Why do military dictatorships become presidential democracies? Mapping the democratic interests of autocratic regimes

An alternative way of categorizing autocracies, popularized by Geddes (1999), consists in distinguishing between personalist, military and single-party regimes. Historically, that distinction has been important in the transition from personalist to hereditary autocracy, and the consequent reduction in conflicts related to succession (Kurrild-Klitgaard 2000; Tullock 2001; Kokkonen and Sundell in press). In the following, most military dictatorships, as categorized by Cheibub et al. (2010) and Bjørnskov and Rode (in press) also would be categorized as military by Geddes (1999), and all communist regimes in the following likewise are single-party. The main difference is that civilian autocracies, in the present category, need not be personalist.

It is worth noting that the same transition appears in all regions of the world—it is thus not a Latin American phenomenon—and regardless of the country’s specific history. Bjørnskov and Rode (in press), for example, show that colonial ancestry (British, French or other colonial history) does not affect coup risks when countries become independent, although it does affect their likelihood of democratizing.

The single case is Peru in the 1990s, when the democratically elected President Alberto Fujimori disregarded the democratic institutions and eventually had to flee the country. After a de facto interim autocratic regime, the constitutional institutions proved robust enough to re-establish democracy without much ado.

In many cases, more than two parties exist or form around democratization. However, the assumption of two parties likely reflects reality when the party system either develops two dominant parties or two approximately coherent blocs of parties (the left and right), as appears to be the case (as documented in the online appendix). When the party system becomes characterized by lax party discipline, as in the United States, it might even form two identifiable groups of politicians across party lines with similar interests on issues of military interest. The assumption of two parties is likely to be problematic only in the case of multiple parties competing in a multi-dimensional issue space, or when the party system develops small pivot parties. Conversely, while an incumbent cannot know prior to democratization the extent to which the parliamentary system develops party discipline, assuming that it does appears to be a rationally conservative choice for both the incumbent government as well as for the present theoretical framework.

A further logical extension of this result is that the incumbent government may have a particular interest in choosing a policy mix that, for example, yields sufficient subsidies to a subset of industries—perhaps best thought of as what was once called the ‘commanding heights’ industries—while achieving the opposite for other industries. That policy mix can, for example, be one in which a subset of industries is protected by high import tariffs while other subsets that potentially could form an alternative majority are exposed to free trade. The mix would reduce their profit margins to the extent that they would have little interest in seeking subsidies—and, thus, represent a threat to the incumbent—because the coordination cost would be prohibitively high relative to the potential gain.

As noted above, in parliamentary systems with multiple parties in which coalitions are necessary to assemble a governing majority, partisan veto players can be effective. However, nothing assures that any constitutional provisions result in either a multiple-party system, a two-party system or a system with dominant parties. As such, presidentialism is the one certain constitutional choice with which to create institutional veto players.

It is worth noting that, theoretically, military competence [as in Eqs. (18) and (19)] may either strengthen or weaken the incentive to democratize, because the net effect depends on the coup risk when not democratizing relative to the risk of losing a democratic election.

An additional result here is that military competence makes it more likely that a presidential system is preferable whenever the condition \(\left( {1 - \mu } \right)\xi m_{0} < \Delta \left( {1 - \theta } \right)\left( {1 - \varphi } \right)m_{p}\) holds, essentially implying the intuitive condition that when the best military spending under autocracy is less than the best possible expected military spending under democracy, a more competent military regime will chose to democratize to a presidential system that minimizes its potential loss.

Conversely, many democratizations from civilian autocracy resulted in new or clearly amended constitutions drafted and introduced after the regime change: 30% introduced a new constitution that was negotiated after democratization. Among military dictatorships, the same pattern has been rare, with only five cases (8%)—Brazil in 1988, the Central African Republic in 1994, Greece in 1975, Poland in 1992 and Venezuela in 1961—in which drafting a new constitution was left to the new regime. The Brazilian case also differs owing to the special nature of the constitutional drafting process in which broad interests within the population were taken into formal account. Indeed, the particular way in which new constitutions were drafted also differs significantly across initially civilian and military regimes: 85% of the new constitutions of democratizing civilian autocracies were drafted by either the legislature or the executive branch, or with heavy influence from either. In military dictatorships that democratized, the drafting process of the more frequent new constitutions tended to be delegated to either a constitutional assembly appointed by the dictatorship or an elected special constitutional legislature assembly. Here, only 44% of the drafting processes included direct influence from either the executive or the legislature.

While it is extremely rare that the president can veto legislation formally, all but one of the military constitutions allow the president to initiate legislation. However, most constitutions also allow the parliament to do so, thereby effectively creating shared competencies. That feature is significantly rarer in democracies initiated by civilian autocracies. Conversely, the president in the former regime type almost never is given the right to dissolve parliament, while that feature is approximately twice as likely to be present in the constitution in the democracies initiated by civilian autocracies.

In addition, it may be worth emphasizing that presidential democracies do not appear more stable than parliamentary types. Although 66% of presidential democracies and only 58% of parliamentary democracies survived the first 10 years after democratization, the difference is far from being significant.