Political economy, information and incentives

Transcription

1 European Economic Review 43 (1999) Presidential Address Political economy, information and incentives Jean-Jacques Laffont* IDEI.R and GREMAQ, University of Toulouse I, Place Anatole France, Toulouse Cedex, France Abstract This paper shows how the instruments of incentive theory can be used to develop some views about the proper design of governments to avoid the capture of politicians and bureaucrats by interest groups. First, treating politicians as informed supervisors to whom economic policy is delegated we show the usefulness of the separation of powers to increase the transaction costs of collusion, the relevance of asymmetric control for avoiding reciprocal favors. The incompleteness of the constitutional contract leaves discretion to politicians who become residual decision makers. We study the trade-offs between greater efficiency obtained by allowing powerful instruments to politicians and less discretion by restricting on the contrary those instruments. Determinants of those trade-offs are the variability of the environment, the extent of asymmetric information about tastes and technologies and the size of majorities. Finally, we show the new theory of incentives for group behavior can be used to determine the transaction costs of collusion under asymmetric information and to which extent these costs relax the constraints imposed on government by collusive behavior Elsevier Science B.V. All rights reserved. JEL classification: H1; D7 Keywords: Political economy; Incentives; Collusion * Tel.: ; fax: ; boe@cict.fr. I am grateful to Jean-Paul Azam, Bruno Biais, Jacques Crémer, David Martimort and Jean Tirole for their comments on the first draft /99/$ see front matter 1999 Elsevier Science B.V. All rights reserved. PII: S ( 9 8 )

2 650 J.-J. Laffont / European Economic Review 43 (1999) Introduction Modern Political Economy was first developed in countries governed by Monarchs. Economists were looking for good policy rules to run the economy. Whatever their beliefs they could not question the postulate that the Monarch was a benevolent agent of the people. They were led to calling good rules the ones maximizing social welfare defined in one way or another as the welfare of the people. However, economists could question the benevolence of the administration, agent of the Monarch to implement policies. The disfunctionings of administrative bodies in charge of the economy of the country were indeed a driving force leading to the birth of liberalism as an economic policy which minimizes public intervention in the economy. Smith (1776) recognized the role of the State in defense, justice, education, public works and institutional design to facilitate private economic activities, with the necessary tax institutions for financing these public goods. But he was very critical of the public administration of this State intervention and he advocated various policy rules having in mind as a main purpose the design of proper incentives for the various administrative bodies. Let us give two examples. He recommended decentralization for the administration of local public goods as the best way to accommodate those incentives. Public works of a local nature should be maintained by local revenue because The abuses which sometimes creep into the local and provincial administration of a local or provincial revenue, how enormous so ever they may appear, are in reality, however always very trifling, in comparison with those which commonly take place in the administration and expenditure of the revenue of a great empire. (p. 689.) He proposed to finance highways, bridges and canals by tolls proportional to the wear and tear, to pay for the expenses and to balance the budget of each public work. Beyond some equity considerations, Smith s main argument for this financing method rather than using the general revenue of the society is a concern about the incentives of the administration to make the proper investments. When high roads, bridges, canals, etc. are in this manner made and supported by the commerce which is carried on by means of them, they can be made only where that commerce requires them, as consequently where it is proper to make them2 A magnificient high road cannot be made through a desert country where there is little or no commerce, or merely because it happens to lead to the country villa of the intendant of the province, or to that of some great lord to whom the intendant finds it

3 J.-J. Laffont / European Economic Review 43 (1999) convenient to make his court. A great bridge cannot be thrown over a river at a place where nobody passes, or merely to embellish the view from the windows of a neighbouring palace: things which sometimes happen, in countries where works of this kind are carried on by any other revenue than that which they themselves are capable of affording. (p. 683.) Political Economy was already for Adam Smith the design of policies which maximize social welfare under the incentive constraints of the administrative bodies in charge of implementing those policies. Modern incentive theory provides tools to explore analytically this paradigm. As administrative or political bodies are usually intermediaries, between the people and the regulated economic activities, who retain discretionary power, we need to go beyond the principal single agent theory to model the design of incentives for the administration. In the first part of this talk we will present briefly two hierarchical models which can serve this purpose and illustrate how they can be used to optimize the structure of government. Asymmetric information is key to the analysis. Given policy choices, asymmetric information determines the distribution of informational rents, which themselves create the stakes of collusive behavior. The optimization of the structure of government arbitrates between economic efficiency requiring various individual and collective incentives and the costly distribution of information rents. The advent of democracy eliminated the convenient myth of a Monarch maximizing social welfare. The American Federalists addressed the task of designing more democratic institutions which would favor social welfare. The constitution founding a given State was supposed to be the mechanism organizing both the representation of the people by elected politicians and the structure of government. The Federalists were well aware of the fact that interest groups would form to influence policy decisions and that appropriate incentives had to be provided to the politicians and to the administration. The Monarch officially was a perfect judge, perfect representative of the people and perfect decision maker. Now the task was much more difficult. How to organize society when the judiciary, executive, legislative branches of government must be delegated by the people to agents who have their private interests and when interest groups form to capture those agents? Can we conceive the constitution as a grand contract which would maximize social welfare under incentive constraints just as the Monarch was doing? Not easily so for several reasons. The first one is that a contract requires penalties to be enforced and ultimately an outside enforcing agency when penalties are resisted. In the absence of the benevolent Monarch, this raises the issue of the incentives for the judiciary branch. The design of incentives for judges is never directly addressed. Rather, their power is limited by the procedures they have to follow and by various possibilities for appeal they cannot for example participate explicitly as a third party in a contract that two economic agents might want to sign with them to

4 652 J.-J. Laffont / European Economic Review 43 (1999) facilitate commitment. The unfortunate consequence is that we do not have a clear vision of the types of contracts that can be enforced by the available judiciary branch, but they are quite incomplete and country specific. Economists usually pursue the analysis by postulating the existence of a benevolent judicial branch or by putting restrictions on contracts which will appear quite ad hoc as long as an economic theory of judges behavior is not developed. In future research, judges will have to be formalized along the same principles as for other economic agents (appealing to moral hazard, asymmetric information, collusion, reputation, etc.). The second reason is that in practice a constitution is necessarily quite incomplete for bounded rationality reasons and residual control rights must be allocated to politicians, for society to be able to cope with unforeseen contingencies. Governmental bodies acquire in this way a possibility of discretionary actions which must be controlled somehow. In general this control is not exercised by classical contracts with penalties which would put politicians under the dependence of judges, but directly by the people through reelection mechanisms designed in the constitution. Here again the type of contract is very incomplete. A normative theory of constitutional design can still be conceived as an optimization of expected social welfare in which the instruments are the electoral processes to select the various governmental bodies, the policy instruments made available to them, the allocation of residual decision rights among them, the checks and balances between the various branches of government under the individual and collective incentive constraints of the various members of society, voters, politicians, bureaucrats, etc. Unfortunately it will be highly dependent on various incomplete contracting assumptions which in general take the form of ad hoc restrictions on the constitutional instruments. In the second part of this talk, we will illustrate with a simple electoral system, the random majority model, one dimension of this type of optimization which trades-off flexibility and discretion in the delegation of economic policy to the winning majority. Information economics again provides natural foundations for the rents influenced by economic policy variables for the control of which majorities fight within the electoral system. The difficulty of the exercice leads many political scientists and economists to favor a more positive approach trying to analyze various existing political institutions. Also, we do not ignore the more pessimistic view articulated by John Quincy Adams according to whom the constitution was calculated to increase the influence, power and wealth of those who had it already. However, it does not seem useless to understand how the constitution should theoretically be designed both to evaluate the current institutions and to make citizens aware of an ideal point toward which they should strive. Furthermore, the positive problem faced by a ruling group in designing its best constitution is of a similar nature as the one studied except for the objective function to be maximized. See also Buchanan and Tullock (1962) and the political economy literature in macroeconomics.

5 J.-J. Laffont / European Economic Review 43 (1999) Beyond the individual stakes for policy choices which can be translated into votes by the political system, groups form to influence the politicians and the bureaucracy. Interest groups can be of different kinds, from the usual political parties which officially participate in the electoral process to consumer groups and producer groups which invite themselves into the process of collective decision making with bribes, campaign contributions and other favors. The lack of satisfactory theories of group behavior is often stressed in political economy: an explicit modelling of coalition formation would surely add to the power of the approach. (Becker, 1983.) The plain truth is that economists know very little about the dynamics of group formation and action. (Tollison, 1989.) In the third part of this paper we will show how recent developments in the theory of incentives for groups can be used to determine the transaction costs of coalition formation stressed by Olson (1965) in his theory of interest groups. 2. Politicians as informed supervisors 2.1. A hierarchical model of government Let us remain for a while in the paradigm of a benevolent constitution which maximizes social welfare under incomplete information, but with no restriction on available instruments. The politician is viewed as an informed supervisor. His role is to partially bridge the constitution s informational gap by obtaining an informative signal. For example, to choose the size of the nuclear program in the 21st century, the politician will gather information about the cost of the future Franco-German nuclear power plant. More generally let q be the size of a public good providing utility S(q) to citizens, S'0, S(0. The production cost of q is θq where θ can take two values θ or θm with θm 'θ and Δθ"θM!θ. θ is private information of the firm producing the public good with ν"pr(θ"θ ) which is common knowledge. To procure this public good the State (constitutional level) gives a transfer t to the firm which has a utility level º"t!θq. Because of distortionary taxation to raise public funds the cost for consumers of this transfer is (1#λ)t, λ'0. The politician obtains a signal σ about θ and receives a transfer s from the State as a function of his report r on σ yielding a utility level»"v(s).

6 654 J.-J. Laffont / European Economic Review 43 (1999) Fig. 1. Utilitarian social welfare is ¼"S(q)!(1#λ) (t#s)#º#». We obtain the following hierarchical structure (see Fig. 1). An optimal constitution must design the incentive structure of the politician and the regulation of the firm on the basis of the information provided by the politician The hard information model ( affont and ¹irole, 1993) Let σ3θ, the signal observed by the politician. More precisely, with probability 1!ξ the politician observes nothing,, and with probability ξ he observes the true value of θ and furthermore he can provide verifiable information about the value of θ. His discretion is that he can pretend he has observed nothing. Suppose first that the politician is benevolent. If σ"θ then his report r is θ and the optimal full information regulation entails S(q)"(1#λ)θ, t"θq, i.e., the marginal social utility of the public good equates the marginal social cost (the Samuelson condition) and º"»"0 and no rent is given up to the firm or to the politician. If σ", then r" and the constitution designs the optimal Baron Myerson regulation (Baron and Myerson, 1992) S(q)"(1#λ)θ, ν S(qN )"(1#λ)θM # λ 1!ν Δθ, This model is based on Tirole (1986).

7 º "Δθ qn, ºM "0. An informational rent must now be given up in general to the efficient firm (º) and, to decrease this rent which is costly for social welfare because of the cost of public funds, the production level of the inefficient type firm (qn ) is decreased. Political economy consists here in recognizing that, unless motivated otherwise, the politician will be captured by the industry and led to protect the firm s rent º, by suppressing the report r"θ. Let us assume v(s)"s50, i.e., risk neutrality with a limited liability constraint. Capture is avoided if the politician s reward when he reports r"θ, i.e., s,is larger than the value for him of the maximal bribe the firm is willing to pay, or s5kδθ qn. k3[0, 1] formalizes in a black box manner the transaction cost of collusion. The lower k the higher the transaction cost. It depends on the morality of the politician but also on the probability of being caught colluding as well as on the constraints which might exist on the type of side transfers possible between the firm and the politician (for example monetary transfers might be too dangerous). The optimal constitution which avoids capture takes into account the expected social cost of these additional transfers λνξkδθ qn J.-J. Laffont / European Economic Review 43 (1999) which occur only when the firm is efficient and identified as such by the politician (with probability νξ). The political economy distortions are two-fold: First, an incentive payment of s is made to the politician when he reports θ. Second, the production level of the inefficient type qn is decreased to account for this additional cost which is increasing in qn S(qN )"(1#λ)θM # λ(1!(1!k)ξ) ν 1!ξ 1!ν Δθ. One feature of this new regulation is that incentives for production are lower (the society is less efficient, more bureaucratic) than under the paradigm of a benevolent politician. Furthermore, the expected rent of the firm is lower as well as the citizens welfare. The firm and consumers suffer from the possibility of collusion. There is no stake of collusion if θ"θm since the firm s rent is zero whatever the politician s report.

8 656 J.-J. Laffont / European Economic Review 43 (1999) Remark 1. The transfer to the politician may be implicit. If he has a private benefit B of holding office, we require that the enhancement of his probability of reelection ΔΠ when he proves that he does well his job (reports θ) is such that ΔΠ ) B"s. To complete the analysis one must be clear about the social cost of those private benefits. A further difficulty is that ΔΠ is endogenous. Remark 2. Tirole (1986) proves a principle of collusion proofness to show that there is no loss of generality in restricting the analysis to collusion-proof allocations. More general mechanisms in which the constitution would try to elicit if the politician and the firm have entered a collusive agreement are not considered since they could be nullified by the colluding partners. Remark 3. It is assumed that the politician and the firm can enter side contracting up to the exogenous discount factor k. A more complete analysis would require a dynamic framework where such side contracting emerges through reputation or an endogeneization of k through asymmetric information as we do in the next section The soft information model (Faure-Grimaud et al., 1998) Quite often the information gathered by policians is soft. In the above model no use of the information could then be made. However, a role for politicians is restored, even with soft information, in the following setup. The politician is now supposed to be risk averse, for example v(s)"1!e, i.e. constant absolute risk aversion. The information technology entails two signals σ and σn with for simplicity Pr(σ/θ)"Pr(σN /θm )"Π5. A benevolent politician always transmits his signal and, after updating its beliefs, the constitution designs the optimal Baron Myerson regulation conditional on the signal transmitted. With a nonbenevolent politician who may collude with the agent, to create incentives for truthful transmission of the signal requires creating some risk for the risk averse politician. The politician is delegated the choice of the agent s contract which then plays the same role as effort in the standard moral hazard problem. The optimal trade-off between the best possible use of the signals and the risk-premium required for information extraction determines the optimal constitutional design. Collusion between the politician and the firm occurs here under incomplete information. The See Laffont and Tirole (1993, Chapter 16), and Boyer and Laffont (1998). See Tirole (1992) and Martimort (1997).

9 transaction costs of collusion are now endogenous and depend on this asymmetric information and on the risk aversion of the politician. The collusion proof constraint reduces here to s 5k Δθ Δq J.-J. Laffont / European Economic Review 43 (1999) where s is the transfer when the report is (θ, σ) and Δq"q(θM, σn )!q(θm, σ) is the stake of collusion when the firm claims θm and the politician hides the signal σ which increases the posterior probability that θ"θ (i.e. reports σ instead of σ). The interesting part is that, for small Δθ, Δq, r: k+ 2 r νπ 1# Δθ Δq. (1!ν)(1!Π) We have now a more fundamental theory of the distortions due to political economy since the transaction costs of collusion are endogenous. For example if politicians are very risk averse, they become useless as informational supervisors since k becomes close to one. Another interesting result is that the optimal information technology Π is neither the uniformative one (Π") nor the perfectly informative one (Π"1). The principal does not want his agents to know each other too well because this facilitates their collusion against him. Equipped with this basic modeling for the optimal design of delegation of supervisory functions to politicians, one can enlarge the analysis to study the structure of government. We will give two such examples An incentive theory of the separation of powers Let us extend the model, in the hard information case, to a cost parameter θ which can take three values θ, θk, θm with θk "θm!δθ, θ"θm!2δθ. For example one can think of the firm s production process as having two stages. In the first one, one unit of intermediate output is produced at a positive cost θm /2!θ, then it is transformed with a Leontieff technology into one unit of final output and the marginal cost of transformation is θm /2!θ, where θ and θ are random variables drawn independently from the same probability distribution on 0, Δθ with respective probabilities (1!ν, ν). Then Pr(θ)"ν, Pr(θK )"2ν(1!ν), Pr(θM )"(1!ν). The supervision technology is composed of two technologies ¹, i"1, 2. Technology ¹ provides with probability ξ hard information about the value of θ. See Laffont and Martimort (1996).

10 658 J.-J. Laffont / European Economic Review 43 (1999) Fig. 2. The question is then: should we have one politician associated with each supervision technology, a separation of powers, or a single politician with the two technologies? Note that the absence of correlation between the two pieces of asymmetric information precludes any yardstick competition argument for separation. Also the benevolence of the constitutional level eliminates any usefulness of multiplying counter-powers to the State. An argument may exist when we recognize that politicians can be captured if a separation of powers between politicians makes capture more difficult. We must compare the optimal collusion-proof constitution when there is a single politician (the integration case) just as we did above but for three possible values of the cost parameter with an alternative structure of two politicians (see Fig. 2), the separation case. For example, the firm is controlled both by the Ministry of Industry and by the Ministry of the Environment. Politicians will make collusive offers to the firm. To the extent that politicians do not collude between themselves or at least collude imperfectly because of the asymmetric information existing between them we can expect some inefficiencies in side contracting with the firm, making less costly for the constitution to design collusion proof institutions. Separation dominates integration in particular when politicians adopt a safe behavior by making only collusive offers which are accepted with probability one. The gain of separation increases with the efficiency of collusive transactions (k) and with the cost of public funds (λ) which can be argued to be high in developing countries. Our conclusion runs against some claims in the literature (Shleifer and Vishny, 1993) according to which centralized corruption is better. Centralized See Neven et al. (1993) for an informal discussion of yardstick competition between regulators. Also, see Persson et al. (1996) for a theory of the separation of powers based on the incompleteness of the constitution.

11 corruption is indeed more efficient, but the constitution or regulator may exploit the inefficiencies of decentralized corruption to eliminate it more easily Checks and balances J.-J. Laffont / European Economic Review 43 (1999) The supervision model of Section 2.1 can be extended one step further to consider three politicians who have private information about their own department and can also exert supervisory functions about the other ones. One may want to compare various designs (see Figs. 3 5). The reciprocal supervision of politicians 2 and 3 is dangerous because it may give rise to reciprocal favors. It can be argued that the transaction costs of collusion in supervision are low when supervision is reciprocal (a simple phone call can settle a collusion in which each politician hides the signal about the other one). Fighting such a collusion is then very costly. It may be better to have an asymmetric control to avoid the opportunity of reciprocal favors (see Laffont and Meleu, 1997). In the new structure politician 1 is not supervised while politician 3 is supervised twice. This should be often dominated by a more symmetric structure: The President supervises the Judiciary Power which supervises the Legislative Power which supervises the President. However, such a structure raises again the possibility of a now 3-party collusion against the constitution with a low transaction cost because of the possibility of indirect reciprocal favors. Fig. 3. Fig. 4.

12 660 J.-J. Laffont / European Economic Review 43 (1999) Fig. 5. To summarize, when politicians are viewed as informed supervisors optimizing the structure of supervision is part of the task of an optimal constitution. 3. Flexibility versus discretion in constitutional design A next step in political economy is to recognize that the constitution is not a complete contract which maximizes expected social welfare under incentive constraints. There may be different reasons for this. One point of view is that the constitution is not written behind the veil of ignorance, and that citizens with different preferences and private information do not agree on a social welfare function to be maximized. However, they agree on voting rules to select politicians who are residual decision makers under some general constitutional constraints such as nondiscrimination, budget balance, polluter pay principle, etc. In a dynamic world, elections constitute the incentive scheme which aligns to some extent the interests of politicians with those of society. Here we would like to maintain the myth of the veil of ignorance and rationalize the incompleteness of the constitution by a precise noncontractibility assumption. This will still provide a possible approach to normative political economy The random majority model Let us return to the public good provision model and let us now distinguish two types of citizens. There is a proportion α of type 1 citizens with utility function S(q) and a proportion (1!α) of type 2 citizens with utility function βs(q), β'1. See also Aghion and Bolton (1998). See Laffont (1996).

13 Citizens agree ex ante on an utilitarian social valuation of the public good which is [α#(1!α)β]s(q). A complete constitution would maximize, for each value of λ, expected social welfare under the incentive and participation constraints of the firm, hence the Baron Myerson regulation: [α#(1!α)β]s(q)"(1#λ)θ, [α#(1!α)β]s(qn )"(1#λ)θM # λν 1!ν Δθ. Suppose that the cost of public funds λ, related to the business conditions, is unknown ex ante and noncontractible. We consider two approaches. The first one is to select ex ante an unconditional policy by maximizing the expectation over λ of social welfare hence [α#(1!α)β]s(q)"(1#eλ)θ, [α#(1!α)β]s(qn )"(1#Eλ)θM # νeλ 1!ν Δθ. Alternatively, one can make the policy conditional on λ by making a third party informed ex post about λ a residual decision maker about the policy. However, as the third party will be able to claim any level for λ, this enables it to optimize its own objective function, i.e., introduces discretion. The constitution will define the process by which this decision maker is selected. Suppose that the proportion α of type 1 citizens fluctuates randomly each period independently. With probability 1/2, α"α*', J.-J. Laffont / European Economic Review 43 (1999) /2, α"1!α*(. In such a context it is natural to settle for a majority system which gives, each period, the decision power to the majority. Each majority in turn delegates to a politician the role of residual decision maker. We assume that, due to reelection constraints, the politician faithfully represents the interests of the majority which has selected him. To complete the model we assume that the firm producing the public good is itself public. The majority in power can appropriate the information rent of the firm, for example by organizing among themselves an auction to choose the manager who is willing to pay ex ante the expected information rent he will capture. Furthermore, the transfers paid to the firm are financed by indirect taxes with a deadweight loss λ and the constitution imposes non discriminating indirect taxes.

14 662 J.-J. Laffont / European Economic Review 43 (1999) Suppose α"α* and majority 1 is in power. It maximizes the expected welfare of type 1 consumers under the incentive and participation constraints of the firm, i.e. s.t. max ν[α*(s(q)!(1#λ)θq)#º ]#(1!ν)[α*(S(qN )!(1#λ)θM qn )#ºM ] Hence ºM 50, º 5ºM #Δθ qn. S(q)"(1#λ)θ, S(qN )"(1#λ)θM # [(1#λ)α*!1]ν Δθ. α*(1!ν) Similarly, for majority 2, we obtain βs(q)"(1#λ)θ, βs(q)"(1#λ)θm # [(1#λ)α*!1]ν Δθ. α*(1!ν) When β"1, public good production for θ"θm is higher than under ex post welfare maximization for both majorities. The reason is that each majority in turn appropriates fully the public firm s information rent, hence a per capita rent of º /α* larger than if it is shared equally in the whole population. Each majority overvalues the information rent in the rent efficiency trade-off and produces too much. Note that the lower α* is, the higher the distortion. When α*"1, there is no distortion because each majority in turn represents the whole population. When α*"1 and β'1, majority 1 produces too little public good in both states of nature and majority 2 too much. In general, the two above effects combine. Delegating production decisions to majorities introduces useful flexibility but also some detrimental discretion. The constitution must arbitrate. Clearly, for any α*(1 and β'1, delegation is dominated by the inflexible constitution for a variance of λ small enough. For any strictly positive variance of λ, delegation dominates the inflexible constitution for β close enough to 1 and α* close enough to 1. For simplicity we assume (1#λ)α*'1. It could also envision more general majority rules.

15 3.2. Constitutional choices J.-J. Laffont / European Economic Review 43 (1999) Suppose that the previous analysis concludes that delegation of public policy is desirable in view of the large uncertainty existing on the noncontractible parameter λ. One may wish to impose further constitutional constraints on politicians than nondiscriminatory indirect taxes. Let us consider a few examples Private versus public ownership If the type 2 consumers own the firm producing the public good, the behavior of majority 2 will be the same as under public ownership since in both cases it appropriates the rent of asymmetric information. On the contrary majority 1 which cannot capture the information rent undervalues it and regulates the firm in a way which results in underproduction. Clearly, ownership matters as in any incomplete contract model. Public or private ownership result in different patterns of distortions and the constitution may want to select ex ante between the two forms of ownerships Incentives for bureaucrats against capture Let us now introduce a hierarchical government with a regulatory commission between the elected politicians and the firm, and let us allow for capture of the commission by the firm. One may argue that, if majority 2 has control, the firm will not enter any collusive agreement with the commission because the political principal coincides with the owner of the firm. On the contrary, under majority 1, the firm will wish to capture the commission to protect its rent. Two constitutional choices can be compared. Either one creates costly incentives for bureaucrats which avoid capture. Such expenses are very useful with majority 1 but useless with majority 2. Or one gives up the supervision role of the bureaucracy with majority 1, by not providing incentives for bureaucrats and letting them be captured in this case. The choice between the two types of bureaucracies relies on an ex ante comparison between supervision gains, transaction costs of capture, the possible regulatory responses to capture and the costs of incentives for the regulatory commission Political price discrimination versus uniform pricing Suppose now that the good produced is private and that the observability of individual consumption (as for electricity, telephone, etc.) makes second degree price discrimination possible. See Laffont (1996) for more details.

16 664 J.-J. Laffont / European Economic Review 43 (1999) Letting price discrimination be used by the political principal is useful because this added flexibility enables to cover the cost of production in a less distortionary way. However, the political price discrimination will be excessive. With the preferences chosen here, type 2 agents who value the most the good are the ones who will have an information rent because of the impossibility of first degree price discrimination. Accordingly, type 1 majority will undervalue the information rent that type 2 capture and this will result in more price discrimination than desirable to decrease this rent. On the contrary type 2 majority overvalues this rent and this results in less price discrimination. Political discrimination is excessively volatile and a constitutionally imposed price uniformity may be better when the political conflicts are high (α* close to 1/2), the cost of public funds is low and its variance also low The marginal cost pricing controversy Coming back to Smith s suggestion of average cost pricing instead of marginal cost pricing, we see now how to model the choice between, say, average cost pricing (the Smith rule) and marginal cost pricing with deficits funded by costly public funds (the Hotelling rule). The delegation of decision making to politicians over the projects to be realized introduces more discretion with the Hotelling rule even though it corresponds to a more efficient pricing rule. For example for small fixed costs one can show that the efficiency losses of the Smith rule are of the second order while those of the Hotelling rule may be of first order, in some circumstances Emergence of environmental incentive regulation It is often the case that efficient rules desirable to deal with complex problems such as the internalization of externalities under incomplete information create more possibilities of manipulation by residual decision makers than cruder rules. With the approach proposed here one can shed some light on the debate about the choice of instruments in environmental regulation which remains to a large extent a mystery so far. To summarize, asymmetric information convexifies the set of reachable utility levels and the fluctuations of policies induced by the majority system become costly. It may be better to limit the ex post efficiency adjustments for insurance purposes. See Laffont (1998). Such an approach follows the pioneering paper of Buchanan and Tullock (1975). See Boyer and Laffont (1998).

17 4. Constitutional response to interest group behavior Politics is about collective decision making. To understand how political parties and interest groups form to influence the process of decision making is at the core of political science and political economy. An example of insight is due to Olson (1965). He has argued that small groups have lower per capita transaction costs and this argument is often used to explain why taxpayers often do not form an interest group while managers of an industry do. As economists we believe that the free rider problem is the major impediment to good collective decision making. The literature on public goods has made a lot of progress in the understanding of mechanisms able to elicit truthful information but it has largely neglected group behavior. Beyond individual incentives to truthful behavior in a collective decision mechanism, we must take into account group incentives. The asymmetric information which exists between agents is a major determinant of the transaction costs affecting group behavior. Accordingly, a good step in the understanding of the formation of political parties and interest groups is a theory of coalition formation under asymmetric information A model of interest group behavior Let us return to the paradigm of a benevolent State which, through a benevolent bureaucracy, regulates two industries which participate in the production of a public good. Each firm produces essential inputs for this good. Firm 1 produces a quantity q of an intermediate good (good 1) which is used by firm 2 to produce a quantity q of final good (good 2). The production technologies are Leontieff and one to one and we denote q"q "q. Each firm i, i"1, 2 (we aggregate each politician with a firm) has private information about its constant marginal cost θ. The firm s cost characteristics, θ, i"1, 2, are drawn from a common knowledge joint distribution on Θ where Θ"θ, θm is the common support of θ and θ. We refer to the probabilities p(θ, θ ) of each state (θ, θ ) as the firms prior beliefs. We denote p "p(θ, θ), p "p(θ, θm )"p "p(θm, θ), p "p(θm, θm ). Finally, to capture the congruence of the firms interests, θ and θ are positively correlated and ρ"p p!p 50 is the degree of positive correlation. Firm i s welfare is º "t!θ q and social welfare is J.-J. Laffont / European Economic Review 43 (1999) ¼"S(q)!(1#λ)(t #t )#º #º. However, see Stigler (1971) and Peltzman (1976).

18 666 J.-J. Laffont / European Economic Review 43 (1999) When extending the principal agent model to multi-agent situations, the theory has largely ignored the possibility of side-contracting by postulating a Bayesian-Nash behavior of the agents. It amounts to presume that the principal can control communication between agents, an often unrealistic assumption. Let us assume at the opposite that the firms can freely communicate and furthermore let us consider the following very favorable modeling of their group behavior. Consider first a mechanism G proposed by the State regulation, which maps any pair of messages (m, m )inm M into a triplet (q, t, t ) of public good production and of transfers paid by the principal to the firms. An uninformed third party, ¹, proposes a side mechanism S composed of a collective manipulation of the messages to the principal and a pair of balanced side transfers y ( ) ), y ( ) ). From the revelation principle there is no loss of generality in assuming that S is a direct mechanism and that and y ( ) ), i"1, 2, map Θ into the set of measures on M and the set of balanced side transfers. Lastly, the third party is benevolent and maximizes the sum º #º of the two colluding firms utilities obtained in playing the composition of the grand mechanism and the side-mechanism. Note first that if there was no collusive behavior, according to the Crémer McLean theorem, and because of the correlation of types (ρ'0), the principal could costlessly extract all the information rents of the agents, an extreme form of yardstick competition. When types are uncorrelated (ρ"0), on the contrary, information rents must be given up and call for output distortions to mitigate those rents. We have therefore a striking discontinuity in the results at ρ"0. How can we design the optimal mechanism when types are correlated and collusion is possible? A major step in the analysis is to establish the weak-collusion proof principle which shows that any perfect Bayesian equilibrium of the two stage game of grand contract offer followed by the third party s side contract offer can be represented by a revelation mechanism offered by the principal for which the best strategy of the third party is to do nothing, i.e., a collusion-proof revelation mechanism. This is analogous to the Revelation Principle and provides a constructive approach to the characterization of the PBE of the game. In addition to the classical individual incentive constraints, we must write the coalition incentive constraints. The optimal mechanism maximizes social welfare under those constraints and the firms participation constraints The structure of optimal mechanisms Collusion-proof constraints matter a lot in shaping the optimal institutions. First the discontinuity at ρ"0 disappears. For ρ small, the collusion-proof See Laffont and Martimort (1998).

19 J.-J. Laffont / European Economic Review 43 (1999) constraints prevent the use of the penalty system embedded in the Crémer McLean mechanism (Cre mer and McLean, 1988), a rather satisfactory solution of this puzzle which explains to some extent the lack of practical success of explicit yardstick mechanisms. The possibility of collusion enables the two firms to collectively extract rents from the principal, i.e., to succeed in forming a pressure group. The output distortions depend of course on the degree of correlation. The usual no distortion at the top subsists and S(qL )"(1#λ)(θ#θM )#λδθ p 2p 1# p p #ρ, S(qN )"2(1#λ)θM #λδθ p #2p!p p /(p #ρ) p, for λ small and when the correlation is small. When the correlation is higher, the mechanism becomes less responsive to information and partial pooling occurs. On the other hand for almost perfect correlation, the principal can afford killing the projects when the reports of the two firms differ and this enables him to approach the first best when ρ goes to 1, i.e., to extract all the information rents. The optimal constitutional response to collusive behavior is then characterized. The usual distortions implied by the free rider problem come from a conflict between the individual incentive and participation constraints. Here, they are quite different and come from a conflict between group incentive constraints and individual participation constraints. These results show how important it is to take into account the limits put by asymmetric information on collusion. For example, for ρ close to one we have a striking difference between the case where collusion takes place under complete information (i.e., without transaction costs) where the principal chooses to have substantial distortions and the case of collusion under incomplete information where the principal approaches the first best. 5. Concluding remarks Political economy is the study of the two-way interaction between economics and politics. We have attempted to show how the instruments of incentive theory may be used to develop some views about the proper design of governments to avoid the capture of politicians and bureaucrats by interest groups. The need for setting up incentives for these agents of the people and for rebalancing the stakes of collusion even at the cost of some allocative inefficiencies, the usefulness of the separation of powers to increase the transaction costs

20 668 J.-J. Laffont / European Economic Review 43 (1999) of collusion, the relevance of asymmetric control for avoiding reciprocal favors are the type of insights that can be derived from an analysis within the usual complete contract paradigm. Political economy is then narrowly defined as the study of the incentive problems due to the delegation of economic policy to politicians treated as economic agents. We also argued that there are a number of reasons (bounded rationality, unforeseeable contingencies) which make the constitution a very incomplete contract and require a broader framework. Even if we attempt to stick to the myth of a benevolent constitution under the veil of ignorance, many degrees of freedom appear in the analysis. Then, we have taken as given an electoral system which selects the politicians who become residual decision makers, and corresponds to a particular specification for the incompleteness of the constitutional contract. Within this framework, we have studied various trade-offs between greater efficiency obtained by allowing powerful policy instruments to politicians and less discretion by restricting on the contrary those instruments. Among the determinants of these trade-offs we have noted the variability of the environment, the extent of asymmetric information about tastes or technologies and the size of the majorities. Coalitional behavior is present all along, sometimes accepted in the case of political parties and some interest groups like unions, syndicates, sometimes rejected, when economic interest groups collude with political parties. Taking into account such collusions is necessary in a proper design of governmental institutions. We have shown how new advances in the theory of incentives for group behavior can be used to determine the transaction costs of collusion induced by asymmetric information and to which extent these costs relax the constraints imposed on governments by collusive behavior. In particular, we have provided a reason for the limit in the effectiveness of yardstick competition as an instrument for the constitutional designer. All these elements are modest steps toward a normative political economy. References Aghion, P., Bolton, P., Incomplete social contracts, Mimeo. Baron, D., Myerson, R., Regulating a monopolist with unknown costs. Econometrica 28, Becker, G., A theory of competition among pressure groups for political influence. Quarterly Journal of Economics 97, Boyer, M., Laffont, J.J., Toward a political theory of the emergence of environmental incentive regulation. Rand Journal of Economics, forthcoming. Buchanan, J., Tullock, G., Polluters profits and political response: Direct control versus taxes. American Economic Review 65, Buchanan, J., Tullock, G., The Calculus of Consent. The University of Michigan Press, Ann Arbor, MI.

21 J.-J. Laffont / European Economic Review 43 (1999) Crémer, J., McLean, R., Full extraction of the surplus in Bayesian and dominant strategy auctions. Econometrica 56, Faure-Grimaud, A., Laffont, J.J., Martimort, D., A theory of supervision with endogenous transaction costs. Mimeo. IDEI, Toulouse. Laffont, J.J., Industrial policy and politics. International Journal of Industrial Organization 14, Laffont, J.J., Frisch, hotelling and the marginal cost pricing controversy. Symposium at the Centennial of Ragnar Frisch, Cambridge University Press, Cambridge, forthcoming. Laffont, J.J., Martimort, D., Separation of regulators against collusive behavior. Rand Journal of Economics, to appear. Laffont, J.J., Martimort, D., Mechanism design with collusion and correlation. Econometrica, forthcoming. Laffont, J.J., Meleu, M., Reciprocal supervision, collusion and organization design. Scandinavian Journal of Economics 99, Laffont, J.J., Tirole, J., A Theory of Incentives in Procurement and Regulation. MIT Press, Cambridge, MA. Martimort, D., The life cycle of regulatory agencies: Dynamic capture and transaction costs. Mimeo. Neven, D., Nutall, R., Seabright, P., Regulatory capture and the design of European merger policy. Merger in Dayligh. CEPR, London. Olson, M., The Logic of Collective Action. Harvard University Press, Cambridge, MA. Peltzman, S., Toward a more general theory of regulation. Journal of Law and Economics 19, Persson, T., Roland, G., Tabellini, G., Separation of powers and accountability: Towards a formal approach to comparative politics. Mimeo. Shleifer, A., Vishny, R., Corruption. Quarterly Journal of Economics 107, Smith, A., The Wealth of Nations. The Modern Library, New York. Stigler, G., The economic theory of regulation. Bell Journal of Economics 2, Tirole, J., Hierarchical and bureaucracies: On the role of collusion in organizations. Journal of Law, Economics and Organization 2, Tirole, J., Collusion and the theory of organization. In: Laffont, J.J. (Ed.), Advances in Economic Theory, vol. 2, Cambridge University Press, Cambridge, pp Tollison, R., Rent seeking, Ch. 23. In: Mueller, D. (Ed.), Perspectives on Public Choice. Cambridge University Press, Cambridge.