political support cosen by te oter Ps. Since te equilibrium contributions must compensate eac P for not carrying out tis treat, te environmental lobby must give te same contributions to every P in order to implement a given ratification probability >, no matter ow large n is. Tus, its expenditures are linearly increasing in n. Proposition implies tat if te environmental lobby s budget constraint binds in equilibrium, its influence declines wit te number of Ps, because tis lobby s budget constraint becomes increasingly tigter. Tus, a testable interaction exists between environmental lobby group strengt and te number of Ps. We formulate tis insigt in te following orollary, wic confirms te environmental lobbying result of te simple model. 5 4 Note tat te unconstrained lobby s weigt in determining te equilibrium is te same in bot countries one. 5 It is wort empasizing tat te bargaining power of any given P does not cange as n increases. Tus, we are not simply assuming tat canging te status-quo is arder wen more Ps are present. Instead, te result follows from te conditions te equilibrium contribution must satisfy, and te fact tat te lobbies are budget constrained. 4

16 orollary. Assume tat te environmental lobby s budget constraint binds, wile te firm lobby s constraint does not bind. Assume furter tat te firm lobby uses trutful contribution scedules. Ten, te influence of te environmental lobby on te equilibrium ratification probability,, is decreasing wit te number of veto players. Proof. Differentiation of 6 and rearranging yield v = e n D β, β n [ as a ] 9 were D is defined under 6. rom Proposition, β / n < 0. Terefore, from Assumption, expression 9 is negative. Q..D. Te next proposition sows tat if te firm lobby is budget-constrained and te environmental lobby uses trutful contributions, te effect of te number of Ps on te firm lobby s campaign contribution expenditures depends on te form of te equilibrium. 6 Proposition 3. Assume tat te firm lobby s budget constraint binds wile te environmental lobby s constraint does not. Assume furter tat te environmental lobby uses trutful contribution scedules. Te effect of an increase in te number of veto players on te firm lobby s contribution expenditures is indeterminate. In particular, te firm lobby s lobbying expenditures i are unaffected by te number of veto players if = for some suc tat = ; ii increase wit te number of veto players if > for all, and ' for some, and increase linearly wen for all ; iii may increase or decrease wit te number of veto players if > for all. Proof. See te Appendix. How could decreasing te ratification probability be more expensive for te firm lobby as a result of more Ps, wen a single P can make tis probability arbitrarily low? As before, in te 6 In stating Proposition 3, we are implicitly assuming tat te specific conditions on te structure of te equilibrium given in te different cases remain valid after te increase in n te effect on lobbying expenditures is indeterminate if we allow for switcing between te cases. 5

17 political equilibrium, a P as to be compensated for not favoring te oter lobby. Moreover, wit externalities among te Ps, any P s ability to favor te environmental lobby by inducing a iger ratification probability depends on wat te oter Ps are doing. Tat is, on te structure of te political equilibrium. or example, wen, [,, olding fixed te oter P s coice, eiter P may unilaterally acieve te environmental lobby s preferred outcome,. In tis case, te firm lobby must terefore also pay bot Ps equally, and its expenditures will be linearly increasing in n. Te implication of Proposition 3 for a budgetconstrained firm lobby is stated in te following orollary. orollary. Assume tat te firm lobby s budget constraint binds wile te environmental lobby s constraint does not. Assume furter tat te environmental lobby uses trutful contribution scedules. Ten, te firm lobby s influence on te equilibrium ratification probability,, may be conditional on te number of veto players. Proof. Differentiation of 7 yields v = f n D β, β n [ as a ]. 0 rom Proposition 3, te sign of β / n is indeterminate, and tus te sign of 0 is indeterminate. An increase in n reduces raises [as no effect on] te influence of te firm lobby if β / n ><[=]0. Q..D. orollary sows tat, in tis more general model, te effect of te number of Ps on te firm lobby s influence on ratification is be ambiguous, and is terefore an empirical question. III. MPIRIAL APPROAH In tis section we test two implications derived from te above teoretical models. i Increased environmental lobbying strengt raises te probability of IA ratification, and te effect of environmental lobbying is conditional on te number of Ps. In particular, te effect is smaller in countries wit a greater number of Ps. ii irm lobbying reduces te probability of IA 6

18 ratification. Weter a greater number of Ps increases or decreases te impact of firm lobbying depends on te structure of te political equilibrium. 7 Data and Main ariables We ave panel data from 70 countries recording weter, and wen, a country ratified te Kyoto Protocol. Time is measured in days. ountries enter te sample on Marc 6, 998, wen te Protocol was opened for ratification. Tey eiter exit te sample by ratifying, or remain at risk until December 5, 00 day 735, te last observation date. 8 Trougout our empirical work, te dependent variable, t, is te probability tat a country ratifies at time t. 9 In tis sample, 88 countries 5.7 % ratified before December 5, 00, wile te remaining countries exited witout ratifying. Among ratifiers, te mean duration was 30 days, wit a standard deviation of 459 days. Te first country to ratify was ii, after 85 days; te last was Sout Korea, after 698 days. Table breaks te sample down into ratifying/non-ratifying and Annex/non-Annex countries. Our explanatory variables and sources are described in Table. A variable of main interest is te number of Ps, n, for wic we employ two different measures. TOPLAYRS_ is a dummy taking a value of one if te parliament in a given country is bicameral, and zero if it is unicameral. Te data comes from Wallack et al. 003, and was supplemented wit data from te Parline database of te Inter Parliamentary Union 004. Te second measure, TOPLAYRS_, is te ecks variable from te World Bank s Database of Political Institutions Beck et al., 00, wic counts te number of decision makers wose agreement is necessary before policies can be canged adusting for weter tese veto players are independent of eac oter, as determined by te level of electoral competitiveness in a system, teir respective party affiliations, and te electoral rules Beck et al., 00, p.. 30 TOPLAYRS_ and TOPLAYRS_ appear to closely follow our teory s focus on te ig-level policy makers involved wit IA ratification decisions. 7 Since Ps may act differently in different countries, resulting in different equilibrium structures, one can expect te interaction effect between industry lobbying and te number of Ps to be ard to identify. Te empirical work offers te opportunity to sed ligt on weter P beavior across countries sows any regularity. 8 Tis cut-off date is due to te limited availability of panel data. It also as te advantage tat countries ratification decisions until tat point in time were less likely to ave been affected by te possibly canging perceived probabilities of te Protocol gaining legal force, as during te Russian ederation sent various conflicting signals regarding its ratification intentions. 9 In te duration analysis, t is te azard of ratification. 30 xamples of TOPLAYRS_ include Nigeria wit, te US and most of te U wit 4, and Denmark and Romania wit 7. India ad te igest score of 5 for most of te sample period. 7

19 However, te correlation between tese P measures is weak Bot are panel data variables. 3 Te next two variables of interest are te sizes of te environmental and firm lobbies. We measure te aggregate size of te environmental lobby NGO by te number of national environmental non-governmental organizations NGOs tat are members of te World onservation Union, a large international organization drawing togeter environmental NGOs from all around te world. 3 or a given population size and oter controls listed below, we view NGO as a proxy for te number of organized environmentalists in a country, i.e. te aggregate intensity of environmental lobbying. 33 To capture te power of te firm industry lobby, we use a number of alternative proxies. Symmetrically to our NGO variable, our first measure reflects firm lobbying strengt troug membersip in te International amber of ommerce I. 34 Several environmental/antiglobalization organizations view I as te main representative of firms interests in te Kyoto process. 35 Since te I as direct access to national governments all over te world troug its national committees we use a dummy to measure weter a country as a national I committee in our sample, 83 countries do. As a less direct measure, we also present results wit a dummy for weter a given country is a fuel exporter UL. Tis reflects te pressure arising from firms directly affected by canges in fuel prices. 36 Several alternative measures are discussed in te robustness section. To isolate te effect of lobbying and political institutions across countries, we control for te general costs and benefits of environmental regulation, as well as environmental preferences of 3 TOPLAYRS_ ad at te time of writing tis paper not yet been computed for 00, terefore we assume tat te 00 values remained uncanged in 00. Tis index is available for 49 countries in our sample. 3 Te World onservation Union claims to be te world s largest and most important conservation network, wit a mission to influence, encourage and assist societies trougout te world to conserve te integrity and diversity of nature see ttp:// Its members include national and international NGOs, government agencies, and scientists in 8 countries. Our NGO measure includes only te national NGOs. 33 A possible obection to tis interpretation is tat in some cases fewer groups migt mean tat te environmental lobby is better organized more concentrated and ence more able to exert political pressure. However, tis would go against finding a significant positive own effect of NGO on ratification. Our robust finding of suc a relationsip below is ard to square wit tis alternative interpretation. NGO s positive correlation wit measures like population, GDP per capita and democratic liberties 0.4, 0.39, and 0.30, respectively also suggests tat larger values of NGO reflect iger and not lower ability to organize political action. 34 Te I claims to be te voice of world business. Because government decisions ave far stronger international repercussions tan in te past I te world's only truly global business organization responds by being more assertive in expressing business views ttp:// 35 See, e.g., ttp:// and ttp:// 8

20 society as a wole. Real income per capita GDPP and population POP are two standard controls. 37 To capture environmental damages, we follow te literature in assuming tat rising sea levels caused by global warming would adversely affect states wit long coastlines redriksson and Gaston, 000; ISLAND is defined as a country s coastline divided by area. 38 In te present context, an important measure of abatement costs is a dummy variable equal to one if te country is in Annex ANNX. Since tese countries face binding requirements wen te Protocol comes into force, teir abatement costs become significantly iger. We also include per capita O emissions OP. Higer levels of OP may imply lower marginal abatement costs, assuming decreasing marginal returns to abatement. 39 We also include a dummy for te former socialist countries in astern urope SLLR, wic ave low marginal abatement costs. Tese are commonly believed to become sellers of tradable permits ence obtain revenues in te international permit trading system establised by te Protocol, and terefore SLLR sould ave a positive effect on te ratification probability. inally, democratic values and institutions ave consistently been found to ave a positive effect on te probability of IA ratification ongleton, 99; Murdoc and Sandler, 997. We terefore include te combined democracy index publised by reedom House DMORAY, wic takes values not free, partially free, and 3 free. 40 Bencmark Results Tables 3 and 4 present bencmark results from pooled Logit regressions. 4 Tis corresponds to asking weter te mecanism suggested by our model is a significant determinant of te probability tat a country ratified te Protocol at some point in time during te first 4.5 years. Tat is, in tese bencmark calculations, we ignore duration. 36 In a Probit analysis, Neumayer 00a found a significant negative effect of UL on te probability of signature of te Kyoto Protocol. 37 or example, if environmental quality is a normal good, we expect citizens in countries wit iger GDP per capita to demand a faster ratification. Similarly, a greater population POP migt imply more exposure to te potentially negative consequences of climate cange, raising te preference for ratification. 38 Te rationale beind using tis ratio rater tan te lengt of coastline is tat, for example, te rise in te sea level is likely to affect te Seycelles islands, wit a coastline of 49 km, more tan Poland. Te latter as a coastline of exactly te same lengt, but is 78 times larger. 39 Alternatively, iger OP may make a country more reluctant to ratify because te cost of inputs may rise relatively more redriksson and Gaston, 000. OP refers to 998 and is time-invariant since no suitable panel data for O emissions is available. 40 We use reedom House s 3-scale classification rater tan te 7-scale classification because te scores are more likely to represent real significant differences between countries. 4 Logit and Probit models are presented by ongleton 99 and Beron et al. 003 wo investigate Montreal Protocol ratification beavior, and by Neumayer 00b wo analyzes te signature rater tan te ratification of te Kyoto Protocol. 9

21 Te first model in Table 3 contains our main controls. ISLAND as a significant positive effect on te ratification azard in every specification. 4 ountries facing larger potential environmental damages due to a ig coastline/area ratio appear more likely to ratify. POPULATION as a significant positive effect, and, consistent wit most of te existing literature, DMORAY is also igly significant in all models. However, GDPP and ANNX are insignificant. Model experiments wit OP emissions and te SLLR dummy. Neiter turns out to be significant. Potential abatement costs or windfall gains do not seem to ave a significant effect on countries ratification beavior towards te Kyoto Protocol. Models 3-6 introduce our measures of Ps and lobby group influence primed models use UL rater tan I. Models 4 and 4 include te interaction of TOPLAYRS_ and NGO, and models 5 and 5 interact TOPLAYRS_ wit te firm lobby variables. inally, models 6 and 6 include bot interactions. Tese models demonstrate te importance of considering lobbying and te institutional environment, as suggested by te teory. NGO is significant and positive only if NGOTOPLAYRS_ is included. Models 4, 4, 6 and 6 suggest tat wile te number of environmental interest groups as a significant positive effect on te ratification probability, te effect is conditional on te number of Ps. In particular, te positive effect of NGO on te ratification probability is reduced in countries wit a greater number of Ps TOPLAYRS_. Tis finding lends support to our teory. A greater number of Ps make lobbying more costly for te environmentalists, wic reduces teir influence in te political process. Turning to te firm lobby measures, wile UL as te expected negative sign trougout, neiter I nor UL is significant, and te interaction terms wit TOPLAYRS_ are all insignificant. Tus, te number of Ps does not appear to ave a significant impact on firm lobbying in te Kyoto context. Table 4 replicates models 3-6 in Table 3, but wit TOPLAYRS_. All significant qualitative findings in Table 3 are robust to tis cange, suggesting a degree of robustness. In addition, in tese models I also as te expected negative sign once its interaction wit TOPLAYRS_ is included, and te interaction term becomes significant in Model 6. Tis finding sows no signs of robustness, owever Trougout, standard errors are adusted for eteroscedasticity and clustered at te country level. 43 Taken togeter, our results suggest tat te strengt of te firm lobby group migt be ard to measure. 0