The impact of liberalization on bureaucratic corruption

Journal of Economic Behavior & Organization 72 (2009) 214–224

Contents lists available at ScienceDirect

Journal of Economic Behavior & Organization journal homepage: www.elsevier.com/locate/jebo

The impact of liberalization on bureaucratic corruption Soham Baksi a,∗ , Pinaki Bose b,1 , Manish Pandey a,2 a b

Department of Economics, University of Winnipeg, 515 Portage Avenue, Winnipeg, Canada R3B2E9 Department of Economics, University of Memphis, Memphis, TN 38152, USA

a r t i c l e

i n f o

Article history: Received 23 August 2007 Received in revised form 1 May 2009 Accepted 4 May 2009 Available online 18 May 2009 JEL classification: D73 O10 O19

a b s t r a c t Liberalization increases the number of goods available for consumption within a country. Since bureaucrats value variety, this raises the marginal utility of accepting a bribe. This “benefit effect” is counteracted by an increasing “cost effect” from corruption deterrence activities that arise due to greater international pressure to curb corruption. The interaction of these two effects can lead to a non-monotonic relation between liberalization and corruption. Moreover, pre-commitment to deterrence activities is shown to be more effective in controlling corruption. Empirical evidence supports the existence of a non-monotonic relation between economic openness and corruption among developing countries. © 2009 Elsevier B.V. All rights reserved.

Keywords: Corruption Bribery Bureaucracy Monitoring Liberalization

1. Introduction In an increasingly integrating world, the nexus between globalization and corruption is an important issue, especially for policy purposes.3 The extant literature on corruption has generally found a negative relation between corruption on one hand and trade, investment and growth on the other (see Rose-Ackerman, 1975, 1978; Mauro, 1995; and the survey by Bardhan, 1997).4 The explanation offered is as follows: by introducing greater foreign competition, trade liberalization reduces monopolistic rents enjoyed by firms and decreases their ability to pay a bribe, thereby reducing bureaucratic

∗ Corresponding author. Tel.: +1 204 2582945; fax: +1 204 7744134. E-mail addresses: [email protected] (S. Baksi), [email protected] (P. Bose), [email protected] (M. Pandey). 1 Tel.: +1 901 6785528. 2 Tel.: +1 204 7869289. 3 Corruption – the misuse of public office for private gains (Bardhan, 1997) – can be “administrative” (misuse by law enforcers) or “grand” (misuse by lawmakers). This paper deals with the former type of corruption only. 4 There is less unanimity on the effect of corruption on static efficiency. While Shleifer and Vishny (1993), Mauro (1998) and Ahlin and Bose (2007) find corruption to be distortionary, Leff (1964) has argued that in a second best world with pre-existing policy induced distortions, corruption may actually improve efficiency. The ability of the lowest cost firm to pay the highest bribe can ensure an efficient outcome in a bargain between a corrupt official and prospective firms, where the former tries to sell a permit to the latter. This argument, of course, presupposes that the lowest cost firm is able to meet the objectives of the permit-program as well as the other firms. On the institutional economics of corruption, see Lambsdorff (2007). 0167-2681/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jebo.2009.05.002

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

215

corruption.5 One exception to this general finding is Ades and Di Tella (1999), who show that if corruptible officials are paid an efficiency wage to induce honest behaviour, the effect of increasing competition on corruption is ambiguous. By reducing profits of oligopolistic firms, competition reduces the efficiency wage as it becomes less attractive to induce honesty, but, at the same time, any level of wage deters more corruption as the gains to corrupt officials fall with competition. As a result of these two contradictory effects, equilibrium corruption can either rise or fall with competition. The empirical evidence on how openness affects corruption across countries has been mixed. In contrast to the ambiguous nature of their theoretical predictions, Ades and Di Tella’s (1999) empirical results showed that foreign competition (as proxied by import to GDP ratio and other variables) negatively affects corruption. Similarly, Wei (2000a,b) and Gatti (2004) found an inverse relation between openness and corruption. Treisman (2000), however, reported a “surprisingly small” negative relation between corruption and openness to trade (measured by import to GDP ratio).6 The ambiguity evident in the literature, notably in Ades and Di Tella (1999) and Treisman (2000), seems to suggest that the relation between openness and corruption is more complex and may even be non-monotonic in many countries due to the interplay of opposing factors that are inherent in globalization.7 A positive relation between openness and corruption is borne out by the initial experience of the transitional economies of Eastern Europe and erstwhile USSR, “where essential steps to privatize the economy and rewrite the rules of commerce after the demise of socialism were often accompanied by widespread corruption” (Transparency International, 2005, p. 271). However, there is recent (1999–2002) evidence to support the view that “the prevalence and costs of some types of corruption are becoming more moderate in many countries in the region” (Transparency International, 2005). The purpose of the present paper is threefold. First, we explore a different (from those explored in the existing literature) channel through which economic openness can induce corruption among public officials. Second, we provide a possible explanation as to why the relation between economic openness and corruption may be non-monotonic in some countries. Lastly, we provide empirical evidence which indicates the existence of such a relationship in some countries. In this sense, our empirical analysis moves away from the existing literature which is preoccupied with estimating the linear relationship between openness and corruption. The different channel, referred to above, is as follows: liberalization typically increases imports, and imports introduce new goods and services to consumers of liberalized economies.8 This effect is likely to be stronger in developing countries, which lack the ability to produce many goods and depend on imports to make them available. Availability of more products increases the marginal utility of consumers’ income, given that they have a taste for variety. In particular, this will also increase the corruptible officials’ marginal utility of bribe income and increase their incentive to be corrupt. Thus liberalization, by increasing the allure of consumerism, strengthens the propensity of officials to be corrupt (at least initially) in our model.9 But there is a second, and countervailing, effect that results from greater openness. A commonly observed attribute of globalization is that it engenders additional pressures on developing countries to reduce corruption and improve governance. For one, greater openness makes domestic corruption more “internationally visible.” Second, international bodies have a greater ability and incentive to engage and pressure more open countries to crack down on corruption.10 For instance, according to Williams and Beare (1999, p. 142), “both the IMF and World Bank have recently introduced reforms to their lending practices making the provision of funds conditional upon the successful implementation of a variety of macroeconomic and anti-corruption reforms.” Similar international anti-corruption initiatives have been launched by the United Nations (UN Convention against Corruption, 2005), African Union (Convention on Preventing and Combating Corruption, 2003), and the Organization of American States (Inter-American Convention against Corruption, 1996).

5 The causality between openness and grand corruption can run in the other direction as well. Corrupt lawmakers are more likely to create regulations involving licenses, permits, quotas, and tariffs so as to facilitate their rent-seeking activities. Such policies can be inimical to trade and business opportunities. 6 In contrast, Williams and Beare (1999, p. 116) have noted “a conviction that corruption has increased to epidemic levels, and that globalization has provided much of the impetus and opportunity for this growth.” 7 The possibility of a non-monotonic relationship between economic growth and corruption has been suggested by Bardhan (1997). While reporting a negative relationship in general for most countries, Bardhan (p. 1329) observed that, “in some countries with the process of modernization and growth corruption may have got worse for some time before getting better.” This, according to him, can arise due to counteracting forces generated by the growth process. On one hand, an expansion of the economy provides public officials with “more opportunities for making money from their decisions” (Bardhan, 1997). On the other hand, a prospering economy can afford to pay its civil servants better (thus reducing their motivation to be corrupt), and may be more likely to install institutions that check corruption. Bardhan (1997), however, does not formally model these forces or their interaction. 8 This is in the spirit of international trade models with product differentiation (such as Krugman, 1980; Lawrence and Spiller, 1983) and Romer, 1994. Klenow and Rodriguez-Clare (1997) estimate that a 1% lower tariff was accompanied by a 0.5% increase in import variety for Costa Rica during 1986–1992. Also, see Mitra (2005). 9 In India, there is widespread perception that the post-liberalization consumerism, spurred on by influx of imports, has increased not only official greed and corruption, but also other illicit and extortionary money making ventures (such as dowry demands). This is reflected in the remarks of the head of the Central Vigilance Commission in a talk delivered at BASF Mumbai in 2002 (see http://unpan1.un.org/intradoc/groups/ public/documents/APCITY/UNPAN019875.pdf): “This brings us to another important social root for corruption that is probably getting more accentuated in recent times. This is the spreading cult of consumerism. . . Consumerism and desire for an ostentatious lifestyle tempt many to make money by hook or crook. Corruption is the result.” 10 See, for examples, the following media articles: “Africa tackles graft, with billions in aid in play” (The New York Times, July 6, 2005), and “EU takes China to task over product safety” (The Globe and Mail, July 23, 2007). International aid agencies and NGOs are often active in pressuring governments of developing countries to crack down on corruption. For example, USAID has developed a wide range of anti-corruption programs. Transparency International and U4 provide resources and training in regulating corruption, and also provide information that exposes corrupt deals, thereby raising domestic and international awareness and pressure for cleaner governance in these countries.

216

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

Our paper models a hierarchical public administrative structure consisting of many corruptible inspectors and an honest regulator (an anti-corruption agency) who undertakes deterrence activities to reduce corruption among the inspectors. We find that an increase in the number of consumption goods increases the marginal utility of accepting a bribe, which, in the absence of deterrence activities, leads to greater corruption among the inspectors (Section 2). In the presence of monitoring by a regulator, however, corruption may not increase monotonically, especially if greater openness also results in greater pressure on the regulator to tackle corruption. This is shown in Section 3 of our paper, where we introduce the regulator. Our paper demonstrates that while greater openness initially leads to increased bureaucratic corruption, the accompanying increases in corruption-reducing activities can cause corruption to fall after some threshold of liberalization has been reached. Thus, our theoretical analysis supports the existence of an inverted-U curve in the relationship between corruption and economic liberalization. The role of pre-commitment by the regulator to her monitoring activity is also analyzed in Section 3. By credibly pre-committing to a monitoring frequency, the regulator is not only able to detect violators (which she is able to do even if she cannot credibly pre-commit) but also able to influence the optimizing behaviour of the corruptible inspectors. Hence, we find corruption to be less under pre-commitment. Section 4 deals with the empirical analysis, where we find evidence of a non-monotonic relation between liberalization and corruption in lower income countries. The last section concludes. 2. Corruption without deterrence Consider an economy where many identical consumers derive utility by consuming n goods. Following Spence (1976) and Dixit and Stiglitz (1977), we assume that consumers have an innate preference for variety, and that each individual’s

n



 1/

utility is given by the CES utility function U = , where xi is the representative individual’s consumption of good x i=1 i i ∈ [1, n], and  ∈ (0, 1) is a constant. Suppose the government imposes a regulation on F firms in this economy, and c denotes each firm’s cost of implementing this regulation.11 For example, the regulation could be one that requires each of the F firms to install a new, more environmentally friendly, technology, which costs c. Further suppose there are M (with M ≥ F) government inspectors to enforce the regulation. For analytical convenience, we assume that one inspector inspects and certifies only one firm.12 The inspectors are corruptible, and the firms are willing to offer a bribe of an amount B to the inspectors for false certification. The maximum amount of bribe a firm will be willing to pay is c, the cost it can save by such bribing.13,14 The inspectors will accept the bribe if utility from it exceeds their disutility (psychic cost or guilt) from being corrupt. The net utility of a corrupt inspector is then



U− =

n 

1/

 xi

− ,

(1)

i=1

where  is the psychic cost associated with being corrupt. The inspectors are heterogeneous in this cost, and  is distributed ¯ For an inspector who does not take a bribe,  = 0. uniformly over the interval [0, ]. We assume that the inspectors earn zero salary income (so that their only income comes from accepting the bribe), and that the prices of all goods are equal to p, normalized to 1. For our purpose, these assumptions simplify the exposition without reducing the robustness of our results.15 Then the indirect utility of a corrupt inspector will be V (n, B, ) = n B − ,

(2)

where  (1 − )/ > 0. Since by not accepting the bribe an inspector will earn zero utility, the inspector who is indifferent between accepting and rejecting the bribe will have a threshold psychic cost of  ∗ = n B.

(3)

˜ Let M(B) ≡ M ∗ /¯ = n BM/¯ denote the number of corrupt inspectors when the bribe amount is B. Keeping in mind that ¯ With one inspector per firm, ˜ ˜ max ≡ M(c) = n cM/. the bribe cannot exceed the cost of the green technology, c, we denote M ∗ ˜ let the equilibrium number of corrupt officials (also bribe-paying firms) be denoted by M . Then, it is easy to see that:

11 Although this not strictly necessary for our results to hold qualitatively, assume for expositional ease that these F domestic firms do not produce any of the n goods that are consumed in the economy. The goods produced by these regulated firms could be either exported or are capital goods. Furthermore, c is not so high that one or more of the F firms have to exit the market when the regulation is imposed. 12 As in Ades and Di Tella (1999), and Acemoglu and Verdier (2000). 13 The scenario that we analyze is one where firms offer bribes to attract officials. Each official has the authority to certify at most one firm. A firm cannot be worse off by choosing bribery over honesty, as the maximum bribe it pays equals the cost of complying with the regulation. 14 The setting of our model is similar to what Sapru (1998, p. 172) notes for India, “The practice of large scale corruption and other forms of bribery among officials has stalled the implementation of pollution control laws to a significant extent. Industry owners commonly perceive that public servants can be bought by monetary incentives. Therefore, industrial polluters reason that they have recourse to cheaper ways than to comply with regulations that may entail significant cost.” 15 In case the inspectors earned positive salary, this would only require an appropriate change of base. Moreover, by holding prices constant, we are abstracting away from the well-known effect of liberalization on firms’ rents and corruption, in order to focus on a different effect. In case liberalization (apart from increasing the number of goods available for consumption) caused prices to decline through competition, this would further increase a corrupt inspector’s indirect utility.

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

217

˜ max < F, then B* = c, and M ˜∗=M ˜ max . This is an interior equilibrium where only some of the F firms pay the maximal (i) If M ¯ i.e. officials with higher psychic cost of corruption remain bribe to an equal number of inspectors.  * is strictly less than , honest. ¯  M ≤ c and M ˜ ∗ = F. In equilibrium the bribe amount is less than (or equal to) c, and all firms ˜ max , then B∗ = F /n (ii) If F ≤ M pay the bribe. If there are more inspectors than firms, not all inspectors will be corrupt. An increase in the number of goods increases the utility from corruption, which induces inspectors with higher psychic ˜ ∗ initially increases with n till the number of corrupt inspectors equals F, the total number of costs to accept bribes. Thus M ˜ firms. At this point, M(c) = F, using which we derive the corresponding number of goods to be at the threshold value of



F ¯ cM

1/

≡ n1 .

(4)

If the number of available goods is less than n1 , we get case (i). Otherwise, case (ii) holds. Proposition 1 follows. Proposition 1. If, due to opening up of trade, the number of goods available for consumption in the economy (n) increases, this ˜ ∗ , keeping the bribe amount unchanged at c (case (i)). When n = n1 , all F initially increases the number of corrupt inspectors M firms offer bribes to the inspectors, and any further increase in n reduces the amount of the bribe, keeping the number of corrupt inspectors unchanged at F (case (ii)). 3. Monitoring and bribery We now introduce a government anti-corruption agency (the “regulator”) that monitors firms to apprehend violators, and assume that the regulator is not corrupt. The regulator randomly monitors a fraction m ∈ [0, 1] of the total number of firms. If a firm is caught violating the regulation, it has pay a penalty of an amount PF , as well as implement the regulation; the corresponding inspector has to pay a penalty of PM . The penalty amounts are decided exogenously (perhaps by legislators or the judiciary), and are outside the regulator’s control. When there is monitoring, the maximum bribe that a firm will be willing to pay is B = c − m(c + PF ).

(5)

For expositional ease, in this section, we assume that there are as many inspectors as firms (i.e. M = F), and confine our analysis to the interior equilibrium where only some inspectors are corrupt and the maximum bribe is paid.16 The expected bribe income of a corrupt inspector, under monitoring, becomes B − mPM . It can be shown that a corrupt inspector’s indirect utility is then V(n, B, ) = n (B − mPM ) − . The inspector, who is indifferent to accepting a bribe, will have a psychic cost of  ∗∗ = n (B − mPM ) = n {c − m(c + PF + PM )},

(6)

where we have substituted the value of B using (5). This gives the number of bribe-accepting inspectors (also bribe-paying firms) as ˜ = M

M ∗∗ M  = n {c − m(c + P)}, ¯ ¯

(7)

˜ as given by (7), represents total corruption. Since now there is a regulator carrying out where P PF + PM . Note that M, ˜ and (ii) deterrence activities, this total corruption can be decomposed into (i) detected (by the regulator) corruption, mM, ˜ unchecked corruption, (1 − m)M. The regulator chooses her monitoring frequency (m) optimally, so as to minimize a weighted sum of three social costs: ˜ (ii) the firms’ cost of complying with the regulation, (i) the cost of, or damage from, unchecked corruption, ˇ0 (1 − m)M, ˜ and (iii) the cost of monitoring, (1/2)ı(mM)2 .17 In the last expression, the parameter ı represents the slope c{M − (1 − m)M}, of the marginal cost of monitoring. Thus, denoting the weights on the first two costs as w1 and w2 , respectively, we assume that the regulator minimizes expected social cost Z, where ˜ + w2 c{M − (1 − m)M} ˜ + Z = w1 ˇ0 (1 − m)M

1 ı(mM)2 . 2

(8)

In the first term on the R.H.S. of (8), parameter ˇ0 denotes the marginal damage from unchecked corruption. If, for example, there is a one-to-one correspondence between unchecked corruption and total pollution (i.e. if each corrupt firm, which escapes detection and does not install the green technology, generates one unit of pollution), then we can also interpret ˇ0 as the marginal pollution damage. Since greater openness leads to greater international pressure on a government to control

16

The maximum bribe would be paid if the inspectors have full bargaining power. The regulator, which represents the upper level of the two-tier pubic administrative structure, works to implement the government’s priorities. As such, the regulator behaves much like a government that minimizes social costs in order to ensure the highest level of welfare. Similar assumptions about the regulating agency’s objective function are made by Grieson and Singh (1990) and Bose (1995). 17

218

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

corruption (or pollution), we expect the regulator, acting on behalf of the government, to put more weight on unchecked corruption damage as liberalization progresses and n increases. To capture this idea, we assume that w1 =

˛n , N

(9)

where ˛ > 0 is a constant, and N denotes the total number of goods produced worldwide, i.e. n ∈ (n0 , N].18 To reduce notation, ˜ is the total number of compliant firms, which comply we define ˇ ˛ˇ0 . In the second term on the R.H.S. of (8), M − (1 − m)M with the regulation (i.e. install the green technology) at a cost of c per firm. The third term on the R.H.S. of (8) represents the regulator’s quadratic cost of monitoring the mM firms. In what follows, we analyze the equilibrium when the regulator is unable to credibly pre-commit to her monitoring frequency, m. Assumption of pre-commitment possibility on part of the regulator would lead to a different equilibrium, which is analyzed in Appendix A.19 It is observed that a non-monotonic relation between liberalization and corruption can exist irrespective of the ability, or otherwise, of the regulator to pre-commit to her monitoring frequency. 3.1. Monitoring without pre-commitment When the regulator cannot credibly pre-commit to her monitoring frequency, we have a simultaneous move game between the regulator and the inspectors. The inspectors decide whether to be corrupt or not, taking m as given. Their reaction function is given by (7). The regulator, on the other hand, minimizes (8) with respect to m, taking total corruption ˜ as given. The first order condition gives the regulator’s best response as a function of total corruption: (M) m∗ =

˜ M((n/N)ˇ − w2 c) . ıM 2

(10)

The second order condition for minimizing Z is satisfied as ıM2 > 0. To ensure an interior equilibrium, where the regulator chooses a positive monitoring frequency, we assume the following lower limit for n: n>

w2 cN ≡ n0 . ˇ

(11)

From (10), we observe that monitoring by the regulator is increasing in the level of total corruption.20 Solving (7) and (10) simultaneously, we get the equilibrium total corruption as ˜ eqm = M

n ıcM 2 ¯ n (c + P)((n/N)ˇ − w2 c) + ıM

.

(12)

Differentiating (12) with respect to n, we get ¯ − n+1 ˇ(c + P)/N} ˜ eqm ∂M n−1 cıM 2 {ıM = . ∂n ¯ + n (c + P)((n/N)ˇ − w2 c)}2 {ıM

(13)

˜ eqm /∂n is greater than, equal to, or less than, zero accordingly as n is less than, equal to, or greater than Thus ∂M ¯ (ıMN/ˇ(c + P))

1/+1

≡ n2 . This leads to the following proposition.

Proposition 2. When the regulator cannot credibly pre-commit to her monitoring frequency, an increase in economic openness initially leads to an increase in the level of total corruption. However, once openness reaches a threshold value (n2 , as defined above), any further increase in openness leads to a decrease in the level of total corruption. How does unchecked corruption (i.e. corruption that evades detection) change as the economy becomes more open? The ˜ eqm )/∂n. This derivative, however, turns out to be a complicated direction of this change depends on the sign of ∂((1 − m∗ )M expression in terms of the parameters, and cannot be unambiguously signed. Hence, we take recourse to a numerical example, and assume the following parameter values: ¯ = 1,

N = 100,

M = 10,

 = 1,

c = 0.1,

ı = 0.01,

P = 0.02,

ˇ = 0.2,

w2 = 0.1.

We show that, for these parameter values, the effect of liberalization on unchecked corruption is non-monotonic. In terms of the assumed parameter values, n2 = 20.4. As well, substituting the parameter values in (10)–(12), we get ˜ eqm = 10n/(0.12n(0.2n − 1) + 10). Note that 0 < m* < 0.83 for all m∗ = (0.1n(0.2n − 1))/(0.12n(0.2n − 1) + 10), n0 = 5 and M ˜ eqm )/∂n is greater than, equal ˜ eqm represents an interior equilibrium for n < 12 or n > 34.4. Moreover, ∂((1 − m∗ )M n > 5, and M

18

The lower limit for n is defined in (11). Pre-commitment can be an issue because an ex-ante optimal monitoring frequency, chosen by the regulator before the inspectors decide whether to be corrupt or not, will not be ex-post-optimal for her to implement, once the inspectors have made their decision. Unless the regulator possesses commitment mechanisms for adhering to announced policies even after they become sub-optimal, her ex-ante monitoring policy will lack credibility with the inspectors. 20 Papers on optimal deterrence, such as Grieson and Singh (1990) and Bose (1995), show this to be the optimal response of regulators in simultaneous move games between potential violators and a regulatory authority. 19

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

219

to, or less than zero according as n is less than, equal to, or greater than 13.5. Thus liberalization has a non-monotonic effect on unchecked corruption, as laid out by the following result. Result 1. Suppose the regulator cannot credibly pre-commit to her monitoring frequency, and parameter values are as given in our numerical example. Then, an increase in economic openness initially leads to an increase in the level of unchecked corruption (as long as n < 13.5) but leads to its decrease later on (when n > 13.5). 3.2. Discussion The regulator chooses her optimal monitoring frequency to equate the marginal benefit of monitoring (i.e. reduced corruption damage) to the marginal cost (consisting of firms’ compliance cost and the regulator’s monitoring cost). From (10), we see that the regulator chooses a positive monitoring frequency as long as weighted marginal corruption damage ((n/N)ˇ) exceeds weighted marginal compliance cost (w2 c). When this does not happen (i.e. (n/N)ˇ ≤ w2 c) the regulator ˜ max , as in Section 2. will choose not to monitor at all (i.e. m* = 0), and both total and unchecked corruptions would equal M For Section 3, however, we exclude this no-monitoring equilibrium by assuming, in (11), that n > n0 ≡ w2 cN/ˇ. When the regulator is unable to pre-commit, she chooses her optimal monitoring frequency as the best response to the given level of corruption. To her, the efficacy of monitoring then only lies in increasing the number of detected violators and decreasing the number of unchecked violators (holding total corruption constant). On the other hand, monitoring under pre-commitment allows the regulator to reduce unchecked corruption both by reducing total corruption (by influencing the inspectors’ decision to be corrupt), as well as by increasing detected corruption. Because of this higher marginal benefit, the monitoring frequency chosen by the regulator when she is able to pre-commit (m** ) will exceed that chosen by her when she is unable to pre-commit (m* ).21 Comparing (10) with appendix Eq. (A.1), using (11) and (12), we have m∗∗ − m∗ =

{n (c

¯ n (c + P)((n/N)ˇ − w2 c){n P((n/N)ˇ − w2 c) + ıM} > 0.  ¯ ¯ + P)((n/N)ˇ − w2 c) + ıM}{2n (c + P)((n/N)ˇ − w2 c) + ıM}

(14)

Since (7) shows that higher monitoring leads to lower total corruption, we have the following proposition. Proposition 3. Both total and unchecked corruption are lower, when the regulator is able to credibly pre-commit to her monitoring frequency, than their corresponding values, when the regulator is unable to do so. Proof.

˜ Follows from (7), (14), and the definition of unchecked corruption, (1 − m)M.

In general, in the presence of monitoring by a regulating (anti-corruption) agency, an increase in economic openness has two opposing effects on the corruptible inspectors. There is a “benefit effect” to the inspectors, as bribery becomes a more “desirable” option in the face of greater variety in consumption. As more consumption goods become available, the marginal utility of bribe increases, which induces more inspectors to become corrupt. On the other hand, greater openness also results in greater international pressures on the government to crack down on corruption (or pollution). This raises deterrence activity by the regulator, and the cost of expected penalty to inspectors and firms. This “cost effect” leads to less corruption. When the level of openness is low, the “benefit effect” exceeds the “cost effect,” and corruption rises with increasing openness.22 However, when the level of openness exceeds a threshold, the “cost effect” dominates the “benefit effect,” and corruption decreases with openness. The result is an inverted-U curve in liberalization and corruption. 4. Empirical evidence 4.1. Strategy and data While a comprehensive test of the model is difficult with available data, our objective in this section is to provide some suggestive evidence of a non-linear relation between corruption and liberalization as predicted by our model. To undertake the empirical investigation we use cross-country data and construct a panel dataset. Two key variables in our model are corruption and liberalization. We use the International Country Risk Guide (ICRG) corruption index introduced by Knack and Keefer (1995). The data are yearly, and cover the period 1984–2005. Lower corruption scores indicate that “high government officials are likely to demand special payments” and that “illegal payments are generally expected throughout lower levels of government” in the form of “bribes connected with import and export licenses, exchange controls, tax assessment, policy protection, or loans” (see Knack and Keefer, 1995, p. 225). The corruption score for countries is from 0

21 That pre-commitment would result in a higher level of deterrence is also evident from Becker (1968) and Grieson and Singh (1990). Our objective is to show that a non-monotonic relation between liberalization and corruption can exist irrespective of the ability, or otherwise, of the regulator to pre-commit to her monitoring frequency. 22 For example, when n is close to n0 , from (10) we have m* is close to zero. With (close to) zero monitoring, the cost effect is (nearly) non-existent and is dominated by the benefit effect.

220

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224 Table 1 Summary statistics: panel data using ICRG. Variable

N

Mean

Range

S.D.

corr ln(GDPpc) impGDP Fuelex

310 310 310 310

2.71 8.65 37.14 14.64

5.46 4.60 128.53 98.03

1.25 1.13 20.73 25.33

corr

lnGDPpc

impGDP lag

fuelex

1.000 −0.621 −0.007 0.249

1.000 0.136 −0.045

1.000 −0.049

1.000

Table 2 Correlations for variables in panel data.

corr lnGDPpc impGDP lag fuelex

to 6, where zero implies higher corruption. For clarity in presentation of our results, we transformed the data by subtracting the index from six, so that high values of the index mean a higher level of corruption. As a proxy for liberalization, we use the ratio of import to GDP. Increasing imports can allow a country to increase the number of goods available to it for consumption.23 Furthermore, in the literature evaluating the determinants of corruption, import to GDP ratio has been used by a number of previous studies as a measure of openness (see Ades and Di Tella, 1999; Treisman, 2000; Gatti, 2004 among others). Data for this variable is obtained from the World Development Indicators (WDI). GDP per capita has been argued to be an important determinant of the level of corruption.24 Hence to assess the relationship between corruption and import/GDP, an important control variable for our estimation would be the logarithm of GDP per capita. Data for GDP per capita in purchasing power parity (PPP) terms is obtained from the WDI. In addition, based on the findings and arguments of Treisman (2000), we include fuel exports as a fraction of total exports as an explanatory variable in our regression to control for natural resource dependence of an economy.25 Our panel data using ICRG data consists of 121 countries. These countries are listed in Appendix B. To smooth out business cycle related fluctuations, and also because annual variation in corruption indices may reflect measurement errors, we take 5-year averages of all variables. Table 1 shows the means, standard deviations, and the range for the 5-year average values of the variables used in the regressions. Correlations between the variables that we use for our analysis are presented in Table 2. Following the arguments presented by Islam (1995) we use a fixed-effect model since this allows us to control for countryspecific fixed effects, such as time-invariant cultural differences, that may affect the level of corruption. Our basic regression equation is given by corri,t = c + ˛1 impGDPi,t−1 + ˛2 impGDPsqi,t−1 + ˇ1 ln(GDPpc)i,t + ˇ2 fuelex + εi,t + i ,

(15)

where at time period t, i represents some country; corr is corruption and is measured using the ICRG measure rescaled as discussed above. To address the issue that cotemporaneous impGDP may be an endogenous variable since it may depend on the level of corruption, we use one period lagged value for the ratio of import to GDP, impGDPt−1 , while impGDPsqt−1 is the squared values of this variable. The logarithm of GDP per capita in PPP terms is represented by ln(GDPpc) and is our proxy for institutional development of a country; fuelex is the ratio of fuel exports to total exports and is a proxy for natural resource rents enjoyed by a country; ε represents the i.i.d. error term while  represents country-specific fixed effects. For an inverted-U shaped relationship between corr and impGDP, we would expect ˛1 > 0 and ˛2 < 0. 4.2. Regression results As mentioned in the introduction, both the “benefit effect” (due to availability of more goods and services) and “cost effect” (due to international pressure to improve quality of governance) of liberalization are likely to be more pronounced in developing (lower GDP per capita) rather than developed (higher GDP per capita) countries. Accordingly, we estimate regression Eq. (15) to assess the relationship between corruption and liberalization for different levels of development. It should be noted that liberalization can affect corruption through alternative channels (arising from changes in inspectors’

23 Import to GDP ratio is one of the best available proxies for changes in variety of goods following liberalization. Although there has been some progress in constructing variety indices across countries (e.g. Broda and Weinstein, 2004), data limitations severely constrain our ability to construct such an index for low income countries. 24 Both Ades and Di Tella (1999) and Treisman (2000) found a negative and significant relationship between corruption and GDP per capita. Neither of these cross-sectional studies used ICRG corruption data. 25 Fuel export is the only variable, among the explanatory variables considered by Treisman (2000), which varies appreciably over time. The other variables would be accounted for by country-specific fixed effects for our panel data estimation.

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

221

Table 3 Regression results fixed-effects model using ICRG corruption measure (dependent variable: corr). Variable

(1) GDPpc < 8000

(2) GDPpc < 12,000

(3) GDPpc < 16,000

(4) GDPpc < 20,000

(5) GDPpc < 24,000

(6) GDPpc < 28,000

(7) All

impGDPt−1 impGDPsqt−1 lnGDPpc fuelex cons

0.045 (0.05) −0.0005 (0.03) 1.747 (0.00) 0.024 (0.06) −11.499 (0.00)

0.034 (0.11) −0.0003 (0.13) 1.897 (0.00) 0.021 (0.08) −12.927 (0.00)

0.034 (0.09) −0.0003 (0.11) 1.942 (0.00) 0.021 (0.08) −13.493 (0.00)

0.039 (0.04) −0.0003 (0.05) 1.871 (0.00) 0.013 (0.11) −13.150 (0.00)

0.036 (0.06) −0.0003 (0.06) 2.012 (0.00) 0.013 (0.06) −14.469 (0.00)

0.035 (0.06) −0.0003 (0.05) 2.061 (0.00) 0.0121 (0.07) −15.022 (0.00)

0.009 (0.54) −0.0001 (0.55) 2.382 (0.00) 0.012 (0.06) −18.280 (0.00)

N Countries R2 (within)

178 71 0.225

206 84 0.210

217 88 0.220

232 94 0.226

246 99 0.232

258 103 0.235

310 121 0.267

Notes. p-Values are in parenthesis. For all variables included in the regression we consider averages over the periods 1984–1988, 1989–1993, 1994–1998 and 1999–2005. impGDPt−1 and impGDPsqt−1 represent one period lagged values for the two variables.

willingness to accept bribes and firms’ willingness to pay bribes), and the regression analysis indicates that such trade-off exists but does not exclude multiple channels. Table 3 presents the results for the estimation for different levels of GDP per capita. Column 1 of the table presents the results for relatively low income countries with GDP per capita of less than $8000 in PPP terms in 2003.26 While, for this group of countries, the estimated coefficient for impGDPt−1 is positive and significant, the estimated coefficient for impGDPsqt−1 is negative and significant. This confirms that even after controlling for the level of GDP per capita, fuel exports and crosscountry heterogeneity, there exists an inverted-U shaped relationship between corruption and liberalization. The estimated coefficients for impGDPt−1 and impGDPsqt−1 suggest that corruption increases for values of import to GDP ratio of less than 45% and decreases thereafter. This is in line with the prediction of our model that as countries liberalize corruption would first increase before reversing direction. While the estimated coefficient for ratio of fuel exports to total exports, fuelex, is, as found by previous literature, positive and significant, the estimated coefficient for ln(GDPpc) is, surprisingly, positive and significant.27 This seems at odds with findings in cross-sectional studies that corruption is negatively related to the level of development. However, Fréchette (2006) and Braun and Di Tella (2004), using ICRG panel data and a fixed-effects model, also obtained a similar result.28 In Columns 2–7 of Table 3 we increase GDP per capita to determine whether a similar relationship between corruption and liberalization holds when countries with higher levels of development are included. As evident, we obtain a significant inverted-U relationship between corruption and liberalization for countries with GDP per capita of less than $28,000, but not for all countries in our sample.29 This suggests that the inverted-U relationship holds only for countries below a certain level of development. To evaluate the robustness of our results we checked for the impact of outliers by excluding observations above 3 deviations of the mean for our variable of interest, impGDP. We find that the inverted-U relationship is robust to such exclusions.30 In addition, we use data obtained from “corruption perceptions index” (CPI) compiled by Transparency International as an alternative measure for corruption for the 1995–2005 period.31 Table 4 presents the results from estimating regression (15) using this measure of corruption.32 As before, we use one period lagged value for impGDP for the regression. Similar to the estimates presented using the ICRG data for corruption, we find a non-monotonic and concave relationship between

26 The lowest upper bound of GDP per capita for which we get significant results is $2000. Considering an upper bound lower than this level yields expected signs for the estimated coefficients; however the consequent reduction in the number of observations makes the estimates statistically insignificant. Also, the choice of year for GDP per capita, for construction of income cut-offs, does not qualitatively alter our results. 27 Even though the ICRG measure of corruption is negatively correlated with ln(GDPpc) in Table 2, and is negatively related to ln(GDPpc) in cross section regressions, after controlling for country-specific fixed effects we find a positive correlation between the two variables. 28 Braun and Di Tella (2004, p. 93) explain their findings as: “Perhaps surprisingly, we find that GDP per capita is positively correlated with corruption in the panel regressions (although it is negatively related in the cross-section). This is consistent with corruption having a pro-cyclical nature. A number of authors have emphasized that ‘moral standards’ are lowered during booms, as greed becomes the dominant force for economic decisions (e.g. Kindleberger, 2000).” 29 The estimate of the coefficient for impGDPsqt−1 is only significant at 11% and 13% for countries with GDP per capita less than $16,000 and $12,000 but is significant at 3% level for countries with GDP per capita less than $8000. 30 The finding of an inverted-U relationship between corr and impGDP is robust to the inclusion of controls for the extent to which a country is democratic. To ascertain the robustness, the democracy indices that we used were ifhpol (Hadenius and Teorell, 2007) and polity2 from the Polity IV: Regime Authority Characteristics and Transitions Datasets (www.systemicpeace.org/inscr/inscr.htm). The results of the robustness checks are available from the authors upon request. 31 Other studies on corruption such as Bardhan (1997) and Treisman (2000) have reported and used CPI as a measure of corruption. Ades and Di Tella (1999, p. 985) have defended the use of subjective indices of corruption for empirical analyses. 32 Even though CPI provides cross-sectional data for a particular year, the sources used for construction of the index vary from one year to the next. Lambsdorff (2005) provides time series data for the 1995–2005 period for 61 countries based on an assessment of sources that continuously enter the index. We use this data for our estimation, after modifying the scores such that they range from 0 (no corruption) to 10 (highly corrupt). Our panel data using the Lambsdorff (2005) time series CPI data for the 1995–2005 period consists of 60 countries as data for Taiwan for import to GDP ratio was not available from the WDI.

222

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

Table 4 Regression results using “corruption perceptions index” (dependent variable: corr). Variable

(1) GDPpc < 8000

(2) GDPpc < 12,000

(3) GDPpc < 16,000

(4) GDPpc < 20,000

(5) GDPpc < 24,000

(6) All

impGDPt−1 impGDPsqt−1 lnGDPpc fuelex cons

0.072 (0.02) −0.0007 (0.04) −1.178 (0.00) 0.0008 (0.91) 15.261 (0.00)

0.056 (0.00) −0.0005 (0.01) −1.029 (0.00) 0.0050 (0.47) 13.994 (0.00)

0.039 (0.04) −0.0003 (0.06) −0.959 (0.00) 0.0058 (0.41) 13.691 (0.00)

0.039 (0.05) −0.0003 (0.10) −0.990 (0.00) 0.0070 (0.35) 13.830 (0.00)

0.040 (0.04) −0.0003 (0.07) −0.874 (0.00) 0.0067 (0.36) 12.618 (0.00)

−0.001 (0.96) 0.0001 (0.47) −0.866 (0.00) 0.0082 (0.22) 12.388 (0.00)

N Countries R2 (within)

204 20 0.069

300 29 0.058

322 31 0.044

365 35 0.037

398 38 0.031

627 60 0.023

Notes. p-Values are in parenthesis. Annual data for all variables is used for this regression. impGDPt−1 and impGDPsqt−1 represent one period lagged values for the two variables.

corruption and the ratio of import to GDP for countries with GDP per capita of less than $24,000 but not for all countries in the sample. 5. Concluding remarks We have shown that an inverted-U curve can describe the incidence of bureaucratic corruption that accompanies increased levels of economic openness in developing countries. This result was derived analytically for total corruption when the regulator is unable to pre-commit (Proposition 2), and numerically for the other cases (Result 1 and Appendix A). Moreover, the regulator’s ability to credibly pre-commit to deterrence activities is shown to have a detrimental effect on corruption (Proposition 3). Empirically we found evidence for a non-monotonic relation between corruption and openness (as proxied by import–GDP ratio) among lower income countries. It should be pointed out that if liberalization decreased (increased) profitability in some sectors of the economy, it would decrease (increase) the willingness of firms in those sectors to pay a bribe, and thus decrease (increase) corruption only among public officials enforcing regulations in those sectors. We, however, did not model this obvious link between liberalization and corruption. In our model, the positive effect of liberalization on corruption comes through another, less obvious, channel: the higher willingness of all officials, enforcing regulations in all sectors of the economy, to accept a bribe (Proposition 1). Thus, our result suggests that the impact of liberalization on corruption will be sectorally more widespread than what is suggested by the obvious channel. Acknowledgements We thank Chris Ahlin, Luciana Echazu, Andrew Hussey, Christopher Kilby, Johann Lambsdorff, Ngo Van Long, Abhirup Sarkar, James Townsend and reviewers of this journal for helpful comments. Pinaki Bose gratefully acknowledges summer support for this research in form of a faculty research grant from the Fogelman College of Business and Economics and the Wang Centre of the University of Memphis. Appendix A. Monitoring with pre-commitment When the regulator can pre-commit to her monitoring frequency, we have a sequential game consisting of the following stages. First, the regulator chooses her monitoring frequency, m, so as to minimize Z. Second, the inspectors decide whether to accept bribes from the firms. Non-bribing firms implement the regulation (i.e. install the green technology). In the third stage, the regulator randomly monitors the firms, and apprehends violators. Firms caught violating the regulation have to pay the penalty PF (the corresponding inspectors have to pay PM ), as well as implement the regulation. To derive the subgame perfect Nash equilibrium of this sequential game, we solve it backwards. The solution of the last two stages is given by (5)–(7). While optimally choosing her monitoring frequency in the first stage, the regulator takes into consideration the reaction of the inspectors to any given m, as specified by (7). Substituting (7) and (9) into (8), and minimizing Z with respect to m, we get the optimal monitoring frequency, from the first order condition, as m∗∗ =

n (2c + P)((n/N)ˇ − w2 c) ¯ 2n (c + P)((n/N)ˇ − w2 c) + ıM

.

(A.1)

The assumption in (11) ensures that m** is a positive fraction. The second order condition for minimizing Z is also satisfied. Substituting m** from (A.1) into (6), we get equilibrium  ** as ∗∗ eqm =

¯ − n P(c + P)((n/N)ˇ − w2 c)} n {cıM .  ¯ 2n (c + P)((n/N)ˇ − w2 c) + ıM

(A.2)

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

223

∗∗ The effect of liberalization on total (respectively, unchecked) corruption depends on the sign of the derivative of eqm ∗∗ ) with respect to n. Analytically, however, these derivatives cannot be signed unambiguously. (respectively, (1 − m∗ )eqm Hence, we take recourse to the numerical example in Section 3.1, and show that, for these parameter values, the effect of liberalization on corruption is non-monotonic. Substitution of the parameter values into (A.1) and (A.2) gives m∗∗ = (0.22n(0.2n − 1))/(0.24n(0.2n − 1) + 10), and ∗∗ ∗∗ represents an interior eqm = (n{1 − 0.0024n(0.2n − 1)})/(0.24n(0.2n − 1) + 10). Note that 0 < m** < 0.92 for all n > 5, and eqm ∗∗ /∂n (respectively, ∂((1 − m∗∗ ) ∗∗ )/∂n) is greater than, equal to, or less than zero equilibrium for n < 48.2. It is seen that ∂eqm eqm according as n is less than, equal to, or greater than 12.9 (respectively, 9.2). Thus, in the numerical example, the impact of liberalization on both total corruption and unchecked corruption is described by an inverted-U shaped curve.

Appendix B. List of countries for ICRG panel data Albania Algeria Angola Argentina Armenia Australia Austria Azerbaijan Bahrain Bangladesh Belarus Belgium Bolivia Botswana Brazil Bulgaria Burkina Faso Cameroon Canada Chile China Colombia Costa Rica Cote d’Ivoire Croatia Cyprus Czech Republic Denmark Dominican Republic Ecuador

Egypt, Arab Rep. El Salvador Estonia Ethiopia Finland France Gabon Gambia, The Germany Ghana Greece Guatemala Guinea Guinea-Bissau Guyana Haiti Honduras Hong Kong, China Hungary Iceland India Indonesia Iran, Islamic Rep. Ireland Israel Italy Jamaica Japan Jordan Kenya

Korea, Rep. Kuwait Latvia Lebanon Lithuania Luxembourg Madagascar Malawi Malaysia Mali Malta Mexico Moldova Mongolia Morocco Mozambique Namibia Netherlands New Zealand Nicaragua Niger Nigeria Norway Oman Pakistan Panama Papua New Guinea Paraguay Peru Philippines

Poland Portugal Romania Saudi Arabia Senegal Slovak Republic Slovenia South Africa Spain Sri Lanka Sudan Suriname Sweden Switzerland Syrian Arab Republic Tanzania Thailand Togo Trinidad and Tobago Tunisia Turkey Uganda Ukraine United Arab Emirates United Kingdom United States Uruguay Vietnam Yemen, Rep. Zambia Zimbabwe

Variable definitions and their data sources Variable

Description and data source

corr

Panel data provided by ICRG for 1984–2005. Source: International Country Risk Group. As a second measure of corruption we use panel data provided by Lambsdorff (2005) for 1995–2005. Source: Transparency International. Real GDP per capita measured in purchasing power parity terms for 1995–2005. Source: World Development Indicators. Import to GDP ratio for 1990–2005. Source: World Development Indicators. Fuel exports as a fraction of total exports. Source: World Development Indicators.

GDPpc impGDP Fuelex

References Acemoglu, D., Verdier, T., 2000. The choice between market failures and corruption. American Economic Review 90, 194–211. Ades, A., Di Tella, R., 1999. Rents, competition, and corruption. American Economic Review 89, 982–993. Ahlin, C., Bose, P., 2007. Bribery, inefficiency, and bureaucratic delay. Journal of Development Economics 84, 465–486. Bardhan, P., 1997. Corruption and development: a review of issues. Journal of Economic Literature 35, 1320–1346. Becker, G., 1968. Crime and punishment: an economic approach. Journal of Political Economy 76, 169–217. Bose, P., 1995. Regulatory errors, optimal fines, and the level of compliance. Journal of Public Economics 56, 475–484. Braun, M., Di Tella, R., 2004. Inflation, inflation variability, and corruption. Economics & Politics 16, 77–100. Broda, C., Weinstein, D., 2004. Variety growth and world welfare. American Economic Review 94, 139–144. Dixit, A., Stiglitz, J., 1977. Monopolistic competition and optimum product diversity. American Economic Review 67, 297–308.

224

S. Baksi et al. / Journal of Economic Behavior & Organization 72 (2009) 214–224

Fréchette, G.R., 2006. Panel Data Analysis of the Time-Varying Determinants of Corruption. CIRANO Working Papers 28. Gatti, R., 2004. Explaining corruption: are open countries less corrupt? Journal of International Development 16, 851–861. Grieson, R., Singh, N., 1990. Regulating externalities through testing. Journal of Public Economics 41, 387–396. Hadenius, A., Teorell, J., 2007. Pathways from authoritarianism. Journal of Democracy 18, 143–156. Islam, N., 1995. Growth empirics: a panel data approach. Quarterly Journal of Economics 110, 1127–1170. Kindleberger, C., 2000. Manias, Panics and Crashes: A History of Financial Crashes. John Wiley, New York. Knack, S., Keefer, P., 1995. Institutions and economic performance: cross-country tests using alternative institutional measures. Economics and Politics 7, 207–227. Klenow, P., Rodriguez-Clare, A., 1997. Quantifying Variety Gains from Trade Liberalization. University of Chicago, mimeo. Krugman, P., 1980. Scale economies, product differentiation and the pattern of trade. American Economic Review 70, 950–959. Lambsdorff, J., 2005. Determining Trends for Perceived Levels of Corruption. University of Passau Discussion Paper V-38-05. Lambsdorff, J., 2007. The Institutional Economics of Corruption and Reform: Theory, Evidence, and Policy. Cambridge University Press, Cambridge. Lawrence, C., Spiller, P., 1983. Product diversity, economies of scale, and international trade. Quarterly Journal of Economics 98, 63–83. Leff, N., 1964. Economic development through bureaucratic corruption. American Behavioral Scientist 8, 8–14. Mauro, P., 1995. Corruption and growth. Quarterly Journal of Economics 110, 681–712. Mauro, P., 1998. Corruption and the composition of government expenditure. Journal of Public Economics 69, 263–279. Mitra, S., 2005. Corruption Reexamined. SSRN Working Paper. Romer, P., 1994. New goods, old theory, and the welfare costs of trade restrictions. Journal of Development Economics 43, 5–38. Rose-Ackerman, S., 1975. The economics of corruption. Journal of Public Economics 4, 187–203. Rose-Ackerman, S., 1978. Corruption: A Study of Political Economy. Academic Press, New York. Sapru, R., 1998. Environmental policy and politics in India. In: Desai, U. (Ed.), Ecological Policy and Politics in Developing Countries: Growth, Democracy and Environment. State University of New York Press, Albany, pp. 153–182. Shleifer, A., Vishny, R., 1993. Corruption. Quarterly Journal of Economics 108, 599–617. Spence, M., 1976. Product selection, fixed costs and monopolistic competition. Review of Economic Studies 43, 217–235. Transparency International, 2005. Global Corruption Report. Online at http://www.transparency.org/publications/gcr. Treisman, D., 2000. The causes of corruption: a cross-national study. Journal of Public Economics 76, 399–457. Wei, S., 2000a. How taxing is corruption on international investors? Review of Economics and Statistics 82, 1–11. Wei, S., 2000b. Natural Openness and Good Government. National Bureau of Economic Research Working Paper No. 7765. Cambridge, MA. Williams, J., Beare, M., 1999. The business of bribery: globalization, economic liberalization, and the ‘problem’ of corruption. Crime, Law, and Social Change 32, 115–146.