Estimating the correlation between autocratic government (let alone its causal effect) and political corruption has produced mixed results. Kleptocratic autocracies are legendary: Robert Mugabe’s Zimbabwe and Ferdinand Marcos’s Philippines are well-known examples. Equally familiar is the case of Singapore, for over half a century a “clean” autocracy, as are (relatively) many Middle Eastern monarchies. But while most of the least corrupt countries are democratic, and most of the most corrupt are autocratic, in many cases experiments with democratization have failed to reduce high corruption. Nevertheless, despite the ease with which repression and co-optation facilitate autocratic rent extraction, Vineeta Yadav and Bumba Mukherjee argue that some autocrats pursue anticorruption quite vigorously, and that over time, corruption in autocracies is about as likely to decline as not. Which autocracies reduce corruption, where, when, and why are the questions that this book sets out to answer.

Yadav and Mukherjee find the answer to these questions concerning autocracies that reduce corruption in “SMEs,” small- and medium-sized enterprises with 15–250 employees, domestic private firms that exist outside of the business elite. These firms are also outside of conventional patronage networks, and this fact, when SMEs are organized nationally, constitutes the demand side for anticorruption. That reducing patronage would be to the greatest economic advantage of these outsider firms is easily assumed, though the exact mechanisms through which this happens are never made exactly clear.

When exactly are SMEs able to form a national organization? According to the authors, when they are sufficiently geographically concentrated to facilitate organizing, electioneering, and monitoring. What then induces the supply of anticorruption? A directly elected multiparty legislature is the answer: At least as long as opposition parties have some clout, enough to give effective articulation to SMEs’ anticorruption demands, the combination of this kind of legislature and an SME national organization can compel autocrats to concede anticorruption. The logic of combining these two types of institutionalization is that SMEs have greater organizational capacity than do individual citizens (with whom they share the burden of anticorruption), and thus can more effectively combine with opposition parties to enforce some accountability on the regime.

The empirical argument is supported by a combination of quantitative analysis and three case studies. These include Jordan (where there always was a national SME group and the legislature emerged in 1989), and Malaysia (where there always was a legislature but SMEs became rapidly more concentrated in the mid-1990s), which serve as examples of cases where both factors are present and corruption declined. Conversely, Uganda serves as a case where neither factor is present and corruption remained. Corruption is defined in the usual way as the abuse of public office for private gain. From the literature we gather that this includes everything from embezzlement and patronage to tolerating demands for bribes by regime-appointed officials. Corruption is measured by expert ratings provided by the International Country Risk Group (ICRG, the longest-running series and therefore the workhorse in most of the analysis), Transparency International’s Corruption Perceptions Index, and a third “poll of polls” indicator from the World Bank. All three produce correlated results, though over different countries and time periods. The study covers 86 (listed in Table 1.1, p. 4) or 95 (frequent in-text references) countries from 1985 to 2010, though complete time series are available in slightly less than half the cases.

The data are set out, described, and calculated as needed with admirable clarity in Chapters 2–4. The attitudes of authoritarian-country SMEs are analyzed in a 27-country World Bank survey: Corruption is a more severe problem for private SMEs than for state-run firms or larger private firms. We get a solid description of where more and less empowered legislatures exist and why and how they can be consequential. We discover that national SME organizations exist in about half of the authoritarian countries and that the geographical dispersion of SMEs (measured at the provincial level and aggregated to the nation) is (along with per capita GDP) a—or perhaps the—most significant predictor of whether or not a national SME association emerges.

Yadav and Mukherjee illustrate their calculation, however, with a map of Iran, which shows that the SMEs there generate a high concentration score (0.77) by being in a small arc around Tehran, surrounded by mountains and deserts, in provinces where state-run industrialization projects are also common. It is easy to see that interaction among firms is facilitated by geography, but the example feels odd because while we get a mention of a brief anticorruption moment in the late 1990s, corruption in Iran remains high. To be consistent with the book’s argument, I guess this is because the legislature remains controlled by the Supreme Leader, or perhaps because the Revolutionary Guard penetrates most firms of any size. In any event, I feel as though an opportunity to get inside the mechanism of change in corruption, or lack of, is missed.

Chapters 5–8 provide the analysis of the main hypothesis, that the combination of a national SME association and an elected multiparty legislature reduces the level of corruption. This puts the burden of inference on an interaction term in a regression setup. The statistical results in Chapter 5 persuade me that the hypothesis is borne out in the data: Many different specifications are presented to account (to some extent) for selection, unobserved heterogeneity, and the like. Time after time the interaction is significant and negative; no matter how you slice or dice the underlying assumptions about the specification, the marginal effect of the interaction has about the same magnitude. When data sets cover different periods, the size of the effect varies, but the sign is always the same. It is not perfect: Several times crucial figures either have misleading legends (6.6, p. 168) or are simply misprinted (7.3 and 7.4 are identical, but said to be different; see p. 182). Nevertheless, I do believe that on average, corruption is lower when a national SME group and a directly elected multiparty legislature jointly confront an autocratic regime.

What exactly does this mean? As so often, the devil is in the details. Are not both of these two aspects of institutionalization more likely to appear in similar circumstances? We know there is one, wealth, that appears in both institutional selection equations (Table 5.6, p. 134). But there are others. For one thing, when a country has an elected multiparty legislature in which the opposition parties are at all effective, it is more of what has come to be called an “anocracy” (a “mixed” polity, with some grafted-on characteristics of a democracy) than it is an autocracy. Polity scores emphasize this: Malaysia is a steady 0.4 through the period, rising to 0.6; Jordan was a –1 until the legislature appeared, and a mid-range –0.2 thereafter.

What happens in an anocracy is that electoral politics, even in small doses, injects uncertainty into the underlying coalition bargain that supports an autocrat, in this case with the business elite. And the uncertainty is consequential, because while Yadav and Mukherjee assume that the continuation of corrupt arrangements is the autocrat’s best strategy going forward, we can no longer be so sure. If the SMEs are main economic drivers of the country (the authors say something like this on p. 17 and elsewhere), then in any bargain they risk being co-opted into the autocrat’s support, as appears to happen in some cases discussed in the book. The institutional combination appeared briefly in Peru (p. 74) and briefly there was anticorruption, and then there was Alberto Fujimori and his Intelligence chief Vladimiro Montesinos! The decline in corruption in Jordan after 1989 (new legislature, opposition majority for the only time in the book’s cases) was minimal, though much bigger after 1999 (with a new king and a whole new support problem to solve). But even the existence of a bargain involving SMEs is problematic, and maybe why even a partially effective legislature was slow to appear in Uganda: It has an ongoing degree of ethnic fractionalization much higher than that of either Malaysia or Jordan.

Despite these quibbles, The Politics of Corruption in Dictatorships is essential reading for corruption scholars. There is fine work on the main hypotheses, and a host of important detail for those motivated to find and use.