Category Archives: Political Science

Illusions of Control

Democracy by Mistake

Daniel Treisman (UCLA)

Abstract: How does democracy emerge from authoritarian rule? Influential theories contend that incumbents deliberately choose to share or surrender power. They do so to prevent revolution, motivate citizens to fight wars, incentivize governments to provide public goods, outbid elite rivals, or limit factional violence. Examining the history of all democratizations since 1800, I show that such deliberate choice arguments may help explain up to one third of cases. In about two thirds, democratization occurred not because incumbent elites chose it but because, in trying to prevent it, they made mistakes that weakened their hold on power. Common mistakes include: calling elections or starting military conflicts, only to lose them; ignoring popular unrest and being overthrown; initiating limited reforms that get out of hand; and selecting a covert democrat as leader. These mistakes reflect well-known cognitive biases such as overconfidence and the illusion of control.


Circulated by NEP-HIS on: 2017-11-05

Revised by Thales Zamberlan Pereira

In his paper “Democracy by Mistake,” Daniel Treisman attempts to provide a new answer to the important question of how democracy emerges from authoritarian rule. Different from previous theories, which attribute the fall of non-democratic regimes to calculated decisions by those in power, Treisman argues that most episodes of democratization happened because dictators – like everyone else – are in fact bad at probabilities. Like all humans, but especially so given their circumstances, dictators are prone to overconfidence and the illusion of control. Living the life of confirmation bias, therefore, non-democratic regimes start to lose their grip on power making avoidable mistakes. Mistakes, of course, are only truly avoidable in hindsight, but the strength in Treisman’s paper relies on his argument that two thirds of his recorded cases of democratization do not fit the usual interpretation that dictators deliberately choose to share or surrender power when democracies start to emerge.

Chile after pinochet

Chile after Pinochet – Source:

In the democratization by choice category, Treisman presents six common arguments from the literature and divides them into three schools of thought – democracy by bargain, splits within ruling circles, and democracy as a peace-making device. The general argument, nonetheless, is that intentional democratization happens when the ruler weights his probabilities and concludes that if he does not reduce his personal power in the short run, he will have worse problems in the long run – e.g., the dictator will have no money to fight wars or will have people with torches at his front door.

In the democratization by “bad choice” category, the mistakes that lead to democratization are, in a simplified way, the following: ignoring warnings and getting overthrown by popular revolt; calling a referendum or election— and losing; initiating or entering a military conflict—  and losing; enacting partial reforms to stabilize the regime— but undermining it; selecting a leader to preserve the regime— who destroys it; and using repression counterproductively.

To divide cases of democratization between intentional and non-intentional, Treisman analyses 218 episodes since 1800 using two definitions of democracy. The first one is a binary concept, called “qualitative,” which uses data from Boix, Miller, and Rosato (2013).  In their definition, democracy exists in a country when elections are free and competitive, the head of government is either directly elected or answerable to an elected parliament, and at least half the male population has the right to vote. The second definition of democratization is called “directional”, and it uses the Polity IV database, which measures countries regimes on a 21-point scale – from -10 for hereditary monarchies to +10 in the case of consolidated democracies. For a country to be considered a democracy, the Polity IV establishes that it must have a score of at least 6.


Some people just can’t let it go

With these categories, how can we interpret the empirical results of the paper? Is it the case that most episodes of democratization happened because dictators, overconfident in their position, made critical mistakes that ultimately undermined their power? The answer, it seems, depends a lot on what the author considers as a mistake. For example, among the episodes of intentional democratization, the end of the military dictatorship in Brazil in 1985 is categorized as a “great compromise.” I agree that this is a good example of intentional democratization due to political compromise, however, one could argue the opposite using the framework presented in the paper. It is known in Brazilian political history that the last military president in Brazil did not begin his term planning to be the last. In fact, his indecision in choosing/supporting a successor led to the strengthening of civil groups demanding increased access to the government (Dimenstein 1985). Therefore, the democratization of Brazil after 1985 also started with a dictator who was overconfident he could maintain the status quo (and who also overestimated his own relative competence). However, is this sufficient to defend the hypothesis that democracy was the outcome of individual mistakes? What about the institutional environment that allowed such significant transformations to happen? It seems that “critical mistakes that undermine power” cannot be restricted to the cognitive biases of “great men.” The fact that democratic reforms are not intentional does not necessarily lead us to the “democracy by mistake” camp.

Another issue of using multiple categories is that sometimes it is not clear which examples are really being used as a mistake. Take the “initiating or entering a military conflict – and losing” category, which account for 6-9 percent of democratization processes, according to the paper. Among the examples is the well-known miscalculated attack by the Argentine military government on the Falkland Islands in 1982. However, Treisman also uses as an example for this category Paraguay’s attack on Brazil in 1864, which started the War of the Triple Alliance. It is strange that these two episodes are bundled together because the death of Solano López (the Paraguayan dictator) was not followed by a democratic government. In fact, the data that Treisman uses also don’t assign this period to a democratic transition: Polity IV only assigns democracy to Paraguay in 1994, and Boix et al. data indicates that Paraguay became democratic in 2003. Moreover, one must not forget that Solano attacked Brazilian officials right after Brazil deposed his allies in Uruguay and potentially ended his access to important continental rivers and the sea. This could be considered a critical mistake, but I couldn’t understand if the author is considering any mistake that weakens authoritarian rule as a valid example, or if his examples are only for when the mistakes turn into a democracy. Does Solano’s choice count as a “leap in the dark” to democracy or not?


Policy IV authority trend for Brazil (with comments)

Treisman asserts that intentionalist theories find weak support in historical cases, but using behavioral science as a mechanism to explain democratic transitions seems insufficient to explain transformations that usually are larger than individuals. Treisman makes arguments such as: “neurological evidence suggests power can impair the ability to process the actions and emotions of others” (p. 28), and “physical and mental deterioration affect[ing] leaders in all systems, they are more likely to impair decision making in autocracies.” (p. 29). Nonetheless, it is surprising that even today dictatorships seem stable in the eyes of those outside it – until the day they are no more. Angola and Zimbabwe are recent examples of this. It seems that there is always the illusion of control, even if dictators stay in power. Isn’t it the case that our mistakes inevitably turn into naive memories that, if not for one detail, make us think that everything could have been different?


Boix, Carles, Michael Miller, and Sebastian Rosato. (2013) “A Complete Data Set of Political Regimes, 1800–2007,” Comparative Political Studies, 46 (12), pp. 1523–1554.

Dimenstein, Gilberto. O Complô que elegeu Tancredo (1985). Rio de Janeiro: Editora JB.