Thèse de doctorat
Résumé : In 2019, the NGO Freedom House entitled its annual report “Democracy in retreat”. This report marked the 13th consecutive year of decline in overall freedom in the World. Freedom House attributes this decline to “the previous wave of democratization rolling back”. Democracy is losing ground in several places: leaders using an authoritarian rhetoric are flourishing whereas popular support for democracy is shrinking (Mounk, 2018). This democratic recession questions the performance of democracies (Diamond, 2015). Previous research showed that democracy is associated with higher growth (Acemoglu et al., 2019), higher education (Gallego, 2010) and higher development (Gerring et al., 2012). Those assessments, however, do not explain the challenges new democratic countries currently face or why many democracies recently rolled back. Comparing the performance of democracies with that of autocracies is not new. Back in ancient Greece, Plato already debated the risks of democracy (Plato, The Republic, Book VIII). Nevertheless, measuring the virtues of political regimes in such static terms has drawbacks. It disregards the history of a political regime and the specificities of political transitions. A tautological argument is that democracy materializes after a democratic transition and autocracy often after an autocratic reversal. These transitions are multifaceted. They breed uncertainty and/or stem from violent outbursts (Acemoglu and Robinson, 2001). They open a period of political instability by fueling claims over power and over wealth in the new polity (Higley and Burton, 1989). New regimes have to face specific challenges: setting new rules, dealing with former elites, stabilizing the new regime and facing new contestants (Clague 1996). Those elements interact with the intrinsic characteristics of political regimes. In order to measure the merits and disadvantages of different regimes, it then seems adequate to understand first what political transitions bring about and second what causes them. This dissertation aims at offering insights on those two interrelated issues. To do so, this dissertation focuses on the instability, the violence and the political risk accompanying political transitions. Its primary objective is not to question the merits of a type of regime over another but to understand the features of political transitions. First, I investigate the consequences of democratic transitions on political risk, defined as a risk of political nature compromising economic activities – e.g. poor investment protections, corruption, and conflict. History illustrates well how transitions cause this kind of risk. For example, the Burundian civil war erupted after the first free and fair elections in the country in 1993. Political transitions also provide the new leaders with incentives to abuse their power to expropriate properties. Understanding political risk – as a correlate of political transitions – adds on the assessment of political regimes found in previous literature. Considering the interplay between political risk and political transitions also sheds light on a so far under-studied consequence of political transitions.Second, I turn to identifying the determinants of a specific dimension of political risk: regime reversals. Regime reversals often occurred in chaotic circumstances as in, 1642 England, 1973 Chile and 1999 Indonesia. The second part of the dissertation analyzes these chaotic circumstances from two different angles. It underlines the long-term structural changes fostering transitions to democracy. It also clarifies the individual reasons why elected politicians may oppose an autocratic reversal.1. The consequences of democratic transitionsPolitical transitions change the way policies are decided. By integrating new people in the decision-process, democratic transitions generate reforms (Giuliano et al, 2013). As a result, the political and economic structures of a country in transition change (Grosjean and Senik, 2011). These changes in a country transiting to democracy shape its development path. Papaioannou and Siourounis (2008a) and Acemoglu et al. (2019) present evidence that democratic transitions foster growth. Aidt and Jensen (2013) observe that they increase public spending. These attractive features of democratic transitions have to be balanced against their less desirable outcomes. Reforms and changing political systems also generate new opportunities for rent-seeking and corruption (Mohtadi and Roe, 2003). By the same token, they may relax property rights to foster redistribution (Clague, 1996). As a consequence, the economic gains from democratic transitions are likely limited and surely unevenly distributed. The first chapter contributes to this literature by assessing the variations of foreign investments before, during and after democratic transitions. It examines the impact of democratic transitions on the attractiveness of new democracies to foreign investors. Using an event-study method on a panel of 115 developing countries from 1970 to 2014, we find that shifting to a democratic regime does not affect Foreign Direct Investment inflows (FDI), on average. We do find, however, that consolidated democratic transitions, defined as those that do not go into reverse for at least five years, actually increase these inflows, with the bulk of the improvement appearing ten years after the transition. Furthermore, when controlling for political risk, the effect of consolidated democratic transitions appears immediately after they occur, suggesting that higher political risk in the early years of the new regime offsets the positive intrinsic effect on FDI. Political transitions, however, do not only impact the economy and how economic transitions operate. They also influence the way different political groups interact. This explains how political transitions affect a specific component of political risk: violence. Democratic transitions have an ambiguous impact on violence. On the one hand, new policies are more aligned with the preferences of the newly-enfranchised population, thus reducing violence (Reynal-Querol, 2002). On the other hand, democratic transitions reduce the de jure power of former elites. These elites may, as a consequence, use violence to compensate for their loss of de jure power (Acemoglu et al., 2008). The overall effect of democratic transitions depends on how they materialize (Cervelatti and Sunde, 2013, Finkel et al., 2015), how they raise the opportunity cost of violence (Collier et al., 2009) and how they change policies in place (Aumann et al., 1983).Chapter 2 directly contributes to this debate. The analysis centers on the influence of enfranchisement on the use of violence as a political resource. This chapter investigates how the Voting Rights Act (VRA) in 1965 United States impacted political violence. The VRA forbade discrimination in voting. Its coverage formula generated both geographic and temporal local discontinuities in its application. The empirical strategy takes advantage of these features by comparing the evolution of political violence in geographically close covered and non-covered counties. Difference-in-differences estimates indicate that the VRA coverage halved the incidence and the probability of the onset of political violence. Extensions also show that redistribution does not explain these dynamics. Instead, I report empirical evidence suggesting that voting became the new institutionalized way to express political preferences. After elections, electoral results are a credible signal of political preferences. Political violence is, as a result, no more needed to correct the lack of representativity of pre-enfranchisement electoral results.As political transitions interplay with political risk, one of the main risk new political regimes face is the risk of reversal (Acemoglu and Robinson, 2001; Svolik, 2015). New autocracies and new democracies face political instability. Ultimately, political instability may result in a new transition reversing the previous one. This may offset any merits of the former transition. The second part of my PhD dissertation aims at establishing the determinants of democratic transitions and autocratic reversals to better understand in which circumstances autocracies and democracies may be at risk.2. Determinants of political transitionsThe “Modernization hypothesis” posits economic development; e.g. high growth, industrialization, urbanization, as a long-term determinant of political transitions (Lipset, 1959). The process of development generates tensions, allowing the emergence of new interest groups that may mobilize and claim more rights (Llavador and Oxoby, 2005). Those divisions promote the creation of new coalitions supporting reforms and forming a new opposition to the ruling elite, e.g. industrialists and workers. Economic development may also facilitate the organization of those groups. Growth, industrialization and urbanization then reduce the cost of protesting and the expected cost of repression as well as they increase civic capital and the demand for public policies (Glaeser and Millett-Steinberg, 2016). All these corollaries simplify the organization of opposition groups demanding a regime change.Chapter 3 tests the modernization hypothesis and the importance of industrialization to spur support for democracy. To do so, it uses data from industrial censuses and Napoleonic plebiscites in XIXth century France. The reported empirical evidence points to the importance of industrialization in prompting support for the new democratic ideals emerging at that time in France. Industrialization remains a main driver of the opposition to Napoleon III autocratic power even after controlling for different measures of “modernization”. Industrialization therefore laid the ground for an actual democratic transition to materialize at the end of the Napoleonic era.A political transition does not only arise as a result of structural changes such as industrialization. They also are the product of political leaders’ individual decisions. These decisions, even if personal, are influenced by politicians’ environments. For example, wars, riots and protests send a signal on the capacity of the opposition to coordinate and on the weakness of the ruling authority (Acemoglu and Robinson, 2001). Facing these signals, a politician may resort to his.her environment to get cues on the best reaction possible. The environments of politicians prompt them to protect or abandon the regime in place in different ways. It may encourage the coordination of coups plotters as emphasized by Naidu et al. (2016) in the case of Haiti’s 1991 coup or more directly monitor politicians’ actions (Olson, 1993; Besley et al., 2017). Politicians’ environments also impact their decisions indirectly by transmitting specific norms (Calvó-Armengol and Jackson, 2009). All in all, political transitions are also the results of politicians’ personal decisions. These decisions are directed by a politician’s experience and environment. Chapter 4 investigates the determinants of these personal decisions and the importance of politicians’ environments in shaping them. To do so, it uses data on the vote by the French parliament on July 10, 1940 of an enabling act that granted full power to Marshall Philippe Pétain, thereby ending the Third French Republic and aligning France with Nazi Germany. It complements the previous literature by focusing on the behavior of political dynasties in that peculiar event. The literature on political dynasties in democracies usually considers dynasties as a homogeneous group and points out their negative effects. By contrast, we argue that political dynasties may differ according to their origin and that democratic dynasties – dynasties whose founder was a defender of democratic ideals – show a stronger support for democracy than other dynasties. Using individual votes in July 1940 and newly-collected data from the biographies of the members of parliament, we observe that members of a democratic dynasty had a 7.6 to 9.0 percentage point higher probability to oppose the act than members of other political dynasties or elected representatives belonging to no political dynasty. Suggestive evidence shows that the effect of democratic dynasties was possibly driven by internalized democratic norms and beliefs.In a general conclusion, I relate the results of these four chapters together to reconcile the different results and nuances they emphasize. I also take the opportunity to underline how the dissertation brings new insights in the study of political transitions and further explore possible avenues it offers for future research.