Covadonga Meseguer offers what very few others have been able to: a quantitative analysis of the policy learning process that tests the relationship between changing beliefs and the choices countries make about market-oriented reform policies. Using Bayesian analysis and probit, she looks for evidence of “rational learning” based on economic growth information in the adoption of four policies: export-oriented development strategies, privatization, the removal of controls for capital mobility, and agreements entered into with the International Monetary Fund (IMF). All of these policies are “aimed at opening up national economies and reducing the role of the state in the economy” (p. 1) and represent a convergence on prevailing models of development, the so-called Washington Consensus. While learning from the success of these policies has often been invoked in the literature to account for this convergence, evidence of actual learning—decisions based on a change in beliefs about the effects of economic policies—has been scarce. Hence, Meseguer's question: “[D]id governments switch to market oriented policies as a result of a learning process?” (p. 3).
The political economy literature, however, offers a number of other explanations to account for this convergence, including democratization and increased responsiveness to the employment demands of the “median voter.” Alternatively, the reform policies might have been externally imposed as part of the conditions required by an international financial institution such as the IMF (or adopted voluntarily by a government eager to shift responsibility for making the changes associated with such conditions). Or the government might have merely emulated others in adopting the change “not because it has been understood to be superior, but because the sheer number of others adopting it makes it good, and enhances its legitimacy” (p. 228).
Meseguer tackles the problem of measuring learning in a sophisticated and convincing way. She operationalizes learning as a process by which governments can “observe their own and others' past experience with policies, and update their beliefs on the basis of these observations. The combination of prior beliefs and available data using Bayes's rule provides posterior beliefs about economic results. I hypothesize that governments choose policies on the basis of these posterior beliefs, which embody learning from previous experience. Period after period, new information is gathered, new posteriors are derived and a new policy choice is made on the basis of what has been learned” (p. 39).
For each of the four policies, the research proceeds in two steps. First, Meseguer determines for each country and for each year that country's expected growth under the policy in question based on its own experience, experience in the region, and experience in the world. She then updates each of these prior beliefs on the basis of the next year's experiences to establish posterior beliefs about growth under the policy for each new year. Average expectations and variances are both calculated, the latter to reflect risk and to test for risk aversion. Secondly, in a probit analysis she uses these expectations, along with controls for alternative hypotheses, as independent variables to determine whether these posterior beliefs can explain the decisions to adopt or sustain each of the four policies.
The results of the analysis are mixed. For some of the policies, the evidence supports a significant role for learning, while for others learning appears to play little if any part. The results for adopting an export-oriented development strategy (shifting from an internal industrial protection approach) illustrate this mix. Only posterior beliefs about the country's own performance are significant predictors of the shift, and as hypothesized, the variability of the country's own performance and that of the world is negatively and significantly related to the shift. When alternative convergence mechanisms are included in the probit model, emulation or herding becomes significant, as does the presumed coercive effect of being under IMF agreements. “All in all,” the author explains, “the greatest effect on the probability of adopting [an export orientation] had to do with the lessons learned from the unreliability of [import substitution industrialization] as a strategy to promote growth” (p. 103).
For the other policy groups, the results are equally complex. Privatization decisions appear to be significantly affected by the updated beliefs about one's own performance, and by regional and world results, reflecting some of the strongest and clearest evidence for learning in the study. When the control variables are included in the model, emulation again emerges as significant, and regional effects appear for the Latin America subsample. For policies surrounding capital account liberalization, the learning effects are less marked. Only the average results from the world appear to have a consistent effect on adopting and sustaining the removal of capital controls. Emulation and the possibly coercive effects of IMF agreements again are significant. Finally, Meseguer considers the effects of learning on decisions to enter into an IMF agreement, with all the conditions that decision imposes on the state. These conditions include some of the polices already investigated, but the author looks at this decision separately to determine if the “IMF programs have costs other than the policies they prescribe” (p. 182). While there is some evidence of learning from world experience on the decision to enter into an IMF agreement, the record for growth under such agreements is relatively poor, and other domestic political and economic variables, including emulation, recent elections, and low reserves, were stronger influences.
Overall, Meseguer highlights the importance of learning from other countries in the world, not only, or in some cases instead of, learning from one's own experience or regional models. Regional influences were strong for the only real policy innovation considered, privatization. These are valuable findings that address persistent debates about the sources of market-oriented reform policies. The unevenness of the record for learning lends credibility to the analysis and gives us confidence in the author's ability to distinguish learning from nonlearning. Also consistently important in accounting for the policy choices, however, was emulation. She interprets this as indicating that emulation and learning are “not exclusive phenomena (p. 222) and can occur together.
While Meseguer distinguishes learning and emulation conceptually, they may become confounded in the analysis, however, especially as learning from world growth averages could generate the appearance of herding. This illustrates a persistent problem with learning analyses: distinguishing learning from other forms of change that do not depend on new causal inferences and cognitive change. Other issues, of which Meseguer is well aware, include focusing only on economic growth to predict policy changes. Political motivations may be equally important, but finding a reasonable metric for that motive has been difficult. Treating the policies, most of which actually involve multiple programmatic elements with diverse records and supporters, as single choices may result in the appearance of more policy convergence than actually exists. Similarly, the diversity of actors and viewpoints raises questions about who is actually learning.
Although these problems are not completely laid to rest, the fact that even some rational learning can be shown to occur makes Learning, Policy Making, and Market Reforms a valuable contribution both substantively and methodologically to the growing focus on policy and government learning. The detailed literature reviews of the controversies surrounding the policies and the mini case studies of the adoption of some of them make the book engaging and appropriate for students of political economy.
