WPS4090 Development and the Interaction of Enforcement Institutions * Amrita Dhillon University of Warwick Jamele Rigolini The World Bank Abstract We examine how institutions that enforce contracts between two parties, producers and consumers, interact in a competitive market with one-sided asymmetric information and productivity shocks. We compare an informal enforcement mechanism, reputation, the efficacy of which is enhanced by consumers investing in "connectedness," with a formal mechanism, legal enforcement, the effectiveness of which can be reduced by producers by means of bribes. When legal enforcement is poor, consumers connect more with one another to improve informal enforcement; in contrast, a well-connected network of consumers reduces producers' incentives to bribe. In equilibrium, the model predicts a positive relationship between the frequency of productivity shocks, bribing, and the use of informal enforcement, providing a physical explanation of why developing countries often fail to have efficient legal systems. Firm-level estimations confirm the partial equilibrium implications of the model. JEL: D02, D7, L14, O12. Keywords: Contracts, Institutions, Corruption, Reputation, Uncertainty. World Bank Policy Research Working Paper 4090, December 2006 The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ.worldbank.org. *We thank Franklin Allen, Jonathan Cave, Philip Keefer, Andy Newman, Carlo Perroni, Johannes Hörner, Roger Myerson, Jennifer Reinganum, Myrna Wooders, seminar participants at Georgetown University, Northwestern University, The Harris School of Public Policy, University of Warwick, Vanderbilt University, the 2005 PET conference, the 2005 Conference on Liberalisation Experiences in Asia, and especially Avinash Dixit for extremely useful comments. 1 Introduction Good institutions have long been recognized as crucial for economic transactions. A large literature has by now documented people's reliance on informal enforcement mechanisms when formal mechanisms work poorly.1 However, less attention has been given to the details of their interaction: how do institutions affect one another? How do they evolve over time? And how do they vary with the level of development? Institutions are not monoliths. They are influenced by technology, politics, and by people's transactions.2 It is on the latter aspect that we focus attention. In this paper we examine how institutions that enforce contracts under asym- metric information, such as courts of law and informal community enforcement, interact with one another, and how economic development - measured, as in Kremer (1993), by the frequency of bad productivity shocks in the production process - af- fects this interaction. We consider two mechanisms that enforce contracts between two parties, producers and consumers, when consumers cannot observe quality: a "formal" enforcement mechanism, legal enforcement, and an "informal" mechanism, reputation. By studying how those institutions affect one another, and how the fre- quency of productivity shocks affects the interaction between the two institutions, we provide a theory of how alternative contract enforcing institutions evolve and interact as a country develops. Our theoretical framework builds on the reputation model of Allen (1984) to al- low for legal enforcement, consumers' investment in connectedness, and stochastic, firm-level productivity shocks.3 We consider a competitive economy where firms produce a good of variable quality, and consumers can observe the quality only after they have bought the good. Bad quality arises either because of a bad produc- tivity shock, or because firms did not put in the effort required to produce high quality, so that a one-sided asymmetric information problem arises. In such an environment, we look at how the efficiency of legal enforcement affects consumers' incentives to "connect" with other consumers to find out about underperforming firms. Vice-versa, we study how the efficiency of the reputation mechanism affects 1See, for instance, Greif (1993) and Greif et al. (1994) for examples of contract enforcement in the Medieval Age; Esfahani (1991) for a discussion on informal enforcement mechanisms in developing countries; Battigalli and Maggi (2004) for how uncertainty and costly contracting influence the choice of contract; Besley et al. (1994) for a comparison of ROSCAS and credit markets; Kandori (1992) and Ellison (1994) for a discussion of the role of community enforcement in repeated games; and the discussions on contract enforcing institutions in Mookherjee (1999), Dixit (2003), Greif (2004), and MacLeod (2006). 2See, for instance, North (1981), Kranton (1996), and Acemoglu et al. (2005). In North (1981) technological progress drives institutional change. In Kranton (1996) the use of reciprocal exchange decreases with the use of money-based exchanges, and in Acemoglu et al. (2005) property rights are endogenously derived from the distribution of political power. 3See also Klein and Leffler (1981), Shapiro (1983), H¨orner (2002), and Kranton (2003). 2 firms' incentives to bribe the judicial system. Therefore, as both consumers' con- nectedness decisions and firms' bribing decisions depend on the relative efficiency of the two mechanisms, formal and informal mechanisms interact via consumers and firms' investment decisions. The model has both "partial" and "general" equilibrium implications. At the partial equilibrium level, results only partly confirm the common belief that formal and informal institutions substitute each other. More precisely, we find that con- sumers invest more in building informal enforcement networks when legal enforce- ment works poorly; on the other hand, however, well-performing informal enforce- ment networks improve legal efficiency, as firms have less incentives to bribe court officials. This last result is consistent with the analyses of Putnam et al. (1993) and Knack and Keefer (1997), who find a positive relationship between social capital (interpreted, however, as trust), institutional quality, and economic performance. At the general equilibrium level, we study how the equilibrium institutional mix is affected by uncertainty in the production process, measured by the frequency of bad productivity shocks. Shocks can be of different types, such as physical (e.g. electricity shortages, transportation difficulties, unskilled labor), socio-political (e.g. crime and corruption), or stemming from policy uncertainty.4 Ceteris paribus, when the frequency of productivity shocks increases, firms bribe more, and consumers connect more with one another. Adding up all the partial effects, we obtain that - up to a certain threshold - increased uncertainty leads to lower legal efficiency, and to increased reliance on informal enforcement mechanisms. Therefore, in accordance with common wisdom, in equilibrium corruption in the legal system is associated with a more widespread use of informal enforcement, via increased uncertainty in the production process. To conclude, we present empirical evidence of the link between uncertainty, cor- ruption, and consumers' incentives to rely on informal enforcement mechanisms. We use firm-level data from the Investment Climate Assessments of the World Bank, which survey more than 28,000 firms in 58 countries, and provide information on firms' characteristics, firms' productivity, and the investment climate. In a context of intra-industry trade, firms are both producers and consumers: we use therefore membership to business association as a proxy for the use of informal enforcement networks. In accordance with the theory, we find that firms that suffer more from corruption, and firms that face a more uncertain production climate, are more likely to be members of business associations. Interestingly, the relationship continues to hold when we look at firms in Sub-Saharan Africa, Asia, Latin America, and East- 4A growing body of literature documents how policy uncertainty is a serious concern for busi- nesses in developing countries. See, for instance, Hallward-Driemeier and Stewart (2004), World Bank (2005), and The Economist Intelligence Unit (2005). 3 ern Europe alone, suggesting that the results are not driven by regional attributes. To the extent that developing countries face higher uncertainty in the production process, both the theoretical and empirical analyses point therefore to a physical ex- planation (i.e., unreliable production processes) for why developing countries often fail to have well-performing legal systems. The paper is organized as follows. Section 2 presents the basic model under exogenous institutions; Section 3 endogenizes connectedness and legal efficiency; Section 4 presents the empirics, and Section 5 concludes. 2 Contract Enforcement under Imperfect Institutions The economy consists of a measure one of consumers, and of firms producing a homogenous good of variable quality. At every period firms can choose to provide high or low effort. If they choose low effort they produce a low quality good which is costless, while if they put in high effort they produce a high quality good with marginal costs c. Firms are also subject to an exogenous "bad" productivity shock that happens with probability 1 - , in which case the good becomes of low quality. Our uncertainty variable 1 - captures in a simple way the difficulty in using observed quality as a signal of effort. It represents production uncertainty faced by firms, such as infrastructural problems, regulation uncertainty, or the prevalence of an unreliable labor force, and as in Kremer (1993) we say that countries (sectors) with a higher "reliability parameter" are more developed. Quality is unobservable to consumers until after they have bought the good, and consumers cannot observe if low quality is due to a bad productivity shock, or due to the firm's decision to produce low quality. Shocks are persistent, and when a firm has faced a bad productivity shock it stops producing high quality forever: hence, consumers face both moral hazard and adverse selection problems. To avoid repetitions, we will call firms that have always produced good quality in the past "good" firms, and firms that have produced bad quality at least once "bad" firms. At each period, consumers and firms meet randomly in the market. Consumers need to buy one unit of the good each period, and derive utility U (p) = U - p from high quality, and utility 0 - p from low quality. The maximum price consumers are willing to pay for high quality is thus p = U, while consumers are not willing to spend money on low quality. However, consumers who decide not to buy from a firm have to wait until next period to randomly trade with another partner. Notice that consumers do not know why a given firm did not produce good quality in the past. Thus, their best reaction is to stop buying from any firm delivering low quality, since - given a high effort equilibrium - firms that produce low quality today will always produce low quality. 4 Nt New Firms Consumers Observe Price and Quality Shock Consumers Buying Prices and Choose Enter. Decisions. Realization. Get Informed. Decisions. Firms. Figure 1: Timing of the stage game. Figure 1 shows the timing of the stage game. In each period there are Nt new firms entering the market and investing a sunk cost of T units in building capacity. The sunk cost allows them to produce up to one unit of output per period, and the number Nt of firms that enter the market in each period is such that entrants face zero expected profits.5 Next, all firms choose prices and quality simultaneously, after which shocks are realized and firms produce either high or low quality. Consumers then observe prices, go randomly to a firm posting a price at which they wish to buy the good, get informed about the firm's history, and decide whether to buy the good or not. Finally, after all transactions have occurred, each firm faces an exogenous probability of closure 1-. An equilibrium is therefore a sequence of prices and quality choices, along with consumers buying decisions, such that consumers maximize utility given the firms strategies, new firms decide whether to enter or not, and all firms in the market choose prices and quality to maximize profits given the consumers strategies (see the Appendix for a formal discussion). In what follows, however, we restrict attention to perfect Bayesian equilibria in stationary strategies that maximize consumers' payoff. In a stationary equilibrium, there are N/(1 - ) firms in the market, and N/(1 - ) good firms in each period. We then consider two institutions that can induce firms to produce high quality: reputation and legal enforcement. Reputation works through the repeated interac- tion of consumers and firms in the market, while legal enforcement works through the reimbursement of consumers who go to court after having experienced bad qual- ity. We denote by j the probability that firm j has to reimburse consumers if it delivers low quality, and by = jdj the average level of legal efficiency in society.6 Similarly, we denote by qi the probability that consumer i is informed 5We depart therefore from conventional models of reputation in two ways (see, among others, Klein and Leffler, 1982, Shapiro, 1983, Allen, 1984, H¨orner, 2002, and Kranton, 2003): first, as in H¨orner (2002), we introduce shocks. Second, as we aim to analyze the interaction of institutions in a competitive setting where firms do not internalize the consequences of their own bribing on institutions, we introduce sunk rather than fixed costs, so that with free entry a single firm cannot capture all the market share. 6Thus, we assume that ex-post quality is verifiable. 5 about the firm she is trading with, and refer to the average level of information in society Q = qidi as connectedness, because consumers need to "connect" to other consumers to be informed about "bad" firms. In this section we assume qi and j to be exogenous and equal across people and firms, so that qi = Q, j = . Finally, we consider a situation where U is sufficiently high so that consumers always prefer firms producing high quality. The expected utility of consumer i in each period is thus equal to: 1 - Ui = (1 - )(1 (1) 1 - U - (1 - (1 - ))p - 1 - - qi)(1 - )p where p is the price of the good, and 1- 1- is the share of "good" firms in the economy. Good firms have a bad shock in the current period with probability 1 - , in which case the consumer gets utility 0-(1-)p as she will be reimbursed with probability . On the other hand, if the consumer meets a bad firm and she is informed, she will not buy from that firm and gets utility U = 0, while if she is uninformed she buys at price p and is reimbursed with probability (1 - ). Notice that consumers' welfare is maximized when p is minimized, and that price competition between firms guarantees that the equilibrium price is the lowest stationary price that is compatible with high quality. Hence, given a price p, the expected payoff of a good firm j from always putting in high effort is: VjH = (1 - j(1 - ))pxH - c · xH + H B (2) R Vj + (1 - )Vj where 1/R is the discount rate and VjB = R(1 - j)pxL/(R - ) represents the discounted profits of a bad firm facing judicial efficiency j.7 On the other hand, if firm j shirks, it faces an expected payoff equal to VjL = (1 - j)pxH + (/R)VjB. In order to sustain high quality we must have that VjH VjL. Thus, as qi = Q and j = are equal across agents and firms, high quality equilibria are sustainable only if: p(, , Q) RxH c (3) (1 - )(xH - xL) + RxH We call inequality (3) the No Milking Condition (see Shapiro, 1983, and Allen, 1984), and the lowest price that satisfies condition (3) the No Milking Price pNM(, , Q). The no milking condition shows that sustaining high effort requires a "carrot and 7 Equation (2) implicitly assumes that consumers stop buying from a bad firms independent from winning or losing in court, as in a high quality equilibrium firms deliver bad quality only if they have been hit by a bad shock. 6 stick" strategy: in order to be able to reward firms for high effort, price must be above marginal costs (the carrot); on the other hand, consumers must also punish shirking firms by boycotting them (the stick). Notice that the no milking price pNM has two components: the marginal cost component c/, and the markup component (represented by the first fraction in (3)). The markup is required to sustain high quality when legal enforcement is less than perfect, and decreases with the efficiency of either institution: Proposition 1 pNM is the lowest stationary price that can be achieved as the out- come of a Perfect Bayesian equilibrium where no firm shirks. Moreover, pNM/Q < 0, and pNM/ < 0. In the Appendix we describe strategies and beliefs that support the high quality equilibrium with firms pricing at pNM.8 Notice that pNM is the most plausible sta- tionary outcome, as competition between firms and free entry ensure that stationary equilibria with prices higher than pNM are not chosen.9 Moreover, if firms were to price lower than pNM consumers would know that they get low quality for sure, and would not buy. Thus, any firm that prices lower than pNM will get no market share, so that there are no stationary separating equilibria in our model. Observe that in the reputational equilibrium (i.e. < 1) firms overinvest in capacity. This is because firms need to price above marginal costs to have the incentives to produce high quality. But because of the free entry of firms, all firms' profits translate into excess capacity (xH < 1). Notice, also, that at high levels of institutional efficiency (, Q), firms' participation constraint can be violated, as firms cannot recover their sunk costs even under full capacity production xH = 1. Therefore, if (, Q) are too high, consumers and firms need to coordinate on a price above pNM. However, we do not discuss this case (see Esfahani, 1991). Our repeat purchase mechanism model is set up to capture in a simple way the im- pact of legal enforcement and consumer connectedness on the incentive compatible price. A more realistic model would have consumers choosing whether to switch from a firm or not based on their experience and the information available from other consumers, and would have firms choosing to exit from the market when mar- ket share becomes too small. It is also true that there are other ways to solve the asymmetric information problem. For instance, firms can decide to form associations to accredit standards (thus signalling their type), or the government may publicize 8As shocks are persistent, the equilibrium is also renegotiation proof. 9Sustaining higher prices would require collusive strategies between firms, which seems unlikely with free entry. 7 information on court cases against firms to inform consumers (see MacLeod, 2006). However, as our focus is to study the interdependence between alternative enforce- ment institutions, we abstract from these issues. 3 Reputation and Legal enforcement as Endogenous In- stitutions We now let consumers invest in their own connectedness to increase the probability with which they are informed about the firm they are trading with. Similarly, we let firms choose how much to bribe court officials to decrease the probability of having to reimburse consumers. To be sure, a more general model would conceivably allow for consumers and firms influencing both variables Q and . Nevertheless, what we want to capture here is the fact that consumers have a comparative advantage in investing in connectedness, while firms have an advantage in bribing. We begin by describing the consumers' maximization problem. Let mc(i) denote consumer i's investment in her own connectedness. Then individual connectedness is equal to q(mc(i)), where q > 0, q < 0, and to exclude corner solutions we assume that q satisfies the Inada conditions q (0) = , q () = 0.10 In deciding how much to invest, consumers take the price pNM, average connectedness Q, and judicial efficiency as given, so that for constant values of , Q, the consumers' maximization problem is: 1 max (4) {mc,t} Rt U(, , q(mc,t), pNM) - mc,t t=0 t=0 Notice that each consumer faces the same maximization problem (4). Thus, ex post qi = Q, and we can use the first order conditions of the maximization problem to characterize the average level of connectedness Q: Proposition 2 For each (, ) there exists a unique level of connectedness Q re- sulting from the consumers' maximization problem (4). Moreover, Q/ < 0, and Q/ < 0. Intuitively, consumers invest in connectedness to be informed about bad firms in 10In a previous version we considered the case where qi also depends on average investment Mc = 1 0 mc(i)di, capturing the idea that individual connectedness increases proportionally more if other consumers also invest in their own connectedness. This adds significant algebraic burden while conveying similar results. 8 the market, and as increases, the share of bad firms decreases. Moreoever, the gain of an extra unit of information per firm is decreasing with the price (which also decreases with ), while the marginal cost is constant. Thus, as increases consumers invest less in connectedness, both because of the direct effect on the share of bad firms, and of the indirect effect on pNM. The same logic holds for judicial efficiency , which captures the net benefits of going to court. We now turn to firm behaviour. Firms can decrease the probability of having to reimburse consumers by bribing. We presume that bribing has decreasing returns, so that (mf) < 0, (mf) > 0, and we also assume that (mf) satisfies the Inada conditions. For constant values of , , Q, the maximization problem of a firm that delivered low quality is: max(1 - (mf))pNMx - mfx (5) mf Notice that firms bribe the court after a case has been brought against them, so that x = xL,H. Using the first order condition of (5), we can characterize judicial efficiency : Proposition 3 Let (mf) satisfy /| | > 1/c. Then for each (, Q) there exists a unique level of judicial efficiency resulting from the firms' maximization problem (5). Moreover, / > 0, and /Q > 0. Intuitively, in the high effort equilibrium an increase in lowers pNM via marginal costs. Therefore, when reliability is high, bad firms have lower incentives to bribe because gains are lower. Bribing, however, also has a general equilibrium effect because it increases the equilibrium price pNM, and hence firms' profits: it is to rule out this perverse effect through that Proposition 3 requires the condition on (mf).11 We conclude by looking at the overall equilibrium institutional mix. Figure 2 shows the consumers' and firms' reaction functions QC(), QF (): the reaction functions are monotonic and opposite in slope, hence the equilibrium is unique. Moreover, an increase in reliability shifts both consumers' and firms' reaction functions down- wards, so that connectedness Q unambiguously decreases with reliability . In con- trast, the effect of changes in reliability on judicial efficiency remains a priori ambiguous, as whether judicial efficiency increases or decreases with depends 11A similar effect also acts on consumers' investment decisions in connectedness q, but for con- sumers the indirect effect has the "right" sign. 9 Q QF (0) Q(0) QF (1 > 0) Q(1 > 0) QC(0) QC(1 > 0) 0 Figure 2: Equilibrium levels of connectedness and of judicial efficiency. on whether the firms' reaction curve QF moves downwards more or less than the consumers' reaction curve QC. At low levels of reliability, however, QF is more elastic than QC, so that bribing also decreases with : Proposition 4 1. Equilibrium connectedness always decreases with the reliability of the produc- tion process. 2. There exists a threshold such that judicial efficiency improves with reliability for < . Thus, in accordance with common wisdom, corruption in the judicial system is associated with a more widespread use of informal enforcement, but via the reliability channel. 4 Empirics In a context of intra-industry trade, firms are often consumers of other firms' prod- ucts, and a stated reason for the existence of firms associations is to get information about other firms. We use this interpretation to test Proposition 2 ­ i.e. that consumer's incentives to connect with one another (by becoming, in our case, mem- 10 bers of business associations) are affected by corruption and by uncertainty in the production climate. We test this implication at the firm level using data from the Investment Climate Assessments (ICA) of the World Bank, which survey more than 28,000 firms in 58 countries between 2000 and 2004, and provide information on firms' characteristics, firms' productivity, and the investment climate. Because not all questions are asked in each country, our final sample consists of 19,000 firms spread over 49 countries. The survey is not nationally representative, but is representative of specific sectors that are consistently surveyed across countries: in the sample, firms in the manufac- turing sector represent 72% of the sample, firms in services 21%, and the remaining 7% of the sample consists of firms operating in agriculture and construction. As a proxy for connectedness we use the variable MEMBER, which indicates whether a firm is a member of a business association. Memberships to business associations are rarely compulsory, and there are usually few binding contractual obligations arising from membership. Therefore, business associations can be con- sidered informal institutions that "connect" their members by mitigating informa- tion asymmetries, providing a basis for solving disputes with trade partners, and easing the transmission of information. This role of business associations is strongly supported by a recent study of Pyle (2005). To be sure, a positive association between corruption, uncertainty, and mem- bership to business associations could potentially also stem from bad governance, leading in the long term to low quality services, and firms' increased incentives to lobby. Although with existing data we are not able to entirely exclude this hy- pothesis, we apply the following safeguards to avoid capturing such a relationship. First, we give the value one to the variable MEMBER only if the firm identifies the ability to resolve disputes, to provide information on domestic product markets, or to accredit standards, as important functions of the business association: these functions should capture the ability of a business association to spread information about other firms, and to coordinate punitive actions, and give less importance to lobbying intentions. Moreover, our analysis is at the firm level, so that the macro- economic relationship between corruption and lobbying should be captured by the country fixed effect. Finally, we construct an uncertainty index from variables that have (at least in the short run) little relationship with governance and lobbying. Specifically, we construct two uncertainty indexes using the following variables: Uic = 0.51 ELECic + 0.35 SKILLic + 0.52 T RANSPic (6) 11 Uic = 0.34 ELECic + 0.29 SKILLic + 0.36 T RANSPic +0.35 CRIMEic + 0.29 MACROic where ELEC, SKILL, TRANSP, CRIME, and MACRO are dummies equal to one if a firm identifies respectively electricity losses, the lack of skills in the workforce, transportation, crime, and macro-economic instability as important business con- straints, and the weights are derived using the method of principal components.12 Note that the second index "expands" the first with with socio-political variables, but is more likely to pick up a relationship between bad governance and lobbying. Finally, as a measure of corruption Cic we use a dummy variable equal to one if firm i identifies corruption as a relevant business constraint. We then estimate the following Probit model: P [MEMBERic = 1] = P [0 + Uic U + Cic C + Yic Y + ic > 0] (7) where i denotes firms and c countries, Uic is the uncertainty index, Cic the corrup- tion index, Yic represents firm characteristics, and the theoretical model predicts that both U and C are positive (see Proposition 2). Notice that the model also predicts a reverse causal relationship between average consumers' connectedness and corruption. However, as we estimate the probability of a single firm being a mem- ber of a business association, it can be shown that the estimates are consistent. Note, also, that we diverge from the theoretical model in two ways. First, we allow different firms to face different levels of corruption. This is because in the model all firms are producing the same product, but in the data firms are substantively different from each other and in principle could "suffer" differently from corruption. Second, strictly speaking the model requires Uic to be the level of uncertainty faced by the business partners of firm i: unfortunately, this information is not available. As firms with similar activities and uncertainty levels tend to trade more often with one another, we assume that the level of uncertainty of each firm is a good proxy of the uncertainty level faced by its business partners. Table 2 presents the results of the Probit estimation: Column 2 shows the estima- tion for all countries, while Columns 3­6 show the estimation by geographic region. In all regressions we have included country and sector dummies, have clustered the errors by country, and have weighted each observation by the total number of ob- 12See Theil (1971) and Alesina and Perotti (1996) for a description of the method of principal components and macro-economic applications. 12 servations in the country. In accordance with the theory, higher levels of corruption and more uncertain production climates are both associated with a higher probabil- ity of firms being members of business associations. The relationship continues to hold when we disaggregate the analysis by region: with the exception of corruption in Sub-Saharan Africa, higher levels of corruption and uncertainty remain positively and significantly related to membership to business associations in all five regional estimations. Table 3 presents the results with the expanded uncertainty index: the results remain basically the same for uncertainty, while the corruption variable be- comes insignificant in Eastern Europe and Central Asia. Lower significance is likely to stem from the new variables added in the uncertainty index, which correlate more with corruption. 5 Conclusions This paper studies the interaction between formal and informal institutions that prevent opportunistic behaviour. Our results are only partly consistent with the common belief that formal and informal institutions substitute each other: we find that when legal enforcement works poorly, consumers invest more in connecting with other consumers to enhance contract enforcement via the reputation mechanism; on the other hand, however, better informal enforcement improves legal enforcement because it reduces firms' incentives to bribe. In fact, in equilibrium we do observe a positive relationship between corruption in the legal system and the use of informal enforcement, but only via increased uncertainty in the production process. Our anal- ysis suggests therefore a physical explanation (i.e., uncertainty) for why developing countries often fail to have well-performing judicial systems. Empirical evidence appears to support the predictions of the model: using firm-level data, we find that firms are more likely to become members of business associations when they see corruption and uncertainty in the production process as severe business contraints. The paper, however, abstains from studying how changes in the institutional mix affect consumers' welfare. When the reliability of the production process in- creases, consumers' welfare is affected both directly by better performing firms, but also indirectly through the institutional mix. As both consumers and firms do not consider the impact of their investment decisions on equilibrium prices, the effect of a change in the institutional mix on consumers' welfare remains a priori ambiguous, and is a subject for future research. 13 References Acemoglu, D., S.Johnson and J.A. Robinson (2005), "Institutions as the Fundamental Cause of Long-Run Growth," forthcoming, Handbook of Economic Growth. Alesina, A., and R. Perotti (1996), "Income Distribution, Political Instability, and Invest- ment," European Economic Review, 40, 1203-28. Allen, F. (1984), "Reputation and Product Quality," The RAND Journal of Economics, 15(3), 311-327. Battigalli, P., and G. Maggi (2004), "Costly Contracting in a Long­Term Relationship," mimeo, Princeton University. Besley, T., S. Coate, and G. Loury (1994), "Rotating Savings and Credit Associations, Credit Markets and Efficiency," The Review of Economic Studies, 61(4), 701-719. Dixit, Avinash (2003), "On modes of economic governance," Econometrica, 71(2), 449-481. Economist Intelligence Unit (2005), "Regulatory Risk: Trends and Strategies for the CRO." Ellison, G. (1994), "Cooperation in the Repeated Prisoners Dilemma with Anonymous Ran- dom Matching," The Review of Economic Studies, 61, 567-588. Esfahani, H. (1991), "Reputation and Uncertainty: Towards an Explanation of Quality problems in competitive LDC markets," The Journal of Development Economics, 35, 1-32. Greif, A. (1993), "Contract Enforceability and Economic Institutions in Early Trade: The Maghribi Traders Coalition," The American Economic Review, 525-548. Greif, A., Milgrom P., and B. Weingast (1994), "Coordination, Commitment and Enforce- ment: The Case of the Merchant Guild," Journal of Political Economy. Greif, A.(2004), "Commitment, Coercion and Markets: The Nature and Dynamics of In- stitutions supporting exchange," In C.Menary and M.M.Shirley (Eds), Handbook for New Institutional Economics, Norwell, MA: Kluwer Academic Publishers. Hallward-Driemeier, M., and D. Stewart (2004), "How Do Investment Climate Conditions Vary Across Countries, Regions and Types of Firms?," background paper prepared for the World Development Report, The World Bank, Washington, DC. H¨orner, J. (2002), "Reputation and Competition," The American Economic Review, 92(3), 644-663. Investment Climate Assessments (2000-2004), The World Bank, http://rru.world-bank.org/ InvestmentClimate/. Klein, B., and K. Leffler (1981), "The Role of Market Forces in Assuring Contractual Per- formance," The Journal of Political Economy, 89(4), 615-641. Knack, S., and P. Keefer (1997), "Does Social Capital Have an Economic Payoff? A Cross- Country Investigation," The Quarterly Journal of Economics, 112(4), 1251-1288. 14 Kandori, M. (1992), "Social Norms and Community Enforcement," The Review of Eco- nomic Studies, 59, 63-80. Kranton, R. (1996), "Reciprocal Exchange: A Self-Sustaining System," The American Eco- nomic Review, 86(4), 830-51. Kranton, R. (2003), "Competition and the Incentive to Produce High Quality," Economica, 70(3), 385-404. Kremer, M. (1993), "The O-Ring Theory of Economic Development," The Quarterly Jour- nal of Economics, 108(3), 551-575. Mookherjee, D. (1999), "Contractual Constraints on Firm Performance in Developing Coun- tries," mimeo, Boston University. MacLeod, W.B. (2006), "Reputations,Relationships and the Enforcement of Incomplete Con- tracts," mimeo, Columbia University. Putnam, R. (1993), Making Democracy Work: Civic Traditions in Modern Italy, Princeton University Press, Princeton, NJ. Pyle, William (2005), "Contractual Disputes and The Channels for Interfirm Communica- tion," The Journal of Law, Economics, & Orgnization, 21(2), 547-75. Shapiro, K. (1983), "Premiums for High Quality Products as Returns to Reputations," The Quarterly Journal of Economics, 98(4), 659-680. Theil, H. (1971), Principle of Econometrics, Wiley, New York. World Bank (2005), World Development Report 2005: A Better Investment Climate For Everyone, Washington, DC. 15 Appendix Proof of Proposition 1 Without loss of generality we only consider the case where Q = 1. The stage game is as follows. At period t there are Nt firms that simultaneously post prices P t = (ptj)j [0,Nt] , and decide whether to produce the high or the low quality good, so that gj = (H, L) for good t firms, and gj = L for bad firms. Consumers i [0, 1] observe the vector of prices P t in the t market, go to a firm j, and decide whether to buy (ati = Bj) or not to buy (ati = NBj). We assume that once consumers have bought the good they can observe the quality perfectly. Thus the stage payoff to consumers at the end of the period is equal to U -ptj if they bought the good and the quality is good, 0 if they did not buy any good, while if quality is bad they get -ptj. On the other hand, the payoff to firm j is equal to (ptj - c) · xtj if it produces the high quality good and xtj consumers bought it, and to ptj · xtj if it produces the bad quality good. Payoffs to firms and consumers in the game as a whole correspond to the discounted sum of payoffs in each period. The game is repeated over an infinite horizon, so that a history ht at period t is a sequence of quality and price vectors (G0, P 0); . . . ; (Gt -1, P t-1), where Gt = (gj, xtj)j t [0,Nt] , and of consumer actions (a0i ); . . . ; (ati-1). Finally, consumers' information sets at time t are defined by all price combinations t = (pj [0, p¯])j [0,Nt] for each possible history ht, which for simplicity we refer to a consumer's information set as (P, ht). The Markov strategies and beliefs that achieve pNM as the outcome of a perfect Bayesian equilibrium are the following: Firms' Strategy. New firms enter as long as expected profits net of sunk costs are positive. Bad firms always produce bad quality and price at pNM. Good firms have the following strategy: 1. If ptj pNM, put in high effort. 2. If ptj < pNM put in low effort. 3. Set ptj = pNM, regardless of history. Consumers' Strategy 1. Do not buy if max ptj < pNM. 2. Match randomly among firms posting a price equal to min ptj|ptj pNM . 3. If a firm has produced bad quality in t - 1, do not buy. Consumers' Beliefs: 1. If ptj < pNM then the firm has produced bad quality with probability one. 2. If ptj pNM then the firm has put in high effort and the probability of getting good quality is , as long as the previous history did not have bad quality. 3. If a firm has produced bad quality in period t - 1 then it will always produce bad quality. It is easy to prove that this strategy profile represents a Nash equilibrium. Notice that firms never face a non-trivial information set, since, given the consumers' strategies, incomplete information about other firms' types does not influence payoffs. Hence, the only imperfect information comes from the simultaneous price game, and it is sufficient to look for subgame perfection in firms' strategies. Consider any subgame off the equilibrium path where prices 16 of some firms are lower or higher than pNM regardless of quality history, or where consumers do not follow their equilibrium strategies. Given consumers strategies in the continuation game, the best response is obviously to price at pNM and put in high effort as long as there is no bad shock. Moreover, if consumers buy at a price lower then pNM they believe that they will get bad quality, and this belief is consistent with firms strategies. Given the permanence of shocks and equilibrium strategies of firms, if a firm produces bad quality once the best response is never to buy from this firm again. Finally, assume that that there exist a stationary price p~ < pNM under which firms put high effort. The no milking condition (3) ensures that, given other firms strategies and consumer's strategies, a firm charging p~ would strictly prefer to cheat in every period. End of Proof. Proof of Proposition 2 Since q(mc) satisfies the Inada conditions the solution of the maximization problem lies in mc,t (0, ), and we can use the first order conditions of the maximization problem to characterize the optimal investment mc,t. Notice that utility is maximized when mc,t maximizes the per period utility. Notice, also, that each consumer faces the same first order conditions, and that in equilibrium Q qi, so that connectedness is characterized by the following condition: G (1 - )(1 (8) 1 - - ) · p(,,q(mc))q (mc) = 1 where p = pNM. By the implicit function theorem the following then holds: G p dq 2 d2q = (1 - )(1 - ) · + p < 0 (9) mc 1 - q dmc dm2c G dq (1 - ) p = < 0 dmc 1 - -p + (1 - ) G = (1 - ) dq 1 - 2 + 2 p < 0 -1 - dmc (1 - ) where 1 - 2 + 2 is minimized for = 1, so that G/ < 0. By the implicit function theorem we then have that mc/ = -G/Gmc < 0, and that mc/ = -G/Gmc < 0. Finally, the fact that G/mc < 0 ensures that there is a unique equilibrium. End of Proof. Proof of Proposition 3 The firms' first order conditions are equal to: - (mf) = 1/pNM. The aggregate firms' implicit function is therefore equal to: F (mf) + (1 - )Q + R = 0 (10) R c 17 and therefore, the partial derivatives are equal to: F (1 - ) = > 0 (11) Q Rc F (1 - )Q + R = > 0 Rc F (R - Q) | | = (mf) + (mf) > - > 0 mf Rc c where the last inequality holds for /| | > 1/c. Using the implicit function theorem we then have that /Q = - FQ/Fmf > 0, and that / = - F/Fmf > 0. End of Proof. Proof of Proposition 4 The partial derivative of the consumers' reaction function QC is as follows: QC 1 - (2 - ) 1 (12) = - (1 - )(1 - ) (1-) (1-)Q+R + |q | / (q )2 where pNM/Q = -pNM(1 - )/((1 - )Q + R). Rewriting judicial efficiency as F (QF (, ), ), notice that F (QF (, ), ) - = 0. Thus, using the implicit function theorem we have that QF / = -(F /)/(F /QF ), which implies that: QF = - (1 - )Q + R (13) (1 - ) The sign of d/d depends on whether |QF /| |QC/|. Hence, judicial efficiency increases if and only if: 1 1 - (2 - ) 1 > (14) (1 - ) (1 - )(1 - ) (1 - ) + C where C = (1 - )Q + R, and = |q | / (q )2. For 0 the inequality (14) is always satisfied, while for = 1 the inequality is never satisfied. End of Proof. 18 Table 1: Variable definitions. Variable Definition Average MEMBER Dummy variable with value one if the firm is a mem- 0.21 ber of a business association and an important func- tion of the business association the ability to resolve disputes, to provide information on domestic product markets, or to accredit standards. CORR Dummy variable with value one if the firm identifies 0.51 corruption as a relevant business constraint. ELEC Dummy variable with value one if the firm identifies 0.38 electricity shortages as a relevant business constraint. SKILL Dummy variable with value one if the firm identi- 0.39 fies the lack of skills in the workforce as a relevant business constraint. TRANSP Dummy variable with value one if the firm identifies 0.27 transportation as a relevant business constraint. CRIME Dummy variable with value one if the firm identifies 0.42 crime & vandalism as a relevant business constraint. MACRO Dummy variable with value one if the firm identi- 0.62 fies macro-economic instability as a relevant business constraint. LARGE Dummy variable with value one if the firm's size is 0.26 large or very large. EXP/IMP Dummy variable with value one if the firm exports 0.12 part of its output, or import at least 40% of its in- puts. FOREIGN Dummy variable with value one if the firm's largest 0.31 shareholder is foreigner. PUBLIC Dummy variable with value one if the firm is owned 0.08 by the state. 19 African North East Asia .11 .08 .47 .18 2 (.02) (.02) (.13) (.04) -.11 (.10) .46 only (.05) 1400 the is 3 South Asia 25. 13. 40. 05. (.10) (.05) (.04) (.04) -.26 .51 (.02) (.15) 2414 el. Algeria lev an . a 5% ndex.i try .08 7 yt Afric (.08) .09 .26 .23 05. (.04) (.08) (.07) (.10) 06. (.16) 2215 the ounc at Sub-Sahar yb t e op sia uncertain A Eur al Significan 09. 30 clustered (.05) 08. 42. 30. .05 (.03) (.06) (.06) (.06) .03 (.07) 8327 Centr eenb el. Eastern & exogenous ev lev a ha 10% with atinL .20 .07 .31 .31 6 meric (.08) (.04) (.03) (.08) .009 (.07) -.04 (.30) 3916 errors the A at Probit and t el ll A 11. 08. 37. 26. .03 (.03) (.02) (.04) (.04) (.04) 05. (.06) 49 19173 Countries Significan dummies, Firm-lev 2: thesis. sectorial able T MEMBER and paren orter in try yt wnershipo are ariable: coun V firm wnershipo ations tries orter/imp errors Dep. Corruption Uncertain Large Exp oreignF Public Observ Coun include Standard . regressions try All coun 20 African North ast E Asia .05 .10 .46 .17 2 only (.004) (.02) (.13) (.04) -.11 (.10) .45 (.06) 1400 the is 3 South Asia 21. .11 39. 06. (.08) (.01) (.03) (.04) -.25 .52 (.02) (.16) 2414 el. Algeria lev an . a 5% try index. .02 7 yt Afric (.07) 14. .24 .23 05. (.05) (.08) (.07) (.10) 06. (.15) the 2215 ounc at Sub-Sahar yb t e ncertainu op sia A Eur al Significan 06. clustered (.05) .08 43. 29. .05 (.03) (.06) (.06) (.06) .03 (.07) 30 8327 Centr eenb el. expanded Eastern & ev lev ha a 10% with atinL .16 .09 .31 .31 .02 6 meric (.07) (.05) (.03) (.08) (.07) -.01 (.30) 3916 errors the A robit at P and t le ll A 08. .10 37. 26. .03 (.03) (.02) (.04) (.04) (.04) 06. (.06) 49 19173 Countries Significan dummies, Firm-lev 3: thesis. sectorial able T MEMBER and paren orter (expanded) in try yt wnershipo are ariable: coun V firm wnershipo ations tries orter/imp errors Dep. Corruption Uncertain Large Exp oreignF Public Observ Coun include Standard . regressions try All coun 21