While democratization has long been a major theme in political science, recent years have seen a renewed interest in the topic of dictatorship. Jennifer Gandhi's book is a welcome addition to this literature. Gandhi argues that most autocrats need popular and elite support to remain in power. Operating through pure coercion is costly and risky. Dictators in her framework are not just concerned with extracting rents from society. They also have policy goals. Many powerful rulers thus accommodate representative political institutions to help them obtain information on public and elite opinion and to facilitate bargaining. They recognize that it is often better to include societal groups in the process of governing than to risk a turn to street protests, violence, or guerilla warfare. They may accommodate wealthy elites that have the capacity to exit for more promising opportunities abroad.

Gandhi begins by defining dictatorships as “regimes in which the rulers acquire power by means other than competitive elections” (p. 7). Her data set is the number of years each country has spent under dictatorship during the period 1946–2002. She distinguishes among monarchs, military strongmen, and civilian leaders. Her data include monarchies in the Middle East, military rulers in Latin America and Asia, autocratic socialist regimes in Europe before 1991, and autocratic regimes that emerged out of the Soviet Union's breakup after 1991.

The author categorizes legislative institutions as appointed, elected, or nonexistent in each country-year, and counts the number of political parties. She then creates a measure of political institutionalization that takes three values: two for a legislature with multiple parties, one for a legislature in which the regime party occupies all the seats, zero for all others. Thus, country-years are coded as zero if there is no legislature or if the legislature includes only members with no party affiliation. Not surprisingly, all things being equal, institutionalization tends to be highest under civilian dictators and lowest under monarchs. Other factors associated with lower levels of institutionalization are high levels of natural resources and of ethnic polarization.

There may be interactions among the independent variables, however. For example, the nature of the dictatorship might be connected to the existence of natural resources or to ethnic diversity. The maintenance of a monarchy could be a function of access to natural resources. Similarly, dictators with stronger policy goals might be more likely to permit or create stronger legislatures. In fact, this seems to be part of Gandhi's conclusion. She hypothesizes that dictators “make enough concessions in the form of rents and policies through institutions to insure that tenures remain secure” (p. 178), and she finds confirming evidence for this claim.

The effort to unpack and differentiate dictatorships is a welcome enterprise that helps one see that dictators are concerned with policy, not just with maintaining power for its own sake and with extracting rents. However, I have a fundamental conceptual problem with Gandhi's method. Her definition of a dictatorship produces a dichotomous variable that puts some countries on one side and some on the other. Then she demonstrates that some countries that she labels dictatorships, in fact, have robust multiparty legislatures. At what point do such rulers cease to be dictators in any meaningful sense? Her very definition of a dictator has created an either/or category out of a continuous or multidimensional variable. For example, although many of her civilian rulers were elected in noncompetitive elections, they must nonetheless satisfy legislatures with strong opposition parties. During what turns out to be the democratic transition, they may gradually accept or even encourage democratic developments. Conversely, civilian leaders elected in competitive elections may move toward dictatorship over time. The author counts those leaders as dictators from their date of election (p. 7, n. 7). Others elected in noncompetitive elections could likely survive a competitive election. She claims that “dictators are dictators because they cannot win competitive elections” (p. 81). But that counterfactual is not part of her definition. She simply asked if the election itself was noncompetitive.

Although Gandhi's statistical results and case studies are enlightening, the puzzle she sets for herself is less of a puzzle if one admits a continuum of government types from full democracy to full dictatorship, with a number of intermediate steps. One intermediate case is a ruler elected in an uncompetitive election who operates under constitutional constraints that produce a multiparty legislature and other institutions, such as a constitutional court, independently elected state governments, and a referendum process. Of course, some transitions occur in a short space of time, such as under a military coup, the disintegration of the Soviet Union, or the fall of the Berlin Wall. However, even in the USSR, the nature of the regime changed dramatically in the years before its disintegration, and in other countries, such as Mexico, the end of one party rule at the presidential level was preceded by growing signs of democratic strength elsewhere. In Peru, why code Alberto Fujimori's regime as a dictatorship from the time he won a hotly contested election and maintain Venezuela as a democracy until the end of the study period in 2002?

In short, I believe that it would be productive to distinguish between transitional regimes and established dictatorships. Can one predict the incidence of gradual versus abrupt transitions? When does a multiparty legislature signal an imminent shift to competitive elections for the head of state? In contrast, when is this type of legislature a stable way to accommodate societal pressures? Gandhi does deal with the survival rates of dictators in her final chapter. She predicts that survival is not tied to the strength of legislative institutions. These institutions are a choice variable for dictators, who will make concessions only when needed to promote their own survival. This is indeed what she finds. However, since many regimes do change during her study period for reasons other than death or disability, her dictators do not always get it right. There are 92 instances where a dictatorship, under her definition, shifts to a democracy; one of these dictatorships involves a monarch, 58 involve military leaders, and 33 involve civilian leaders. Other dictators lose power to other dictators, not always voluntarily. Some countries go in and out of the dictatorship category, as in the fluctuating regimes in Latin America. Others have made a one-time shift to what appears to be a stable democracy.

It would be interesting to compare cases where a gradual erosion of dictatorial powers leads to full democracy with cases where the shift is abrupt and perhaps violent or systemic. A more dynamic view of regime change and a move away from the simple dictatorship/nondictatorship dichotomy would add more nuance to the important set of issues that Gandhi raises inPolitical Institutions under Dictatorship. At the same time, she is to be commended for her effort to incorporate forms of bargaining and consent into the analysis of dictatorship. One hopes that in future work she will extend this analysis to take further account of hybrid and transitional regimes.