Inducing independence: A strategic model of World Bank assistance and legal reform

Abstract

Legal reforms matter for economic growth and democratic consolidation. As part of the “second-generation reforms”, international financial institutions have sought to build the rule of law by funding a vast array of legal and judicial reform projects throughout the developing world. Yet aside from scattered anecdotal evidence, the general effects of international assistance on legal reform and the rule of law remain poorly understood. This article addresses this gap by developing a theoretical framework that explores the strategic interaction among international financial institutions, national governments and non-governmental actors. Using original data on World Bank legal and judicial reform projects, we show that World Bank assistance can in fact encourage some types of incumbent governments to promote reforms that increase judicial independence.

Keywords

International aid judicial independence judicial reform legal reform rule of law World Bank

As part of the “second-generation reforms”, international financial institutions have sought to build the rule of law¹ by funding a vast array of legal reform projects throughout the developing world.² The basic logic underlying this recent wave of financial transfers stems from the widespread view that efficient and impartial judiciaries are the key to sustainable economic development and democratic consolidation (e.g. World Bank, 2003, 2004). Yet despite the growing involvement of international lending agencies in such reforms, standard theories about the origins of judicial independence and the rule of law are largely confined to analyzing the strategic interaction among domestic political actors (e.g. Ferejohn, 1999; Ramseyer, 1994; Weingast, 1997). Aside from scattered anecdotal evidence, the general effects of international aid on legal reform and the rule of law remain poorly understood.

This article addresses this gap by developing a general theoretical framework that explores the strategic interaction among international financial institutions, national governments and non-governmental actors. We begin with the simple but widespread observation that formal institutional protections are rarely sufficient for explaining judicial independence: nearly every modern constitution mandates an independent judiciary. However, in practice, independence is relatively scarce, and while no single set of actors has a monopoly on undermining judicial independence, it is often the case that incumbent politicians have both the temptation and capacity to threaten judicial autonomy.

Under the broad rubric of rational choice theory, scholars have thus variously sought to explain the “puzzle of judicial independence” as a solution to commitment problems between legislators and interest groups (Landes and Posner, 1975), an information-generating mechanism (McCubbins and Schwartz, 1984; Rogers, 2001; Stephenson, 2004), a device for blame avoidance (Salzberger, 1993; Magaloni and Sánchez, 2001), an “insurance policy” for parties facing electoral uncertainty (Ramseyer, 1994; Stephenson, 2003a), a product of political fragmentation (Cooter and Ginsburg, 1996; Iaryczower et al., 2002; Rios-Figueroa, 2007; Tsebelis, 2002), or as the result of strategic judicial decision-making (Epstein and Knight, 1998; Ferejohn and Weingast, 1992; Helmke, 2005; Spiller and Gely, 1990; Vanberg, 2001). Although this body of research has generated enormous insight into the incentives and constraints underlying judicial independence, most strategic-based accounts either entirely ignore international efforts at bolstering the rule of law, or assume that certain domestic conditions must already be in place for internationally backed efforts to succeed (e.g. Stephenson, 2003b).

We take a different tack. We develop a simple game-theoretic model to explore the conditions under which international institutional involvement increases the willingness of domestic politicians to promote independence-enhancing judicial reforms. In particular, we consider the role that non-governmental actors (e.g. bar associations and various legal non-governmental organizations) play in helping the World Bank to reduce the moral hazard problem inherent in internationally sponsored reforms. Game-theoretic modeling is a particularly useful tool, given our interest in the strategic underpinnings of the interactions between the donor organization and national-level actors and of the resulting moral hazard problem. We show that World Bank aid can both raise the bar for domestic actors and provide them with incentives for clearing it. Precisely by demanding third-party support for legal reform, the World Bank may be able to indirectly encourage some types of incumbent governments to promote judicial reforms that increase independence, which would not otherwise do so.

Taken together, our findings provide a fresh perspective on established empirical regularities and add to emerging debates over the effects of international aid. On the one hand, for example, our model suggests another possible mechanism for the association between judicial independence and democracy. Whereas much of the recent literature maintains that only highly competitive democracies are likely to enjoy judicial independence, our findings suggest an alternative route. With a merely pluralistic society—in which non-governmental groups are able to participate in policy reform—in place, international lending agencies may be able to commit sufficient resources that, in turn, help encourage incumbent governments to pursue independence-enhancing judicial reforms.

On the other hand, our findings also contribute to the nascent literature on the effects of international aid. Although research looking at the economic effects of aid is divided,³ studies of the impact of aid on governance in recipient countries present rather pessimistic results. For instance, Knack (2001) finds that higher aid levels decrease the quality of governance, one measure of which is the rule of law. Our theoretical expectation supports the results of Knack’s analysis only in part. Certain types of governments will not tolerate judicial independence regardless of the World Bank’s involvement level; however, some governments can be swayed by the promise of aid and the threat of its suspension.

International aid, judicial reform and “consensus-based” monitoring

Promoting legal and judicial reform was not always a priority at the World Bank. Beginning in the 1970s, the international financial community’s view of reforms was limited to reducing the level of state involvement in the economy in favor of open markets, de-regulation and low tariffs. However, experience soon showed that simply “dismantling” the state was ultimately counter-productive (World Bank, 2003: 12). In light of problems with privatization in the former Soviet Union, the Asian financial crisis and the persistence of poverty and income inequality throughout Latin America and Africa, by the 1990s the World Bank and other international agencies had begun a new phase of institution-building reforms. Touting the importance of a well-functioning legal system capable of protecting property and other individual rights, the Bank’s current mantra is that legal and judicial reform is essential for encouraging domestic and foreign investment, creating jobs and alleviating poverty (World Bank, 2003: 2).

Although initially the World Bank limited its mandate to funding technical legal and judicial reforms that had “direct and obvious” implications for economic development,⁴ such as strengthening judicial administrative capacity, over time the Bank began to take a broader view of its mission. In a 1996 report on initial judicial reform efforts in Latin America, for example, the Bank called for expanding its agenda to include judicial independence as “an imperative feature of any judicial reform project” (Heymann and Lundburg, 2004: 3). Although still wary of involving itself too heavily in domestic politics, policy experts warned of the futility of creating efficient courts which, nevertheless, still lacked the power to uphold the rule of law (World Bank, 2003: 2).

Throughout this period, the Bank also came to believe that the main determinant of success is the level of domestic consensus in favor of reform. The various actors whose support the Bank views as crucial include the executive branch (including the Ministry of Justice), the judiciary (including judges and the judicial bureaucracy), the bar association and various non-governmental organizations, such as human rights groups. Yet, as the Bank has also acknowledged, achieving consensus among these various actors is hardly automatic. In addition to these actors’ disagreements over which reforms matter most, a more fundamental problem is that many actors benefit from maintaining the status quo (Dakolias, 1996: 170).

Consider the executive branch. From Peru and Venezuela to Russia and Argentina, presidents are notorious for dominating policy by subordinating other branches of government, including the judiciary (e.g. Carey and Shugart, 1998). In addition to increasing the chances of the court issuing a contrary opinion, granting independence to judges is also likely to increase public support for the judiciary, thus raising the costs to the government for disobeying such opinions (Carruba, 2002; Staton, 2003). Perhaps not surprisingly then, in such contexts, judicial “reform” often serves merely as a pretext for governments interested in seizing greater judicial control.

For international lenders, the main problem thus lies in the difficulty of determining whether a government’s commitment to independence-enhancing reforms is credible. For example, despite former Peruvian President Alberto Fujimori’s stated commitment to a World Bank-sponsored judicial reform project, the government had routinely interfered in the independence of the judiciary since Fujimori’s autogolpe in 1992 (Kennedy School Government Case Program 1779.1; Heymann and Lundburg, 2004). Although the country had successfully carried through other neo-liberal reforms (e.g. Stokes, 2001), the Bank rightly feared that its resources would be used to further consolidate the government’s control over the judicial system.

The difficulty in directly monitoring the success of judicial reforms only further compounds lenders’ problems. In addition to the fact that the results of many reforms may take years to become manifest, objective measures, particularly regarding judicial independence, are notoriously hard to develop (Stephenson, 2003b). From the Bank’s point of view, this means that both the long-run progress toward the Bank’s objectives and short-run violations of the terms of the loan are difficult to gauge.⁵

In light of these problems, how can the Bank maximize the chances of success? In this article, we explore the idea that the Bank’s demand for consensus between governmental and non-governmental actors may, indirectly, help in the reform process, particularly for the duration of the loan. Specifically, we argue that, although international agencies often have difficulty in directly assessing a government’s commitment to judicial reform, particularly at the stage of project development, the involvement of non-governmental actors can be instrumental in the implementation of World Bank projects. Starting with the assumption that most non-governmental actors stand to benefit from an independent judiciary, while the incumbent government may or may not, we explore how the level of collaboration between the government with reform preferences unknown to the donor, on the one hand, and third-party actors with known pro-reform preferences, on the other hand, may increase the likelihood of project success.

The model

Our model focuses on the interaction of three players: the World Bank, a national government and non-governmental actors (NGAs). The sequence of play is summarized in Figure 1. To capture the fundamental difficulty that international agencies face in determining whether domestic governments are truly committed to reform, we model this as a simple game of incomplete information in which governments know their preferred level of commitment to judicial reform, but the Bank does not. The assumption of incomplete information is consistent with existing research that points out that international bureaucrats are unlikely to have full knowledge of recipient countries and governments’ intentions (e.g. Murrell, 2002).

Figure 1.

Game tree.

Thus, the game begins with Nature’s move, which randomly selects a type of the government. Government types (t) are assumed to be distributed uniformly on an interval [0, 1] ranging from governments that eschew all aspects of legal and judicial reform (t = 0) to governments that are in full accordance with the Bank’s goals of establishing an independent and effective judiciary (t = 1). Next, without knowing the government’s type with certainty, the World Bank makes a decision on the size of its aid commitment. Contributions of all players are normalized so that they can be represented by the [0, 1] interval. Thus, the World Bank commits w∈[0, 1] to the judicial reform project. The government then has an opportunity to choose its contribution to the project i∈[0, 1]. Conceptually, we view the government’s contribution as the effort it makes to realize the Bank’s objectives of judicial impartiality and effectiveness. Following these two decisions, non-governmental actors decide what their support level a∈[0, 1] will be, which we conceptualize as the degree to which they choose to participate in the reform process. We assume that NGAs share a common interest in achieving improvements in judicial and legal institutions in their country. Finally, the World Bank concludes the game by choosing either to release the committed funds completely (the project receives w) or to suspend them (the project receives only w/2⁶).

The utility function of the World Bank is defined as

u_{WB} (w) = ia + r λ w - {(λ w - ia)}^{2 .}

We assume that the World Bank derives utility from shared commitment to judicial reform on the part of the government and non-governmental actors. Thus, the term ia represents the level of shared commitment: if one actor (or both) fails to give support to the project (i.e. i = 0 and/or a = 0), the World Bank receives the lowest utility, all else being equal. rλw is the term that captures the Bank’s utility from releasing aid, where λw is the disbursed amount of aid and r∈[0, 1] is a measure of the World Bank’s interest in a particular project. Finally, we assume that the World Bank attempts to match the level of shared commitment between the government and NGAs in determining its financial support for the project. Thus, the last term in the World Bank’s utility function reduces the Bank’s payoff whenever the level of funding exceeds or falls short of the domestic support for the project.

The national government has the following utility function:

u_{G} (i) = - {(i - t)}^{2} + λ w .

The first term in the utility function reflects the idea that the government receives negative utility from offering more or less support (i) for judicial reforms than its type (t) requires. At the same time, the government always benefits from funding it receives from the World Bank (λw).

Non-governmental actors’ utility function is defined as

u_{NGA} (a) = - {(a - i)}^{2} - {(ia - λ w)}^{2} .

The first term in this utility function represents the notion that NGAs seek to match the level of commitment to the judicial reform demonstrated by the government. For example, if the government is willing to back the reforms in a way that sufficiently accords with the Bank’s mandate, NGAs endorse the government’s reforms. By contrast, if the government is prone to subverting the mandate, NGAs should avoid involvement in aid projects. NGAs also care about external funding for reform: the last term in the NGAs’ utility function decreases their payoff whenever aid exceeds or falls short of the level of shared commitment to the reforms.

This is a sequential game of incomplete information, in which the uninformed player (i.e. the World Bank) moves first. However, the government chooses its support level for the legal and judicial reform project (i), which is observable by the Bank, and thus, through the impact of its decision on the NGAs’ participation level, provides the World Bank with sufficient information to make the decision to suspend or disburse funding at the last node of the game. Since at this decision node the Bank can differentiate between the absence of cooperation on the part of the government and some positive level of support, the Bank acts under complete information, and we apply backwards induction to find the equilibrium of this game.

At the final node, the World Bank chooses whether or not to suspend the funding that it has committed.⁷ The World Bank will fully release funding when the following condition holds:

ia \geq \frac{3}{4} w - \frac{1}{2} r = (ia) * or c * (consensus)

(1)

We can re-write condition (1) to reflect the effect of NGA participation on the Bank’s decision:

a \geq \frac{3 w - 2 r}{4 i} = a *

(2)

If NGAs’ level of support is less than a*, the World Bank will punish the government by releasing only a part of promised funding ( $λ w$ ).

Since NGAs are forward looking and possess all the information from previous decision nodes, they know whether or not conditions (1) and (2) hold. Given this, they choose their optimal involvement level:

a_{p} = \frac{i (1 + λ w)}{1 + i^{2}}

(3)

a_p is NGAs’ preferred level of support for a project in the absence of the external constraint a*, imposed by the World Bank.⁸ However, NGAs know that there are two possible scenarios: a_p $\geq$ a* and a_p < a*. If a_p $\geq$ a*, the World Bank does not suspend aid (i.e. λ = 1). NGAs set a_p at this level when the following condition holds:

i \geq \sqrt{\frac{3 w - 2 r}{4 + w + 2 r}} = \hat{i}

(4)

This means that NGAs are willing to set their support for the project at the level sufficient to deter the World Bank from suspending funds whenever the government displays a sufficient commitment to judicial reform ( $i \geq \hat{i}$ ). Otherwise, NGAs give less support than is required to sustain aid (a_p < a*). The basic logic is that NGAs want to prevent “non-cooperative” governments from receiving an aid windfall, which would merely increase the resources at the government’s disposal.

Next, consider the government’s response. In a world without external constraints, the government’s preferred level of legal reform is i_p, where

i_{p} = t

(5)

However, external constraints induce the government to compare i_p with $\hat{i}$ because its choice will affect the decisions of NGAs and ultimately the World Bank at the final decision node. There are two possible cases to consider:

When $i_{p} \geq \hat{i}$ , the government contributes i_p, its optimal choice, and the World Bank releases λw = w. Only governments with a sufficiently high commitment to reform (t) will voluntarily push for more reform than the minimal requirement that would satisfy NGAs and the Bank:

t \geq \sqrt{\frac{3 w - 2 r}{4 + w + 2 r}} = t_{1} *, where t_{1} * is the first cut point .

(6)

When $i_{p} < \hat{i}$ , governments prefer levels of support for judicial reform that are less than or equal to $\hat{i}$ ; however, not all governments whose i_p is less than $\hat{i}$ actually choose i_p. For some, an increase in i (up to $\hat{i}$ ) is sufficiently compensated by receiving the full amount of World Bank funding. Thus, some governments contribute $\hat{i}$ in spite of their underlying preferences because this prevents the World Bank from suspending its aid, which adequately compensates for the government’s weakened control over the judiciary. Such governments have t less than t₁ * defined above but greater than t₂ *:

t \geq \sqrt{\frac{3 w - 2 r}{4 + w + 2 r}} - \frac{\sqrt{2 w}}{2} = t_{2} *, where t_{2} * is the second cut point .

(7)

Finally, governments with t < t₂ * contribute i_p < $\hat{i}$ and receive λw = $w / 2$ from the World Bank. Thus, as Figure 2 shows, there are two categories of governments: “cooperative” ( $i_{p} \geq \hat{i}$ ) and “non-cooperative” ( $i_{p} < \hat{i}$ ).

Figure 2.

Government types.

At the first decision node, the World Bank has to decide on the optimal level of its financial commitment. The World Bank uses the information on the probability distribution of government types and corresponding expected contributions from different types of governments. The optimal level of w is presented in Figure 3.⁹

Figure 3.

Optimal level of w as a function of r.

Empirical implications

According to the model, the World Bank’s decision to suspend aid depends on whether condition (1) is satisfied. In other words, the Bank gives full aid if domestic consensus meets or exceeds c*, the threshold level of consensus. We can state this theoretical expectation as follows:

Expectation 1.1: As the level of domestic consensus increases, the World Bank is less likely to suspend its funding.

Note that, as Figure 4 shows, w and r have opposite effects on the threshold: the greater the World Bank’s interest in the project (r), the lower the threshold; the higher the level of committed aid (w), the higher the threshold. This leads to two additional expectations regarding the likelihood of aid suspension:

Expectation 1.2: As the amount of committed aid increases, the World Bank is more likely to suspend its funding.

Expectation 1.3: As the level of the World Bank’s interest in the project increases, the World Bank is less likely to suspend its funding.

Figure 4.

The World Bank’s decision to suspend.

In turn, NGAs’ preferred level of support for the project depends on the domestic government’s willingness to implement judicial reform and the amount of World Bank funding. As Figure 5 shows, NGAs are most likely to fully participate in reforms when both the World Bank and the government give their highest level of support for the project. By contrast, if the domestic government does not support the project, NGAs’ support remains at its lowest regardless of the aid amount the World Bank commits. This suggests two testable implications regarding the participation of non-governmental actors:

Expectation 2.1: NGAs are more likely to lend support to judicial reform projects as the level of the government’s support increases.

Expectation 2.2: NGAs are more likely to lend support to judicial reform projects as the amount of World Bank funding increases.

Figure 5.

NGAs’ preferred level of support.

Figure 6 represents the first cut point. Government types that are on or above the plane will voluntarily grant more independence to the judiciary and thus both NGAs and the World Bank will be compelled to give their support to judicial reform projects. These government types have pro-reform preferences, and hence are in no danger of losing any share of committed aid. Figure 7 shows the second cut point, which separates government types that can be induced to respect judicial independence from those that cannot. Here, a combination of three factors determines whether the government will satisfy the donor’s performance requirements and receive the full amount of committed funding. First, the government’s type shapes the government’s support level: government types below the second cut point will not grant more independence to the judiciary even though they expect to have their aid suspended. Two other factors affect the cut point: the donor’s aid commitment and interest level. Taken together, these theoretical conclusions suggest the following empirical implications regarding project outcomes:

Expectation 3.1: Governments with stronger pro-reform preferences will be more likely to complete projects satisfactorily.

Expectation 3.2: As the amount of World Bank funding increases, projects will be more likely to be completed satisfactorily when governments favor reforms.

Expectation 3.3: As the level of World Bank interest increases, projects will be less likely to be completed satisfactorily when governments favor reforms.

Figure 6.

The first cut point.

Figure 7.

The second cut point.

The final theoretical expectation links the World Bank’s interest in supporting judicial reform in a given country to the size of the Bank’s initial aid commitment. As Figure 3 shows, once all subsequent choices made by the other actors have been taken into account, the level of aid is essentially a linear function of the donor’s interest. The expected positive relationship can be summarized as follows:

Expectation 4: As the level of the World Bank’s interest increases, the amount of committed funding increases as well.

Empirical analysis

This section draws on original data to make the first empirical probes of observable implications of the strategic interaction among the World Bank, national governments and non-governmental actors. To construct the World Bank Judicial Reforms Dataset, we relied on the World Bank Projects and Operations Database,¹⁰ which allowed us to generate a comprehensive list of all World Bank projects requiring some judicial or legal reforms.¹¹ For each project, we coded the following information: the start and end dates of the project, the amount of aid committed by the World Bank, whether the aid was suspended, various project assessment measures provided by the World Bank, and whether or not non-governmental organizations participated in the reform project. Currently, the World Bank Judicial Reforms Dataset contains 274 closed or dropped projects from 104 developing and post-communist countries around the world. However, owing to unavailable project evaluations and missing data, the number of observations in our analyses ranges from 60 to 63.

Table 1 lists the variables used in our analyses, including the range of values for each variable and formal model parameters to which they correspond. Whereas some of these variables map easily onto the parameters discussed in the previous section, others require additional explication. For example, capturing the level of World Bank aid (w) is relatively straightforward with our data (i.e. World Bank aid is the total amount of aid in US dollars committed by the World Bank to a given project). The variable Suspend is similarly straightforward.¹² Capturing the level of World Bank interest (r) systematically with available data, however, is more difficult. Here, we create a proxy for World Bank interest using a product of two measures, the rule of law (Rule of law) and government effectiveness (Govt effectiveness), taken from the World Bank Governance Indicators Dataset for each country at the start of each project. Rule of law captures “perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence”, while Government effectiveness represents “perceptions of the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies” (Kaufmann et al., 2009: 6). The logic of these measures is that the World Bank is likely to have the greatest interest in funding projects in countries that have both the greatest need (measured by low rule of law scores) and capacity for judicial reform (measured by high government effectiveness scores).

Table 1.

Variables from the World Bank Judicial Reform Dataset

Variable	Description	Corresponding model parameter
WB aid	Total amount of World Bank (WB) aid committed to a project (range from US$0 to 505.5)	w = Level of World Bank funding
WB interest	Product of two measures: government effectiveness and rule of law, at project approval (range from 0.13 to 0.37)	r = Level of World Bank interest
Suspend	Dummy variable that captures whether the World Bank suspends funding in a project (0 = not suspend; 1 = suspend)	λ = World Bank aid suspension
NGA dummy	Dummy variable that captures whether NGAs participate in the World Bank reform project (0 = no participation; 1 = participation)	a = NGAs’ support level
Law and order difference	Difference between ICRG Law and Order scores from the project approval year and the last year of the project (range from −3 to 1)	i = Government commitment to judicial reform
Consensus	Interactive term: NGA dummy×Law and order difference (range from −1 to 1)	ia = Measure of domestic actors’ consensus on reform
Law and order	ICRG Law and Order scores from the project approval year (range from 1 to 6)	t = Proxy measure of government type
Outcome	Dummy variable that captures the World Bank’s assessment of project performance (0 = unsatisfactory; 1 = satisfactory)	Used as a proxy measure for project success

The variable NGA dummy indicates whether or not non-governmental organizations (which include bar associations, lawyers groups, human rights organizations and other civil groups) were mentioned in the Bank’s project reports as participating in the reform project. Although this measure is not refined in terms of the extent of NGA participation, it serves as a convenient proxy of a, providing a rough and ready indicator of whether or not the reform process was exclusively dominated by the government.

Perhaps the most difficult parameters to capture with the existing data are those associated with the national government. For example, in our theoretical model, NGAs’ participation is contingent on a certain level of national government support (i). Measuring government support objectively, however, is tricky at best. This is especially the case given that we wish to avoid conflating situations where the government may support reform financially, but does so with the intention of creating a subordinate judiciary. As a first cut, here we employ an ordinal variable (Law and order difference) based on the difference in Law and Order scores contained in the International Country Risk Guide (ICRG) from the year prior to the World Bank project beginning and the last year of the project. Although our ability to infer the government’s contribution from such scores is far from perfect,¹³ a positive difference in scores (i.e. the legal environment improves over the course of the project) suggests that the national government used the World Bank resources in a way that roughly corresponded to the World Bank mandate, while a zero or negative difference in scores likely indicates the reverse.¹⁴

Finally, to create a proxy for the type of government the World Bank faces (t), we use the ICRG Law and Order score (Law and order) reported for the year prior to when the project began. Law and order is coded as a combination of two components: the law component is an evaluation of the strength and impartiality of a country’s legal system, while the order component is an evaluation of adherence to the law. We gauge whether different types of governments are more or less likely to satisfy reform conditions attached to aid with a dummy variable (Outcome), which is based on the World Bank’s final assessment of a given project.¹⁵

The first set of questions that comes from our theoretical framework revolves around the basic idea that the Bank’s decision to suspend aid depends on whether governments and non-governmental actors mutually support legal and judicial reform. In the foregoing discussion, we established that the threshold for consensus (c*) that the World Bank requires to avoid suspension is related inversely to the amount of committed aid (Expectation 1.2) and proportionally to the Bank’s interest in a given project (Expectation 1.3). Put simply, higher levels of aid increase the consensus threshold that must be reached; higher levels of interest lower it. At the same time, as the actual level of domestic consensus grows, the recipient country will find it easier to clear the World Bank’s consensus threshold (Expectation 1.1). To explore empirically these propositions, we estimate the following logit model:

Suspend = β_{1} WBaid + β_{2} WBinterest + β_{3} Consensus

(model 1)

The coefficients in model 1 presented in Table 2 are in the predicted directions and reach statistical significance at conventional levels. More informative are the substantive effects derived from evaluating the probability of suspension under different values of the explanatory variables shown in Table 3. When all variables are set at their means, for example, the probability of suspension is just 9.1%. As Consensus decreases from its maximum to the minimum value, while the other regressors remain fixed, the likelihood of suspension grows from 8.7 to 9.5%, which is consistent with Expectation 1.1. When aid grows, while we hold the levels of interest and consensus at their means, this has a much more sizeable effect on the probability of suspension: when the amount of World Bank aid reaches its maximum, the likelihood of suspension is 66.3%, in accordance with Expectation 1.2. Conversely, and consistently with Expectation 1.3, holding aid and consensus constant and increasing donor interest from the mean to its highest level produces a small decline in the likelihood of suspension—from 9.1 to 4%.

Table 2.

Empirical model of strategic interaction among the World Bank, national governments and non-governmental actors

	Model 1	Model 2	Model 3	Model 4
Dependent variable	Suspend	NGA dummy	Outcome	WB aid
Model estimated	Logit	Logit	Logit	Tobit
WB aid	0.006*	0.004*	0.006*
	(0.004)	(0.002)	(0.003)
	[+]	[+]	[+]
WB interest	−10.222***	−5.135***	−3.639	188.980***
	(2.027)	(1.607)	(3.297)	(46.339)
	[−]	[−]	[−]	[+]
Consensus	−0.057*
	(0.032)
	[−]
Law and order difference		1.094***
		(0.383)
		[+]
Law and order			0.393**
			(0.166)
			[+]
N	60	63	61	60

Note: *p < 0.1; **p < 0.05; ***p < 0.01 (two-tailed). Standard errors are reported in parentheses. Expected relationships are in brackets.

Table 3.

Predicted probabilities of suspension, NGA support, and satisfactory evaluation

Variables	Probability of suspension = 1 (model 1)	Probability of NGA dummy = 1 (model 2)	Probability of outcome = 1 (model 3)
WB aid (min)	0.069	0.156	0.702
WB aid (mean)	0.091	0.186	0.736
WB aid (max)	0.663	0.576	0.824
WB interest (min)	0.171	0.237
WB interest (mean)	0.091	0.186
WB interest (max)	0.040	0.126
Consensus (min)	0.095
Consensus (mean)	0.091
Consensus (max)	0.087
Law and order difference (min)		0.012
Law and order difference (mean)		0.186
Law and order difference (max)		0.510
Law and order (min)			0.605
Law and order (mean)			0.736
Law and order (max)			0.866

Note: Predicted probabilities are calculated at minimum, mean and maximum values of a given independent variable that has reached statistical significance, while others are held at their mean values.

Next, we examine the link between the decision of non-governmental actors to support reform and levels of support given by both the World Bank and the national government. Here, the basic intuition we seek to test is whether national governments’ willingness to contribute to the success of the reform has the expected positive effect on non-governmental actors’ support, and whether NGA participation is affected by an increase in World Bank aid. The coefficients reported for model 2 in Table 2 are based on the following logit model:

N G A d u m m y = β_{1} WBaid + β_{2} WBinterest + β_{3} LawOrderDiff

(model 2)

The results from model 2 are fully consistent with our theoretical expectations and reach statistical significance at conventional levels. Table 3 examines substantive effects of the donor’s aid commitment and government contributions on NGA participation. We start with a baseline that fixes all regressors at their mean values, and calculate the probability of non-governmental actors participating in a project. This baseline probability is 18.6%. Increasing the government’s level of support (Law and order difference) to the maximum value, while the other regressors remain at their means, increases the likelihood of NGA participation to 51%. When we vary World Bank aid from its mean to its maximum, the probability that NGAs participate increases substantially—by 39 percentage points. Finally, we observe an increase in the probability of NGA involvement by 5.1 percentage points when World Bank interest declines from its mean to the minimum value, while the remaining regressors are fixed at their mean values. Thus, Expectations 2.1–2.3 receive empirical support in our analysis, and the effects of the donor’s and governments’ actions on NGAs’ decision to get involved are substantively significant.

In the next set of results, we explore the idea that different “types” of governments may be more or less influenced by the World Bank’s assistance and interest level. As an initial test of this set of theoretical expectations, we use Law and Order scores as a proxy for the type of government and then examine whether “cooperative” government types (i.e. those above the second cut point) are more responsive than “non-cooperative” government types (i.e. those below the second cut point). In order to determine government type, we (1) estimate the following ordinary least squares (OLS) model: Law and order = β₁ WBaid +β₂ WBinterest; (2) calculate predicted values of Law and order, which then serve as the second cut point represented by t₂ * in the theoretical model; and (3) generate a binary government type variable (Ctype) that takes the value of 1 if a government’s Law and order score is greater than or equal to the predicted value, and 0 otherwise. We then interact World Bank aid and World Bank interest with this binary variable because our theoretical results suggest that only “cooperative” governments are affected by these two factors, and estimate the following logit model:¹⁶

Outcome = β_{1} WBaid * Ctype + β_{2} WBinterest * Ctype + β_{3} Law and order

(model 3)

Table 2 shows that governments with pro-reform preferences, as expected, are more likely to receive a positive evaluation from the World Bank. While the probability of a positive evaluation is quite high even for governments that do not wish to grant more independence to the judiciary—that is, 60.5%—the likelihood of a positive outcome rises to 73.6 and 86.6% when the measure of government type (Law and order) takes its mean and maximum values, respectively, while holding other regressors at their means (see Table 3, column 3). We also find that the performance evaluation of “cooperative” types improves when aid increases, but declines when the donor’s interest grows, although the latter result is not statistically significant. The probability of “cooperative” types receiving a positive evaluation increases by 8.8 percentage points when World Bank aid increases from its mean to its maximum.

Finally, we test the theoretical expectation linking the donor’s level of interest in supporting a legal reform project in a given recipient country with the size of the donor’s aid commitment. Since, in equilibrium, the World Bank’s interest is the sole determinant of the level of the donor’s commitment, we estimate the following tobit model:

WBaid = β_{1} WBinterest

(model 4)

The last column of Table 2 shows empirical support for the hypothesized positive relationship between the World Bank’s interest and commitment. To illustrate this relationship, Figure 8 presents a plot of the predicted size of the World Bank’s aid. The donor’s commitment increases by approximately US$50 million as the level of the donor’s interest ranges from the minimum to the maximum value. This funding increase is substantial: the median aid amount is only US$20.5 million in our dataset, while the average is approximately US$60 million. Therefore, a country with the greatest need for external assistance (i.e. with the lowest rule of law score) and the best ability to implement required reforms (i.e. with the highest government effectiveness score) attracts a larger aid commitment than a country without these characteristics.

Figure 8.

Predicted World Bank aid commitment (based on model 4).

Robustness checks: the effect of World Bank assistance on de facto and de jure judicial independence

To conduct additional tests of the effect of World Bank projects on judicial independence in recipient countries, we collected data for measures gauging changes in de jure and de facto levels of judicial independence during project implementation. Melton and Ginsburg (2012) code six binary variables that reflect key dimensions of judicial independence: (1) presence/absence of an explicit statement in the country’s constitution regarding the independence of the judiciary; (2) presence/absence of a life-term judicial tenure; (3) presence/absence of selection procedures for judicial appointments; (4) stringency of procedures for judicial removal; (5) limited/unlimited removal conditions; and (6) presence/absence of judicial salary insulation. In addition, Melton and Ginsburg create indicators of de jure independence, using these six measures. We follow this approach and generate two de jure indices—an additive index of all six measures, and an additive index of three selection and removal procedure indicators (i.e. variables 3–5). Data for the de facto measure of judicial independence are from Linzer and Staton (2011). Their approach is to use eight variables from previous studies of judicial independence to run a model that yields estimated levels of de facto independence.

In order to construct dependent variables appropriate for our robustness checks, we calculate the change in these measures between project initiation and completion years. Table 4 reports our robustness checks with these variables. We replaced our original measure (the Outcome dummy) and replicated the analysis using the two-step estimation procedure, as previously. The only difference is that, instead of using logit in the second stage, we used OLS in models with Δde jure independence (6), Δde jure independence (3) and Δde facto independence as dependent variables.

Table 4.

Robustness checks for the government performance model using measures of de facto and de jure judicial independence

	Model 4 (from Table 2)
Dependent variable	Outcome dummy	Δde jure independence (6)	Δde jure independence (3)	Δde facto independence
Model estimated	Logit	OLS	OLS	OLS
WB aid	0.006*	0.003**	0.001*	0.000
	(0.003)	(0.001)	(0.001)	(0.000)
	[+]
WB interest	−3.639	−2.619***	−1.829***	−0.060
	(3.297)	(0.777)	(0.587)	(0.065)
	[−]
Law and order	0.393**	0.159***	0.098***	0.002
	(0.166)	(0.036)	(0.027)	(0.003)
	[+]
N	61	106	106	110

Note: *p < 0.1; **p < 0.05; ***p < 0.01 (two-tailed). Standard errors are reported in parentheses. Expected relationships are in brackets.

Note that the key difference between the de facto and de jure aspects of judicial independence is that the former is a long-term indicator and hence, from the standpoint of judicial reform, takes a long time to evolve as a result of policy changes implemented by the government (Feld and Voigt, 2003; Hayo and Voigt, 2007; Melton and Ginsburg, 2012). Thus, Melton and Ginsburg (2012: 18) construct a de facto measure by averaging “Linzer and Staton’s measure for each country from 1980–2003”. The latter is, on the contrary, a short-term indicator; therefore, judicial reforms are likely to result in de jure changes that can be observed soon after the reforms are implemented. Given the relatively short time horizon of the legal and judicial projects funded by the World Bank (on average, approximately 6 years from initiation to completion), we expected to see differences in empirical results based on these two aspects of judicial independence. In fact, we do find that aid has the expected positive effect on de jure judicial independence, but no influence on de facto independence.¹⁷

Is there evidence of selection bias?

The positive relationship between World Bank assistance and legal and judicial reforms implemented by recipient governments could be spurious if recipients with better legal and judicial institutions tend to apply for such assistance and, hence, a self-selection mechanism is at work. To address this important issue, we compare two subsets of countries in our dataset—that is, those that receive legal and judicial reform projects from the World Bank and those that do not. First, there can be selection on the type of countries (IBRD or IDA) that become recipients of legal and judicial projects.¹⁸ To address this possibility, we collected data on all World Bank projects. The World Bank project database currently has information on 13,630 projects, initiated between 1947 and 2013; approximately 80% of these projects are closed, and the rest are still active. Note that the World Bank implements projects for other international organizations, such as the Global Environment Facility, so we coded two dummy variables to account for IDA and IBRD countries. The IDA dummy takes the value of 1 if a given project received funding from IDA, and 0 otherwise. Similarly, the IBRD dummy takes the value of 1 if a given project received funding from the IBRD, and 0 otherwise. The total numbers of all IBRD and IDA projects are approximately the same: 5909 and 6017, respectively. We then evaluated the likelihood of receiving a legal and judicial project based on the country type (IBRD vs IDA), and found that, for both types of countries, legal and judicial reforms constituted very similar shares of their project portfolios: 5.39% for IDA, and 4.35% for IBRD. Therefore, there is no evidence that one group of recipients is significantly more likely to undertake judicial reforms than the other.

Second, we examine whether recipients with more judicial independence selected themselves into World Bank projects requiring legal and judicial reform because it would be quite easy for them to comply with the terms of the agreement with the World Bank. An important consequence of such self-selection would be our inability to attribute any positive relationship between World Bank assistance and project performance to the “inducement” effect of aid. We conducted bivariate (difference of means) tests and found that countries that received legal and judicial projects were not significantly different from other aid recipients in terms of their de jure or de facto levels of judicial independence. The distributions of the key measures of the legal and judicial environment in recipient countries prior to project initiation are not different at conventional significance levels across the two values of the legal project dummy (coded as 1 for legal and judicial projects, and 0 for all other projects). The only result suggesting a statistically significant difference between countries with legal and judicial projects and other countries is based on the de facto measure of judicial independence—a positive and statistically significant difference between countries with legal and judicial projects and other countries indicates that the first group has a slightly more independent judiciary. The substantive significance, however, is negligible. In sum, countries applying for legal and judicial projects do not differ from other aid recipients in terms of their legal and judicial environment at project initiation.

Finally, we compare the size of World Bank aid projects requiring legal and judicial reform with that of the other projects. We find that legal and judicial projects are not part of larger lending arrangements: legal and judicial projects receive, on average, less financing than other types of projects. Table 5 summarizes World Bank assistance for these two subsets of projects, based on: (1) overall assistance; (2) IDA assistance; and (3) IBRD assistance. In all three cases, legal and judicial projects receive significantly less aid than other types of projects. For instance, an average legal and judicial project receives approximately US$57 million in assistance, while an average aid commitment associated with other types of projects is around US$73 million. Thus, recipients of legal and judicial projects are not lured by the promise of outsized inflows of aid, in exchange for the agreement to conduct reforms.

Table 5.

Summary statistics of aid allocations for legal and judicial and other types of projects

	Observations	Mean	Standard deviation	Minimum	Maximum
Other types of projects
Total WB aid	11,017	72.941	134.170	0.1	3750
IBRD aid	11,017	50.542	131.029	0	3750
IDA aid	11,017	22.400	51.366	0	1200
Legal and judicial projects
Total WB aid	565	56.842	94.865	0.5	750
IBRD aid	565	41.479	95.489	0	750
IDA aid	565	15.363	31.235	0	300

Conclusion

International institutions play an increasingly large role in promoting judicial and legal reform in developing countries around the world. Yet, most analyses of judicial reform are confined to the strategic choices domestic political actors make. This article aims to bridge this gap by exploring the strategic interaction among the World Bank, national governments and non-governmental actors. Taken together, the theoretical framework yields a series of intuitively appealing predictions about the conditions under which the Bank is likely to suspend aid, the conditions under which NGAs are likely to participate in reforms, and the likelihood of reform success among different types of governments. Using original systematic data on World Bank projects, the empirical evidence presented in the article is preliminary, but promising.

The amount of World Bank aid clearly affects international and domestic actors alike. Increasing aid volumes raise the threshold for domestic consensus; hence, suspension is more likely at higher levels of aid. At the same time, increasing aid may also induce better performance on the part of some types of national governments. Specifically, the larger the amount of resources at stake, the more support for reforms NGAs demand from governments. In turn, the higher the amount of aid, the more governments stand to lose by not complying.

Our findings raise a deeper question about the endurance of judicial reforms. In our model, the only stick that the Bank carries is the threat to suspend aid. Thus, once all of the aid is released, an obvious question is why judicial systems in governments that fall between the first and second cut points in the model do not simply revert to the status quo. One answer may lie in the emerging judicial politics literature on public support and legitimacy (Carruba, 2002; Staton, 2003; Vanberg, 2001). Thus, although the government’s initial support is contingent on the level of foreign assistance, once such reforms are relatively successful, the enhanced public reputation of the judiciary may help to sustain independence. In other words, when courts are weak or unpopular, international involvement may provide a “jump start” to the legitimacy-building process.

This, in turn, leads to two final implications. First, given the previously noted fact that the results of such reforms often take time to become manifest, the duration of the Bank’s assistance programs should be considerably longer in countries that are at a greater risk of reversion. Second, among such cases the Bank should be particularly anxious to claim success for its projects. Thus, there should be a greater discrepancy in such cases between the World Bank’s ex post assessment of project effectiveness and objective measures (such as Law and Order scores) than among countries that are at lower risk. These remain questions for future research.

Footnotes

Acknowledgements

We would like to thank our anonymous reviewers, the editor, Randy Stone, Eddie Malesky, seminar participants at the Watson Center for Conflict and Cooperation Seminar at the University of Rochester, and panelists at the 2005 annual meeting of the Midwest Political Science Association for their valuable feedback on previous versions of this article. We are also grateful to James Melton and Jeffrey Staton for sharing their data, and to Ann Marshall and Erica Owen for outstanding research assistance. All errors remain our own.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Notes

References

Burnside

Dollar

(2000) Aid, policies and growth. The American Economic Review 90(4): 847–868.

Carey

Shugart

(eds) (1998) Executive Decree Authority. New York: Cambridge University Press.

Carruba

(2002) On the evolution of adjudication and monitoring mechanisms. Paper presented at the annual meeting of the Southern Political Science Association.

Collier

Dollar

(2001) Can the world cut poverty in half? How policy reform and effective aid can meet international development goals. World Development 29(11): 1787–1802.

Cooter

Ginsburg

(1996) Comparative judicial discretion: An empirical test of economic models. International Review of Law and Economics 16: 245–313.

Dakolias

(1996) A strategy for judicial reform: The experience in Latin America. Virginia Journal of International Law 36(1): 168–231.

Dalgaard

Hansen

(2001) On aid, growth and good policies. The Journal of Development Studies 37(6): 17–42.

Epstein

Knight

(1998) The Choices Justices Make. Washington, DC: Congressional Quarterly Press.

Feld

Voigt

(2003) Economic growth and judicial independence: Cross-country evidence using a new set of indicators. European Journal of Political Economy 19(3): 497–527.

10.

Ferejohn

(1999) Independent judges, dependent judiciary: Explaining judicial independence. Southern California Law Review (72): 353–374.

11.

Ferejohn

Weingast

(1992) A positive theory of statutory interpretation. International Review of Law and Economics 12(2): 263–279.

12.

Guillaumont

Chauvet

(2001) Aid and performance: A reassessment. The Journal of Development Studies 37(6): 66–92.

13.

Hayo

Voigt

(2007) Explaining de facto judicial independence. International Review of Law and Economics 27: 269–290.

14.

Helmke

(2005) Courts under Constraints: Judges, Generals, and Presidents in Argentina. New York: Cambridge University Press.

15.

Heymann

Lundberg

(2004) Aiding or Abetting? World Bank and the 1997 Judicial Reform Project. Kennedy School of Government Case Program 1779.1.

16.

Iaryczower

Spiller

Tomassi

(2002) Judicial decision-making in unstable environments: Argentina 1935–1998. American Journal of Political Science 46(4): 699–716.

17.

Kaufmann

Kraay

Mastruzzi

(2009) Governance Matters VIII: Aggregate and Individual Governance Indicators, 1996–2008. World Bank Policy Research Working Paper 4798.

18.

Knack

(2001) Aid dependence and the quality of governance: Cross-country empirical tests. Southern Economic Journal 68(2): 310–329.

19.

Landes

Posner

(1975) The independent judiciary in an interest-group perspective. Journal of Law & Economics 18(3): 875–901.

20.

Levy

(1987a) Anticipated development assistance, temporary relief aid, and consumption behavior of low-income countries. The Economic Journal 97(386): 446–458.

21.

Levy

(1987b) Does concessionary aid lead to higher investment rates in low-income countries? The Review of Economics and Statistics 69(1): 152–156.

22.

Levy

(1988) Aid and growth in Sub-Saharan Africa: The recent experience. European Economic Review 32(9): 1777–1795.

23.

Linzer

Staton

(2011) A measurement model for synthesizing multiple comparative indicators: The case of judicial independence. Paper presented at the annual meeting of the American Political Science Association, Seattle, WA.

24.

Magaloni

Sánchez

(2001) Empowering courts as constitutional veto players: Presidential delegation and the New Mexican Supreme Court. Paper presented at the annual meeting of the American Political Science Association, San Francisco, CA.

25.

McCubbins

Schwartz

(1984) Congressional oversight overlooked: Police patrols versus fire alarms. American Journal of Political Science 28(1): 165–179.

26.

Melton

Ginsburg

(2012) Does De Jure Judicial Independence Really Matter? A Reevaluation of Explanations for Judicial Independence. University of Chicago Institute for Law and Economics, Olin Research Paper 612.

27.

Mosley

Hudson

Horrell

(1987) Aid, the public sector and the market in less developed countries. The Economic Journal 97(387): 616–641.

28.

Murrell

(2002) The interactions of donors, contractors, and recipients. In: Martens

Mummert

Murrell

Seabright

(eds) The Institutional Economics of Foreign Aid. Cambridge: Cambridge University Press.

29.

Ramseyer

(1994) The puzzling (in)dependence of courts: A comparative approach. Journal of Legal Studies 23(2): 721–747.

30.

Rios-Figueroa

(2007) Fragmentation of power and the emergence of an effective judiciary in Mexico, 1994–2002. Latin American Politics and Society 49(1): 31–57.

31.

Rogers

(2001) Information and judicial review: A signaling game of legislative–judicial interaction. American Journal of Political Science 45(1): 84–99.

32.

Salzberger

(1993) A positive analysis of the doctrine of the separation of powers, or: Why do we have an independent judiciary? International Review of Law and Economics 13(4): 349–379.

33.

Spiller

Gely

(1990) A rational choice theory of Supreme Court statutory decisions with applications to the State Farm and Grove City cases. Journal of Law, Economics, and Organization 6(2): 263–300.

34.

Staton

(2003) It can happen in a hurry: Rational learning about judicial legitimacy. Paper presented at the annual meeting of the American Political Science Association.

35.

Stephenson

(2003a) “When the devil turns …”: The political foundations of independent judicial review. Journal of Legal Studies 32(1): 59–89.

36.

Stephenson

(2003b) Judicial Independence: What it is, How it can be Measured, why it Occurs. Available at: http://www1.worldbank.org/publicsector/legal/judicialindependence.htm.

37.

Stephenson

(2004) Court of public opinion government accountability and judicial power. Journal of Law, Economics, and Organization 20(2): 379–399.

38.

Stokes

(2001) Mandates and Democracy: Neoliberalism by Surprise. New York: Cambridge University Press.

39.

Svensson

(1999) Aid, growth and democracy. Economics and Politics 11(3): 275–297.

40.

Tsebelis

(2002) Veto Players: How Political Institutions Work. New York: Russell Sage Foundation.

41.

Vanberg

(2001) Legislative–executive relations: A game theoretic approach. American Journal of Political Science 45(2): 346–361.

42.

Weingast

(1997) The political foundations of democracy and the rule of law. American Political Science Review 91(2): 245–263.

43.

World Bank (2003) Legal and Judicial Reform: Strategic Directions. Available at: http://www4.worldbank.org/legal/leglr/publications_ljr.html

44.

World Bank (2004) Initiatives in Legal and Judicial Reform. Available at: http://www4.worldbank.org/legal/leglr/publications_ljr.html