The Case and Arguments

Nielson and Tierney's case study begins with a campaign launched in the early 1980s by a set of environmental nongovernmental organizations (NGOs) seeking to force the Bank to reform its environmental behavior amidst evidence that the Bank had funded a set of projects that resulted in extensive environmental degradation.²

One example was the infamous Polonoroeste project in Brazil, where more than $450 million in loans to promote road-building and agricultural development in the state of Rondônia triggered a massive migration of colonists that overwhelmed the Bank's efforts. According to Rich and others,³ this migration resulted in slash and burn agriculture that in turn caused an enormous amount of deforestation, while malaria, measles, and influenza epidemics spread throughout the indigenous populations and settlers.

Nielson and Tierney recount how the Bank avoided taking meaningful action until NGO allies in the U.S. Congress threatened to withhold the U.S. contribution to the World Bank's International Development Agency (IDA) replenishment.⁴

Yet a variety of reforms undertaken by the Bank, including a major 1987 reorganization and expansion of its environmental staff, failed to satisfy environmentalists and the Bank's major shareholders as more evidence of the Bank's lax behavior surfaced. Two widely cited examples include a 1992 independent study commissioned by the Bank on its controversial Sardar Sarovar dam project in India, which found the Bank was not enforcing its own environmental and resettlement policies. The study concluded that “There appears to be an institutional numbness at the Bank and in India to environmental matters.”⁵ An in-house report reviewing the quality of the Bank's portfolio, meanwhile, found extensive deterioration, with the number of “unsatisfactory” projects increasing from 15 percent of the sample reviewed in fiscal 1981 to 37.5 percent in fiscal 1991.⁶

Nielson and Tierney argue that the Bank adopted a set of “sweeping institutional reforms” in 1994, reflecting convergence of environmental preferences among the Bank's major shareholders.⁷

The Bank's response was to “significantly” alter its portfolio “by increasing environmental lending and decreasing projects that caused environmental harm.”⁸ The reforms cited by the authors included fire-alarm mechanisms such as the creation of an inspection panel to investigate claims against Bank activities by private groups who believe the Bank has not followed its procedures; increased reporting requirements, such as new public information documents; the requirement of two distinct environmental assessments for projects that are environmentally sensitive;⁹ and a dramatic increase in environmental staff and their location within the Bank.¹⁰

Nielson and Tierney then assess whether the Bank's environmental “lending behavior” has changed, reflecting principal demands, by analyzing board-approved projects between 1980 and 2000, specifically the percentages of projects in different lending categories and percentages of project dollars in different lending categories. They use a three-year rolling average for both measures, to smooth out fluctuations and to control for the increase in number and size of Bank projects over time. They look for increases in lending for projects that “actively attempt to preserve and improve the environment,” such as projects with primary goals of alleviating pollution or improving environmental infrastructure, and a decline in “projects that can harm the environment, particularly ‘traditional’ lending in the areas of energy/electricity, industry, oil/gas exploration, transportation and urban development”¹¹

Loans were coded as “either primarily environmental or not.”¹² In terms of dollars lent, the authors conclude that the increase in the environmental percentage of Bank lending is statistically significant at the 0.01 level for 1994–2000. However, they also observe that stand-alone environmental dollars declined after 1996, contrary to their expectations. They note that this decline is flattened by the inclusion funds from the Global Environment Facility (GEF), a separate facility funded by donor governments to offer grants to poor countries for projects addressing global environmental problems. The World Bank, United Nations Development Programme (UNDP), and United Nations Environment Programme (UNEP) are the GEF's implementing agencies. As for stand-alone environmental projects, the authors found an increase from 4.6 percent from 1980–86 to 7.3 percent for 1994–2000; the addition of Bank-managed GEF projects brings the increase to 14.9 percent. They conclude that these changes are “punctuated and significant.”¹³

Measuring Environmental Lending and “Behavior”

Nielson and Tierney's statistical measurement and analysis of Bank lending are problematic in several areas, revealing, in part, that precision on these issues is difficult. First, the categories used to code projects are too simplistic and do not correctly distinguish areas where the portfolio may have changed to reflect greater attention to environmental issues. Including in the category of projects that “can harm the environment” projects and lending dollars in energy/electricity, transportation, and urban development ignores the fact that many energy, transportation, and urban development projects have significant environmental components. Energy projects may improve energy conservation and efficiency, transportation projects may emphasize road maintenance or public transportation versus building new highways, and urban development may include projects to reduce air and water pollution.

The reverse is also true. While projects categorized by the authors as environmentally harmful may, in fact, have positive environmental components, projects categorized as environmentally beneficial may be environmentally harmful because there is no way to know from their coding whether projects classified as “stand-alone environment” have actually done a poor job of following Bank policies and have resulted in environmental degradation. The authors implicitly recognize that projects appearing to be environmentally friendly may indeed cause environmental degradation on the ground, and that is why they exclude forestry projects from their data set. World Bank forestry projects have generated controversy in cases where they were supposed to improve the environment but did the opposite.¹⁴

This choice to remove forestry projects implies that implementation outcomes must be taken into account when coding loans. At best, the authors appear to be measuring some limited and distorted reflection of Bank intention through types and lending amounts of a specific set of projects, but this will not reflect behavior.

An important exercise, then, is to more clearly define the complex issues involved in determining and measuring what World Bank environmental “behavior” does and does not mean. At minimum, the definition must address whether internal Bank policies have influenced project design, even in so-called traditional projects. The Bank itself has moved away from stand-alone environmental projects and turned its attention to “mainstreaming” environmental safeguards into its work. The definition must also take into account what the Bank actually does on the ground, although it is also true that many of the things than can go wrong in the implementation process may not reflect agent shirking. Nonetheless, projects that attempt to improve the environment on paper, but are cancelled or implemented in such a way that they are environmentally damaging, do not reflect significant positive changes in the Bank's lending portfolio.

The challenge of accurate measurement raises the issue of data sources. The authors do not provide the source(s) for their data, other than noting that it includes board-approved projects, and mentioning that environmental line-items are included in pre-implementation loan appraisal documents. Pre-implementation data, including the most widely used Bank data—figures from its annual reports on board-approved loans—do not reflect what happened after board approval, which includes the critical information on whether projects were cancelled, partially implemented, or poorly implemented. Loan commitment, disbursement, cancellation, and implementation data are all quite different. But even if one focuses on environmental components and objectives in board-approved loans as a partial explanation for improvement in environmental behavior, without looking at implementation, the Bank itself has noted that such components and objectives were at a “historic low” of 5 percent of the portfolio in fiscal 2002, down from 14 percent in 1993.¹⁵

Ultimately, if one does not know what happened with Bank loans in project design and implementation, one cannot tell whether pressure from Bank principals reined in Bank officials. Focusing on intention rather than analyzing behavior or performance, then, cannot reveal whether agents take seriously principal attempts to tighten up contracting and oversight tools, or whether they are simply jumping through necessary bureaucratic hoops in a window-dressing exercise. Indeed, perhaps the most common criticism of the Bank from NGOs is that there is a large gap between the Bank's rhetoric and intentions and on-the-ground reality.

The authors' finding that it is the inclusion of GEF dollars and projects that notably boosts their measurement of environmental dollars and projects also does not support their central argument that improvement in environmental lending since 1994 reflects actions taken by principals to rein in agents. Even though the World Bank manages more than half of the GEF's funds, the Bank's actions in this area show that Bank environmental lending is boosted by a separate grant-driven mechanism, rather than new oversight and monitoring mechanisms for all of its work. Indeed, GEF grant components that are “mainstreamed” into Bank projects follow existing Bank policies and procedures. In fact, Bank critics have argued in the past that GEF grants allow the Bank to “buy environmental respectability” while continuing a business-as-usual approach.¹⁶

In sum, at the end of Nielson and Tierney's article, one is no closer to understanding whether significant environmental reform has occurred at the Bank, and if so, why.

Environmental Reform at the World Bank

Another problem with the article is that the puzzle—explaining major reforms in 1994 or 1993–94 that resulted in significant changes in the Bank's behavior—starts with a mistaken premise. This period of time was not a benchmark in terms of Bank reform, and the changes in lending behavior are not as definitive as the authors assert. Consider an alternative perspective: that environmental change at the Bank is a relatively slow, ongoing process, with forward steps often followed by backward or sideways steps.¹⁷

Important “forward steps” include the major organizational changes of 1987, when the Bank created a central environmental department plus six regional technical offices to act as “environmental watchdogs” over its projects; the development of safeguard policies such as environmental assessment procedures in 1989 (revised in 1991 and 1999), to avert or mitigate a project's possible adverse environmental impacts; focused efforts to promote loans with primary or significant environmental components; and other strategies to help member countries develop environmental institutions, policies and programs.¹⁸ The oversight mechanisms created to monitor Bank staff behavior since the mid-1990s include the internal Quality Assurance Group (1996) to assess the quality of projects, supervision, and analytical work; and a new Quality Assurance and Compliance Unit (2000) to oversee Bank staff compliance with safeguard policies. Finally, the Bank has raised its profile in global environmental governance through its role as implementing agency for the GEF (established in 1991) and the Montreal Protocol's Multilateral Fund (established in 1990), in addition to other activities. One important change that undeniably did occur in 1994 was the Bank's adoption of a public information disclosure policy, which makes available a variety of documents that were once confidential.

Yet few of these reforms have escaped criticism. With respect to the post-1994 changes, there has been internal and external criticism of the safeguard policies for their sometimes limited impact on project design. Indeed, a 1996 internal Bank report noted that “very few EAs [environmental assessments] influence project design”¹⁹

Although the Inspection Panel (created in 1993, and beginning operations in 1994) is widely seen as an important step in improving Bank accountability, it is not yet the successful oversight tool envisioned, given the mixed reviews it has received regarding its impact on Bank lending behavior. Fox's analysis of the Panel's performance, for example, shows the Panel has seen limited results and ambiguous impact, resulting in “limited leverage” in changing Bank behavior. Obstacles include the fact that the Panel serves at the Board's discretion; that the Board can reject its recommendations; and that many factors complicate the ability of civil society actors to file claims.²⁰

It may be the case the reforms such as the Inspection Panel will gain strength over time and have a clear impact on lending behavior, but the link between reform and a reduction in the gap between Bank intention and activities must be explained and not assumed. The authors recognize some “conspicuous absences” in Bank reform, reflected in ongoing frustration of Bank critics, but this observation is a minor point in the article and does not detract from the central assertion that Bank behavior has changed significantly.

I would argue that backward or sideways steps include the organizational restructuring resulting from Bank president James Wolfensohn's 1996 Strategic Compact, aimed at improving Bank performance. The restructuring shifted power within the Bank to country directors in charge of country budgets. The move was applauded as a means to increase the Bank's accountability to client countries, but the fact is that many recipient countries and country directors see environmental issues as relatively lower priorities than they may be for some of the Bank's major shareholders. The result has been country directors calling less often on environmental staff for projects featuring major environmental goals and continuing mixed performance with respect to the implementation of environmental policies. As the Bank's Operations Evaluation Department (OED) noted in 2002, “Bank management, concerned with an ever-growing development agenda, has not been consistent in its commitment to the environment; and managers have not been held strictly accountable for complying with the Bank's environmental policies.” It concluded that, “The modest extent of mainstreaming the environment into the Bank's overall program is disturbing”²¹

Implications for Using P-A Models to Explain IO Behavior

The challenges Nielson and Tierney face in applying principal-agent models to the environmental behavior of the World Bank highlight two important areas for future research and debate. The first is whether or not P-A models can explain the puzzle of the absence or unevenness of expected change at the Bank, despite efforts by principals to institute reforms seeking to improve environmental performance. One approach, which I have argued elsewhere, is to focus more attention on problems with the delegation side of the relationship.²²

Delegation problems occur when states regularly asks IOs to take on conflicting and complex activities that are difficult to institutionalize and implement. This means that performance problems may not simply reflect errant institutional agents, but may find their roots in the struggle IOs face in confronting mission creep, and trying to implement goals that are difficult to specify or juggle. In the case of the World Bank, acting as financial institution and development agency, while balancing environmental issues with a multitude of other goals, both contribute to mixed efforts to address environmental issues. Another, more cynical approach, might look for evidence that the member state principals were never really serious about pushing for deep reforms within the Bank—that the “gap” between mandate and performance is not a real one.

Second is the challenge of more precisely defining IO “behavior” within a P-A model. In the field of international relations most scholars applying P-A relationships to IOs focus on the relationship between member state principals to IO agents. This angle is helpful in analyzing why states choose to delegate or how IOs gain autonomy.²³

Yet stopping the chain of delegation, and hence analysis of performance, with the organization means there is no way to explain persistent IO performance problems that do not directly reflect IO autonomy as an agent. While IOs are characterized by a set of collective principals that may create opportunities for institutional autonomy, the recognition that IOs themselves may also be usefully analyzed as principals when they enter into contracts with recipient countries reveals additional opportunities for agency losses.

International relations scholars may be uncomfortable with conceptualizing the relationship between an IO such as the World Bank and a recipient country within a P-A framework. After all, one may argue that the recipient country is a client, not an employee, or that the country, not the Bank, has the final authority to build a new power plant or highway. Nonetheless, the relationship between international financial institutions (IFIs) and loan recipients is a close enough fit to the P-A model to usefully illuminate many of the challenges the IFIs face in providing incentives for countries to agree to and implement loan conditionality, and as such deserves greater attention in the IO literature. P-A models, after all, are commonly used to explain the relationship between lenders and borrowers. Lenders use screening, monitoring, and other tools to mitigate against problems of moral hazard and adverse selection endemic to any P-A relationship where information and interests are asymmetric. In the case of the International Monetary Fund (IMF) and World Bank, the organizations use conditionality in loan agreements to further donor objectives, which may be financial, economic, legal, environmental, and so on. Loan disbursements may be stopped when countries are not fulfilling the agreed-on terms.

In fact, a number of scholars within the field of development economics use P-A models to explain poor performance by major aid organizations such as the IMF and World Bank, by framing the problem as one between the IFI as principal to the recipient country as agent in an explicit effort to explain where and why IFI conditionality fails to elicit the expected behavior from the latter.²⁴

Within this literature, scholars have also explored the ways in which the IFI-recipient relationship differs from standard lending, or other private contractual, relationships. For example, it is often politically difficult for an IFI to punish a recipient country not keeping its side of the conditionality bargain.²⁵ Rich has argued that in the case of the World Bank, there has been institutional complicity in corruption in large borrowers such as Russia and Indonesia, where the Bank has failed to cut off disbursements amidst evidence that the borrower has violated major conditions.²⁶ There is also an unusual twist with IFIs, given the fact that recipient countries are also principals because they are represented on the institutions' board of directors. Some IMF officials find the analysis of recipient countries as agents of IFI principals to be politically awkward, given that such a representation hardly conforms with the idea IFIs are touting of recipient countries being in the “driver's seat” of their country programs.²⁷

The question of who exactly is the IFI's agent is also not always clear. World Bank loans, for example, are often disbursed through different actors, including the relevant recipient country ministry, the local government authority, the specific organization for whom the aid is intended, and so on.²⁸

Finally, P-A models are not equipped to identify myriad sources of implementation problems in recipient countries. World Bank project implementation, for example, may be hampered by weak local institutions, overly ambitious expectations, problems with supervision, disbursement, management, and so on. These problems may or may not reflect agency slack. These complexities are all areas for future research aimed at producing more precise explanations for IO performance. For example, careful identification of the source of implementation problems may point attention to whether the problem lies somewhere within the IFI or the recipient country, and whether the cause may or may not be traced to agency slack. Analysis of member states as principals to IO agents may highlight a different set of problems than analysis of IOs as principals to member state recipients.

There is great scope for using P-A models to locate specific areas where performance pathologies occur along the chain of delegation, recognizing the dual role the IO may play as both agent and principal. For example, if an institution is designed to be highly autonomous, one should not assume that significant behavioral change will follow pressure from major member state principals.²⁹

Alternatively, an institution that has made major efforts to institutionalize reforms may find it is thwarted by incentive failures within the aid delivery process.³⁰ P-A relationships need not encompass every step of delegation between a member state and the actor(s) implementing a loan, but there is much work to be done in identifying the major sets of relationships and assessing their distinctive impact on IO activities.

IOs virtually across the board are under fire for a variety of delegation and performance problems. The success or failure of these organizations is dependent, in large part, on how complex P-A relationships are crafted and managed. Nielson and Tierney open the door to considering how these models may be applied to understanding the performance of the World Bank in particular, and pushing the analytical envelope on some of the issues raised will contribute to understanding how agency models offer a powerful tool for understanding IO behavior and performance. No other theoretical approach to the study of IOs is flexible enough to examine the interactions among the politics and preferences of member states, the internal incentives facing IO staff, and recipient country actors involved in IO activities. P-A models, properly applied, can greatly advance rationalist approaches to understanding IOs and their performance.