Autocracy as a Safety Valve for Democracy s Elites: Evidence from British Colonies

Transcription

1 Autocracy as a Safety Valve for Democracy s Elites: Evidence from British Colonies Christian Dippel June 25, 2014 Abstract Democracies frequently turn into autocracies, with detrimental effects on socioeconomic development. The coups behind these regime changes are usually viewed as originating outside the democratic process. This paper explores the contrasting possibility that they may be initiated by elected elites, if these elites prefer being ruled by an autocrat over being marginalized in the democratic process. Identification of this channel is difficult because coups are illegal, so that support for them tends to be unobservable. Instead, I study legal autocratic regime-change in 19th-century British colonies, where the constitutional rules allowed locally elected parliaments to abolish themselves and invite the colonial governor to rule autocratically in their stead. I identify 11 such regime changes and find (i) that in their lead-up, increasing political competition from freed slaves was tilting policies away from elites preferences, (ii) that variation in political competition explains the incidence and timing of regime changes, (iii) that regime changes subsequently tilted policies toward elites preferences, and (iv) that they maintained elites insider access to government, thus keeping the affected colonies on paths of low public-good provision and oligarchic rule. Keywords: Democracy, Autocracy, Regime Change, Elite Persistence, Oligarchy, Institutions, Franchise Extension. I thank Scott Ashworth, Leah Boustan, Ann Carlos, Melissa Dell, Stan Engerman, Fred Finan, Price Fishback, Morgan Kousser, Gary Libecap, Suresh Naidu, Ken Shotts, Guido Tabellini, Francesco Trebbi, Dan Trefler, Stephane Wolton, and Alex Whalley, as well as seminar participants at Caltech, Stanford, UBC, UCI, Yale, Chicago Harris, the Boulder IBS workshop, LACEA Political Economy meetings, and NBER POL and DAE meetings for valuable discussions and insightful comments. I also thank the UCLA Burkle Center and the UCLA Center for Economic History for financial support. UCLA Anderson School of Management ( christian.dippel@anderson.ucla.edu)

2 If we want things to stay as they are, they will have to change. Giuseppe di Lampedusa, The Leopard 1 Introduction A rich political economy literature has provided explanations for how autocracies may (gradually or suddenly) become more democratic (Lipset [1959], Acemoglu and Robinson [2000], Acemoglu and Robinson [2001], Lizzeri and Persico [2004], Brückner and Ciccone [2011], Besley et al. [2012]). However, there is also frequent movement in the opposite direction, with coups turning democracies into autocracies. Not only are these regime changes not uncommon, they also have significant detrimental effects on subsequent economic development (Meyersson [2014]). 1 Why then, do democracies turn back into autocracies? Most answers have focused on cross-sectional or slowmoving characteristics like the length of countries democratic experience and the stability of their institutions (Przeworski et al. [2000], Haggard and Kaufman [2012]) or economic inequality (Boix [2003], Acemoglu and Robinson [2006], Dutt and Mitra [2008]). While these answers highlight fundamental factors that explain the probability that a democracy fails, they have left the actual causes of the coups that turn democracies into autocracies relatively unexplored. This paper asks whether autocratic regime change may be initiated by incumbent democratic elites inside parliament, when they are threatened by increasing political competition. There is evidence that this mechanism was important in many real-world autocratic regime changes. In the 1967 coup d état in Greece, for example, politicians of the incumbent Conservative Party openly invited a military coup, fearing that the left-leaning Center Union Party would gain a parliamentary majority in the upcoming election (Kassimeris [2006]). Similarly, the 1971 military coup in Turkey was apparently supported by conservative parliamentary elements fearing increasing influence of both left- and right-wing parties and trade unions (Feroz [2002]). In Sierra Leone, when the incumbent prime minister, Albert Margai, narrowly lost the 1967 election to Siaka Stevens, he had planned ahead for this contingency and had the latter deposed by a military coup within hours after taking office (Cartwright [1970]). Recently, the New York Times reported that Thailand s military coup in May 2014 was the culmination of months of maneuvering by the Bangkok 1 Meyersson [2014], who estimates the economic effects of coups, reports that globally, out of 232 coups in 94 states from 1950 to 2000, around one quarter overthrew democratically elected governments. 1

3 establishment to sideline the populist movement that has won every national election since 2001, and that the establishment was seeking to suspend democracy, at least in the short term, as it struggled to unseat a ruling party it has been unable to defeat at the polls (Fuller [2014]). This is exactly the mechanism that this paper seeks to identify empirically, the only difference being that in the Caribbean the regime switches happened preemptively, i.e., before elites were defeated at the polls, and legally, by a vote in the elected legislature. A theoretical framework that fits this argument well is the one in Acemoglu and Robinson [2001], who extend the model of redistributive taxation in Meltzer and Richard [1981] to study democratic regime change, with elites trading off the cost of having a non-elite median voter against the threat of a costly revolt by the non-elite. 2 A natural extension of this framework is to allow elites to defer power to an autocrat who can eliminate the threat of revolt. This would be more attractive when the increase in political competition was more permanent, where the nonelite s objectives were more acrimonious to the elite and where the autocrat had objectives that were more closely aligned with the elite. 3 Because coups are by definition illegal, finding out whether they were supported or initiated by an incumbent democratic elite is difficult, even ex post. As an alternative, this paper focuses on a unique institutional rule in British colonies, where local parliaments could legally abolish themselves and invite the Crown to run the government autocratically with unconstrained executive power. Specifically, I focus on the 19th century Caribbean, because this is where all of these events actually occurred. Like the original American colonies and unlike any of the African colonies, 12 of the 15 Caribbean plantation colonies began the 19th-century with strong locally elected assemblies. By century s end, parliaments in 11 of the 12 had voluntarily abolished all elected elements from their government. This paper tests the hypothesis that these internally driven switches from democracy to autocracy were elites response to increasing political competition from former slaves, who had been freed by Britain s Empire-wide abolition of slavery in Using data from the Colonial Blue Books annual statistical accounts that were sent to Lon- 2 In the Meltzer and Richard [1981] framework, political competition is the threat of having a non-elite median voter set policies. The more unequal a society is, the further apart are the elite s and non-elite s policy bliss points. 3 In Acemoglu and Robinson [2006], elite rule is largely synonymous with non-democracy, but one can easily imagine making a three-way distinction between an elite-dominated democracy based on a narrow franchise, a broad democracy, and an autocracy. This distinction is implicit in the related literature on the extension of the franchise (Acemoglu and Robinson [2000], Lizzeri and Persico [2004]). 4 This political pressure varied within the Caribbean, depending on how easily freedmen could claim ownership of 2

4 don from each individual colony to report on local conditions the paper presents four findings, organized chronologically around the event of a regime switch to autocracy. First, I show that freedmen had gained real political influence in the lead-up to the regime switches. Under the democratic regime, there was a clear electoral cycle in public spending favoring the rural poor. In election years, parliaments voted systematically more expenditure on education and on a basket of other public goods that freedmen favored but elites did not, relative to other years in the same electoral cycle. 5 Second, I show that the instance and timing of the switches to autocracy can be explained through increases in political competition, as measured by local electoral turnover. 6 Because there is considerable variation in electoral turnover both across colonies and within colonies over time, I estimate the effect of electoral turnover on regime change both in a pooled cross section specification and in panel specifications. Remarkably, the estimated effect is practically identical in both specifications. A 10 % increase in electoral turnover raises the probability that a parliament voted to abolish itself by 2 % in a given year. To address the possibility that omitted factors that correlate with electoral turnover are driving the differential timing of regime switches, I control directly for a number of potential confounding drivers of regime change. I first control for the possibility that colonies learned from each others actions by including the stock of colonies that had already abolished their parliaments. I then control for increased attractiveness of the regime change after the the colonial administration put down the Indian Mutiny in 1857, which signalled a willingness to put down local peasant uprisings. Finally, I control flexibly for linear trends and year fixed effects. The results are robust to all these specifications. Because I hypothesize a causal link from electoral turnover to the probability of regime change, land, which gave the right both to vote and to run for office. 5 Educational expenditure was the expenditure category that was most consistently available. The basket of other public goods contains all public goods associated with public health. On both dimensions, the interests of freedmen and planter elites diverged strongly. 6 Electoral turnover is not the ideal measure of political competition, but it is the only one available here. Reassuringly, electoral turnover strongly correlates with two other, preferable, measures of political competition, which are only available for subsamples. The first is the franchise (the number of voters per capita). Franchise-data was reported only after 1854, which leaves too little data to be useful for estimation at the colony level. But at the sub-colony parish level, where the cross-sectional sample size is increased to over 100, I find that increases in the franchise significantly increased electoral turnover, both in specifications with parish fixed effects (identifying within parishes over time) and with electoral-cycle fixed effects (identifying across parishes within a colony). Second, for Jamaica only, I can calculate the rate of entry of new families into parliament from Jamaica s full parliamentary history in Roby [1831]. (To assess the flow of new entrants, one needs to accurately measure the stock of old political families from a reasonably long prior history, which the Blue Books do not provide.) For the 14 elections that took place in Jamaica from 1800 to 1865, there is a strong correlation between electoral turnover and political entry. 3

5 I also explore sources of exogenous variation underlying the panel-variation in electoral turnover. The historical record suggests that the exogenous introduction of the Caribbean Incumbered Estate Act (IEA), which facilitated the sale of bankrupt plantation land, strongly increased political competition by increasing land transactions, and that it did so differentially depending on local conditions that determined plantations profitability. Consistent with this, I find that the introduction of the IEA raised both electoral turnover and the probability of regime change, but that it did so significantly less where local conditions meant plantations were more profitable. Pushing the identifying assumptions a bit further by assuming that the introduction of the IEA, interacted with local conditions, influenced regime change only through its effect on electoral turnover, I also present instrumental variable (IV) estimates of the effect of electoral turnover on regime change. The IV estimates broadly confirm the baseline results, although they are larger, suggesting that when electoral turnover exogenously increased by 10 %, the probability that parliament voted to abolish itself was between 4 and 15 % larger in a given year. Third, I study the effects of switching from democracy to autocracy by comparing public expenditures before and after the regime switch. Across specifications, spending on education and on other public goods declined with the dismantling of elected parliaments, both in specifications with only common time-trends (where the control is colonies that had not yet switched), and in specifications with colony-specific time trends (where the control is also each colony s own counterfactual trend had it not switched to autocracy). I also investigate more closely the time-path of public expenditure, specifically around the years of regime change. There is some evidence that public-good provision actually increased from one year before to one year after the switch to autocracy, perhaps because elites wanted to secure a smooth transition. Fourth, I check directly whether elites were able to maintain insider access to the colonial autocrat after the regime switch. Two facts are consistent with this insider-access hypothesis: One, old political families continued to hold about 80 % of appointed legislative positions as late as 20 years after the regime switch, and two, within each colony, these old families came disproportionately from the plantation parishes. One issue that panels with few cross-sectional units need to address is proper clustering. Standard clustering relies on asymptotic theory that is unlikely to apply in a panel with only 12 colonyclusters. I therefore also report p-values for wild bootstrapped standard errors for all the main 4

6 results, as suggested in Cameron and Miller [2013]. This paper speaks to a well-established literature on the consolidation and stability of democratic regimes (Haggard and Kaufman [1997], Przeworski et al. [2000], Boix [2003], Acemoglu and Robinson [2006], Haggard and Kaufman [2012]). In that literature, political competition often leads to a strengthening of democracy: In Acemoglu and Robinson [2001], elites make political concessions such as extending the franchise in order to preserve their economic rents, a mechanism for which Brückner and Ciccone [2011] provide compelling empirical support in African data. In Besley et al. [2012], elites strengthen checks and balances in response to a higher likelihood of a non-elite future government. By contrast, this paper shows that political competition may lead elites to abolish democracy altogether and initiate a regime switch to autocracy if the cost of a non-elite democratic government is high relative to that of an autocratic regime. 7,8 This paper also speaks to a broader literature on the nexus of endogenous institutional change, elite persistence, and economic growth (North and Weingast [1989]). The evidence presented here fits best into the theory of the simultaneous change and persistence in institutions, in which changes in de jure institutions have no consequences because they are offset by changes in elites de facto collective action (Acemoglu and Robinson [2008]). 9 Most closely related are studies of the politics of the post-reconstruction U.S. South, where literacy tests, discriminatory schooling, the Ku Klux Clan, and Jim Crow laws were successfully used to reestablish the planter elites dominance over the democratic process after the Union Army pulled out in 1877 (Kousser [1974], Margo [1990], Naidu [2009]). Several noteworthy differences, which I explore in Section 5.1, exist between the U.S. South and the Caribbean. These differences meant that a Southern equilibrium 7 Because this will be the case only when the autocrat s actions are reasonably constrained, the Caribbean experience in the 19th century is more akin to the history of military coups in Greece, Turkey, and parts of Latin America than it is to those in post-colonial Africa. 8 There is also a strand of this literature where economic growth means that elites benefit more from broader representation than they lose from giving up their dominance of the democratic process, a central tenet of modernization theory (Lipset [1959]). In Acemoglu and Robinson [2000], revolts by the poor become costlier as economic prosperity increases. In Lizzeri and Persico [2004], the interests of wealthy capitalists and workers are aligned against the landed gentry, and capitalist elites therefore seek to strengthen the workers. These mechanisms help explain the widespread expansion of the franchise to the poor in many parts of the world in the 19th century (Engerman and Sokoloff [2005]), but none of these mechanisms were at play in the Caribbean. 9 Meyersson [2014] studies the effects of coups on development outcomes and shows that their negative effect is bigger when the replaced regime was more democratic. In this context, the Caribbean may be an intermediate case, where a very unequal democracy was replaced by a relatively stable and technocratic autocracy. 5

7 was not feasible for Caribbean planter elites. 10,11 This paper also speaks to an influential line of research that has advocated historical colonialism as a natural experiment to identify the causal effect of institutions on long-run development (Acemoglu et al. [2001, 2002], Engerman and Sokoloff [2002], Feyrer and Sacerdote [2009]). However, empirical measures of concrete colonial institutions and their change over time have remained elusive. 12 To my knowledge, this is the first paper that studies panel-variation of concrete colonial institutions. Section 5.4 presents and discusses some additional evidence on this data variation in a wider set of British colonies. Finally, this paper contributes to a large body of economic history on colonialism and the Caribbean. See, respectively, Lange [2004] and Iyer [2010], as well as Engerman [1982], Engerman [1984], Holt [1991], and Henry and Miller [2009] for references. This paper is organized as follows: Section 2 provides background, Section 3 discusses the data, Section 4 presents the four key results, Section 5 discusses the results, and Section 6 concludes. 2 Background: The Political Consequences of the Abolition of Slavery This study focuses on the Caribbean because this is where the events of interest occurred. But there is a deeper appeal in this focus on the Caribbean, because its colonies were highly comparable in their histories and economic structures, and were arguably exposed to the same shocks over time. Most importantly, there was a single, well-identified driving force behind all regime changes in the data, which can be explored for empirical analysis. The Caribbean colonies were settled in the 17th century by British smallhold farmers, much like in the Chesapeake Bay. These smallhold farmers sought to control taxes and spending through local assemblies (Taylor [2002, p. 246]), which seriously curtailed the powers of the governors in 10 Nonetheless, Caribbean elites response of abandoning democratic institutions was not without equivalent in the U.S. South: The county government of Dallas County in Alabama asked the governor to dissolve it when a black majority was looming in 1880, and the county government was not reestablished until 1900 (Kousser [1999]). 11 Major changes in democratic institutions did occur in the U.S. South after the 1965 Civil Rights Act, when municipal governments electoral rules changed in black-dominated U.S. cities to entrench white votes (Trebbi et al. [2008]). Unlike in the 19th-century Caribbean, these institutional changes were not enough to prevent major concessions in public spending to the newly enfranchised (Husted and Kenny [1997]). 12 Some recent exceptions that study concrete colonial institutions are Banerjee and Iyer [2005] and Dell [2010]. However, they study the persistent effects of initial institutions. 6

8 the colonies (Morrell and Parker [1969, p. 435]. This assembly system remained even after large-scale sugar plantations had completely displaced the smaller tobacco plantations in the early 18th century. The franchise was everywhere highly concentrated 13, but this reflected merely the small number of whites everywhere (Dunn and Parker [1972]). Indeed, the legacy of the Caribbean s smallhold origins meant that typically all whites were enfranchised, because the property holdings required to vote continued to be low. They had typically stayed at their original 10-acre threshold (Wrong [1923]). 14 After slavery was abolished by an Act of English Parliament in 1834, many freedmen left the plantations and began to purchase smallholds. 15,16 Black leaders quickly recognized that smallhold expansion meant franchise expansion. 17 In Jamaica for example, Baptist ministers tried to mobilize the dormant black electorate. They encouraged their members to purchase freeholds and register to vote (Holt [1991]). This gradual enfranchisement of the freedmen led to an increase in political competition: The planters steadily lost their political dominance as disputatious assemblies were infiltrated by men of color independent of the plantation economy (Burroughs [1999]). 18 Finally, from 1854 to 1877, 11 of the Caribbean parliaments simply abolished themselves. Parliament would vote on a bill to abolish parliament and invite the Crown to write a new constitution for the colony. In all cases, this new constitution followed the standard template of Crown Colony rule: All functions of government were controlled by the colonial administration, with the governor appointing the local legislature and judiciary. Locally elected parliament was replaced with a legislature that consisted partly of colonial officials and partly of local appointees. Historians have argued that this voluntary dismantling of Caribbean parliaments was planter elites response to the political competition of freedmen: While they had traditionally jealously guarded [the as- 13 Wrong [1923, p. 69] states that throughout the pre-abolition Caribbean, it was distinctly the exception for a member of the legislature to be returned by more than 10 votes. 14 The franchise in the Caribbean, as elsewhere in the British colonies, was obtained primarily through land ownership. The franchise rules reported in the Blue Books show that there is very little variation in the 10-acre threshold across colonies. 15 Slavery ended in 1836, but abolition only really became effective in 1838, after a two-year transitional period of quasi-indenture called Apprenticeship. 16 This happened through purchases of marginal plantation lands or Crown lands. Squatting, pervasive throughout the Caribbean, also gave legal title after 12 years on private land and 60 years on Crown lands (Craton [1997, p. 390]). 17 Contrary to the situation in Europe, the franchise in the Caribbean was therefore expanding without any change to the de jure rules that regulated the right to vote. 18 The main conflict of interest between the planters and freedmen was on the issue of taxation and public-good provision. Freedmen s primary concerns were land redistribution and public-good provision. Planters were naturally averse to land redistribution and were also disinterested in the expansion of education and health services (Sewell [1861, p. 39], Dookhan [1977], Brizan [1984, p. 163]). 7

9 Table 1: Descriptives & Abbrev. & Year & Year of & Population & Pop-Share & Density & Slave-price Colony & & Founded & Regime Change & 1836 & White, 1836 & 1836 & 1836, in Antigua & ATG & 1632 & 1867 & 35,188 & 5.4 & 125 & 35 Barbados & BRB & 1629 & & 105,812 & 12.8 & 246 & 38.8 Dominica & DOM & 1763 & 1862 & 16,207 & 3.9 & 21 & 28.7 Grenada & GRD & 1763 & 1876 & 17,751 & 2.6 & 52 & 41.2 Honduras (Belize) & HON & 1638 & 1869 & 8,235 & 4.2 & 4 & 94.2 Jamaica & JAM & 1655 & 1860 & 381,951 & 8.2 & 34 & 31 Montserrat & MON & 1634 & 1861 & 6,647 & 4.3 & 65 & 25.3 Nevis & NEV & 1623 & 1867 & 7,434 & 5.4 & 80 & 21.4 St. Kitts & STK & 1628 & 1867 & 21,578 & 6.4 & 113 & 29.7 St. Vincent & STV & 1763 & 1866 & 26,659 & 4.7 & 69 & 39.5 Tobago & TOB & 1763 & 1873 & 11,456 & 2.3 & 38 & 41.7 Virgin Islands & VIR & 1672 & 1855 & 7,471 & 12.4 & 49 & 23.1 Guyana & GUY & 1803 & n/a & 66,561 & 0.7 & 6 & 87.4 St. Lucia & SLU & 1803 & n/a & 17,005 & 11.3 & 27 & 50.3 Trinidad & TRI & 1797 & n/a & 34,650 & 8 & 7 & 83.6 Note: The Data in this table are from the Colonial Blue Books and from Martin (1839). Year of regime change is not applicable for the 3 late colonies, which were founded as Crown Colonies. The two mainland colonies, Honduras and Guyana, had very big hinterlands, so I use 10 % of their area as an approximation to their de facto extent at the time. Density is calculated as population per square kilometer. semblies] against interference by the colonial adminstration (Wrong [1923, p. 70]), Lowes [1994, p. 35] argues that in the end, the demand of an increasingly restive nonwhite middle class for a voice in island affairs proved the greater fear and they voted themselves out of office. Similarly, Ashdown [1979, p. 34] argues that the colonies gave up their elected assemblies voluntarily, for in most cases the white, privileged classes preferred direct imperial government to the government of the colored classes who were slowly obtaining greater representation in the legislative councils. 3 Data 3.1 The British Caribbean The 15 British plantation colonies in the Caribbean were founded in three waves. The first wave Antigua, Barbados, Honduras, Jamaica, Montserrat, Nevis, St. Kitts, and the Virgin Islands were founded in the 1600s by small-scale British planters. The second wave Dominica, Tobago, St. Vincent, and Grenada were annexed from France at the end of the Seven Years War in They were resettled by sugar planters from the other Caribbean islands and from the start were endowed with the same representative institutions (Ragatz [1928, p. 112]). The last three colonies 8

10 Trinidad, St. Lucia, and Guyana were ceded to Britain between 1797 and By then, the Crown had started to assert more authority over its colonies, so that these three late colonies were already founded under Crown Colony rule (Will [1970]). Table 1 lists the colonies with their foundation dates, their characteristics at the time of abolition, and the dates they transitioned toward Crown Rule. 19,20 Table 1 includes the three late colonies for completeness, but these are naturally not included in the empirics, so that the effective sample consists of the 12 highly comparable plantation colonies that started out under the Assembly system. Figure 1 plots the raw electoral turnover data leading up to constitutional changes for the 12 colonies. Noteworthy variation exists both in the cross-section and in the time-series. In the cross-section, consider the contrast between the Virgin Islands, the first switcher, and Barbados, the only colony that never switched. In Barbados, turnover was below 40 percent in every one of its elections. By contrast, in the Virgin Islands, turnover was above 40 percent in every one of its elections. In the time-series, electoral turnover was generally trending up before the regime switches. There are some instances, such as in Dominica and Grenada, where the regime switch followed a rapid increase, and some, such as in Dominica and Grenada, where the increase was more gradual. 3.2 Data Sources The main data source for this paper are the British Colonial Blue Books, annual statistical accounts that were sent to London from each individual colony to report on local conditions. 21 From 1836, the books Councils and Assemblies section reported the names of all local politicians, with elec- 19 While all 11 colonies eventually switched to full Crown rule, ten of them first changed their constitution to a transitional semi-representative system with an elected minority. Historians agree that this was a transitional mechanism with the clear end-goal of securing a majority for the full switch toward Crown Colony rule (Craig-James [2000, p. 256], Brizan [1984, p. 204]). The Colonial Office List s description of Grenada s constitutional history reads, The constitution was remodeled by an Act on October 7th 1875 and a single legislative chamber was established, [which] consisted of 8 members elected by the people and 9 nominated by the Crown. This Assembly at its first meeting on February 9th 1976 addressed the Queen that it had passed a bill for its own extinction (Britain [1879, p. 188]). Since colonies only stayed in the transitional state for six years on average, there is little informational content in that period; I focus on the first constitutional change. 20 For two colonies, I report an even earlier date: Jamaica and Montserrat actually changed the rules governing their franchise to limit political competition. In both cases, this led to active resistance, which still culminated in the abolition of their parliaments. Since limiting the franchise had the same objective as dismantling parliament, I use the dates of the change in franchise rules in those two cases, which preceded the regime switch by a few years. Jamaica switched to Crown rule in 1865 and Montserrat in For years before the 1890s, at most two copies exist of each Blue Book, one in the issuing colony s archives and one in the British National Archives, in London, where this data was hand-collected. 9

11 Figure 1: Constitutional Change and Electoral Turnover tion dates and the parish they represented. Also from 1836, the Comparative Tables of Revenue and Expenditure reported detailed breakdowns of taxation and spending by category, though not by parish. From 1854, the books Political Franchise section also reported the number of registered voters for the last election by parish. Figure 1 in the appendix shows an example of the Councils and Assemblies section for Barbados 1847 Blue Book. 3.3 Data Structure The first result, in 4.1, focuses on political competition within an electoral cycle. This result naturally only uses data from before the switch to autocracy. The third result, in 4.3, compares outcomes under the two regimes; it is the only exercise for which the full panel is used. The fourth result, in 4.4, focuses on elite-persistence after the switch to autocracy only, which naturally studies data on from after the switch. For the second result, in 4.2, the data is set up as duration data, because the regime switches were absorbing and because the main regressor, electoral turnover, no longer exists as data after the switch. For each colony, the outcome data is set up as a time-series of 0s followed by a single 1 in the year of the regime change, with which that colony s data end. Table 1 in 10

12 the appendix illustrates the structure of duration data for Antigua s 1853, 1860 and 1867 elections, ending in the regime change. 3.4 Measuring Political Competition Electoral turnover suffers from several deficiencies as a measure of political competition. On the one hand, some turnover within the elite is likely. On the other hand, an increase in political competition could take the form of a one-time permanent entry of competitors into politics, followed by limited subsequent electoral turnover. Unfortunately, electoral turnover is the only available measure of political competition in this data. 22 Fortunately, two better measures of political competition are available for sub-samples of the data, and can be correlated with electoral competition for those sub-samples as a validity check. The first measure is the franchise (the number of voters per capita). Franchise data was only reported after 1854, which leaves too short a panel to be useful for estimation at the colony level. But at the sub-colony parish level, the cross-sectional sample size is increased to over 100, which gives enough power to test for a partial correlation between the franchise and electoral turnover at the parish level. The second measure is political entry. To assess the flow of entrants into politics, one needs to accurately measure the stock of established political families from a reasonably long prior history, which the Blue Books unfortunately do not provide. However, for Jamaica only, these data exist in a complete history of Jamaica s parliament Roby [1831]. In correlating franchise and electoral turnover at the parish level, I can compare data within parishes over time or within a colony (and electoral cycle) across parishes. The following regression nests both comparisons: electoral turnover ip,el cycle = γlog(reg. voters) ip,el cycle + ϕ ip + φ el cycle + ɛ ipt, (1) where φ el cycle are colony-specific electoral-cycle fixed effects and ϕ ip are fixed effects for parish p in colony i. Colony-specific electoral-cycle fixed effects are more conservative than colony fixed effects because they imply a comparison of parishes across a given colony only within the same electoral cycle. Columns 1 and 2 in Table 2 show that electoral turnover and the franchise are 22 The Blue Books do not provide enough additional information about the listed politicians to infer their political affiliation. Similarly, there is no obvious way to link names to race or economic class. 11

13 Table 2: Does Electoral Turnover Measure Political Competition Dependent: log(reg. voters/population) Number of political newcomers in one electoral cycle in Jamaica (1) (2) (3) Elect.Turnover 0.258*** 0.137** *** (2.646) (2.164) (3.245) Unit of observation parish-year parish-year Jamaican elections fixed effects: parish parish & el.cycle Observations 1,363 1, R Note: This tables validate electoral turnover as a measure of political competition, using two additional data sources. In columns 1 2, electoral turnover is related to the franchise (per capita registered voters) at the parish level, from the Blue Books. 10 % points higher electoral turnover is associated with a 1.5% 3% higher number of voters (The franchise datawas reported only after 1854, so cannot be used to explain regime changes at the colony level.) Column 3 uses Roby's (1831) history of Jamaican parliament to relate the number of names that appeared for the first time ever to electoral turnover in that electoral cycle. 10 % points higher electoral turnover is associated with the entrance of 3 new political families. (This measure of entrance into politics requires data on the full previous political history, only available for Jamaica.) In columns 1 2, s.e. are two way clustered at the parish and the colony electoral cycle level. In Column 3, s.e. are not clustered. significantly positively correlated, whether I compare only within-parish over time in column 1 or within-parish over time and across parishes within a colony in column 2. Electoral turnover falls between 0 and 1 so Table 2 says that 10 percentage-points higher electoral turnover is associated with a 1.5% - 3% higher number of voters. Column 3 is more straightforward because the number of observations is limited to the 14 elections that took place in Jamaica from 1800 to A simple regression of entry on electoral turnover shows a strong correlation between the two. 10 percentage-points higher electoral turnover is associated with the entrance of 3 new political families. In combination, these results thus lend considerable credence to electoral turnover as a measure of political competition, in the specific context studied here. 4 Four Key Results 4.1 Evidence of Political Pressure on Elites in the Lead-Up to the Regime Switch If political competition from freedmen had real bite, then politicians should have had to cater to freedmen in election years, leading to electoral cycles (Besley and Case [1995]). The historical record suggests that the interests of the freedmen and the elite were conflicting on several dimensions, including taxation sources, land redistribution, and public spending (Holt [1991, ch. 7]). I focus on public spending because this is the only one of these dimensions on which there is data 12

14 Table 3: Electoral Cycles: Did Elites Cater to Freedmen at Election Time? Dependent: log(educational expenditure) log(sanitation & health expenditure) (1) (2) (3) (4) (5) (6) D(electoral year) 0.200** 0.184** 0.174** 0.224** 0.210** 0.180** (2.283) (2.263) (2.057) (2.617) (2.470) (1.992) log(total exp) 0.830** 0.675* 0.387* (2.447) (1.864) (1.802) (1.301) Fixed Effects decade decade Observations Note: All specifications include colony electoral cycle fixed effects so that the election year effect is identififed relative to the rest of the electoral cycle. Reported regresssions use only the data before the switch since there are no elections after. D(electoral year) equals 1 only in the year before the election. The number of observations is less than for the regressions relating institutional change to politcial competition, because the Tables of Revenue and Expenditure did not report fine categories in the early years. in the Blue Books. Two expenditure categories were consistently reported in the Blue Books, and evinced clear conflict between freedmen and planters: spending on education and expenditures related to health provision. Both were either irrelevant to the planters or could be provided privately (Lowes [1994]). 23 I therefore look for evidence of increased spending on education and health-provision in election years, i.e., when politicians had re-election concerns. Specifically, I estimate Policies it = βd(election-year) it + γx it + φ el cycle + ɛ it, (2) where φ el cycle is a fixed effect for an electoral cycle in a colony so that the effect of an election year is identified only relative to other years in the same electoral cycle. Table 3 reports the results of this estimating specification (2). Columns 1 through 3 report on educational expenditure and columns 4 through 6 on expenditure for health provision. The first column includes only colony-electoral-cycle fixed effects, i.e., each electoral cycle is colonyspecific. The second column also controls for more high-frequency variation in total expenditure. The third column also adds decade fixed effects to further controls for any broad shifts in public spending. There is consistent evidence for the existence of an electoral cycle. Within an electoral cycle, politicians spent about 20 % more on both categories in election years. This evidence suggests that the freedmen were a political force to be reckoned with in the period after abolition and leading up to the regime changes. Because specification (2) uses within-electoral-cycle varia- 23 The children of the Caribbean great planters were almost always educated in England, and even their wives frequently lived there (Lowes [1994]). 13

15 tion, as opposed to within-colony, standard errors should be clustered at the electoral-cycle level. And with 163 electoral cycles in the data, there should not be an issue with the asymptotics of clustering (Cameron and Miller [2013]). Indeed the p-values for wild-bootstrapped standard errors (clustered at the colony level), are and in columns 2 and 5 respectively. 24 This is smaller than the p-values of and , obtained with standard clustering at the electoralcycle level. 4.2 Did Political Competition Cause Regime Switches to Autocracy To test the main hypothesis that regime changes to autocracy were elites defense against political competition, I regress Regime Switch it = βpolitical competition it + φ i + φ t + ɛ it, (3) where the dependent Regime Switch it takes value 1 in the year when the switch occurs and value 0 in all years before then. 25 Political competition is measured by electoral turnover, φ i are colony fixed effects, and φ t measures a time-trend in several different ways across specifications. Because Figure 1 displayed both considerable cross-sectional and time-series variation in electoral competition, I separately report pooled cross-sectional results (without φ i ) in table 4, and fixed-effects results in table 5. Panel A of table 4 reports the results of the pooled cross-sectional specification. Column 1 suggests that 10 percentage-points higher electoral turnover at the previous election is associated with about a 2 % (0.189*0.1) higher probability of abolishing elected parliament in a given year. In columns 2 to 5, I address the possibility that these results are biased by time-variant forces that drive the probability of regime change and correlate with political competition. In column 2, I simply include a common linear trend. In column 3, I control for a peer effect, with regime change in one colony triggering change in the neighboring colonies, by including the number of already transitioned colonies. Another possibility is that over time, switching to Crown rule became more attractive for other reasons. For example, the violent suppression of the Indian 24 A p-value of means that there was not a single draw in the bootstrap procedure, that generated a t-value above the one obtained with standard clustering. 25 With duration data, the data ends in a single 1 for each colony that switched to autocracy. 14

16 Panel A: Table 4: Regime-Change and Electoral Turnover Does Electoral Turnover Explain Regime Change? (1) (2) (3) (4) (5) Electoral Turnover 0.189*** 0.215*** 0.221*** 0.210*** 0.202*** (3.397) (4.223) (3.940) (4.580) (4.316) R Panel B: Does Electoral Turnover Explain Regime Change, with Colony Fixed Effects? Electoral Turnover 0.277*** 0.229** 0.237** 0.241** 0.220** (3.234) (2.534) (2.784) (2.946) (2.519) R Controls: year No.(Switched) D(post-1857) year-fe Note: This table reports on regressions of regime changes on electoral turnover in a panel. The data is organized as duration data, i.e. each colony's time series is a series of 0s ending in a single 1. Intuitively, the duration setup is chosen because regime changes are "absorbing," i.e. there is no switching back from autocracy. Mechanically, the duration setup is necessary, because the regressor, electoral turnover, is not defined after the regime change. Panel A reports coefficients for OLS with different controls, Panel B also includes colony fixed effects, which wash out any differences in average electoral turnover in Figure 1. Overall, 10 percentage point higher electoral turnover leads to a 2% (0.189*0.1) higher probability of abolishing parliament in a given year. N = 337 in all columns, all s.e. are clustered at the level of a colony level. Mutiny of 1857 demonstrated to local colonial elites that the colonial administration could be counted on to suppress local rebellions. (Burroughs [1999, p. 181]). 26 In column 4, I therefore control for a post-1857 indicator. Finally, in column 5, I include year fixed effects. The results appear overall robust to the inclusion of all these controls. Panel B of table 4 replicates these results with fixed effects, i.e., using identifying only off within-colony variation. The coefficient estimates are remarkably stable across the two panels. All regressions in table 4 are clustered at the colony level. Ordinarily, when the number of clusters is as low as 12, standard clustering can underestimate the true standard errors. A wild bootstrap is the appropriate way to address this and wild-bootstrapped p-values can be considerably larger than the standard ones (Cameron and Miller [2013]). However, in the duration data setup here, which is naturally characterized by long series of 0s which end in a single 1, this turns out not to be the case. The p-values corresponding to the colony-clustered standard errors in column 5 of tables 4 and 5 are, respectively, and The p-values for the wild bootstrap for those same two regressions are both smaller; and respectively Since the Crown could impose abolition on the Caribbean colonies, it is reasonable to ask why it could not impose Crown rule. The historical record suggests it was simply not important enough and that the demonstrated willingness of Caribbean assemblies to bring the local political process to a halt made colonial administrators unwilling to force Crown rule onto the Caribbean colonies (Wrong [1923]). 27 A p-value of means that there was not a single bootstrap-draw for which the wild t-statistic exceeded the standard t-statistic reported in tables 4 and 5. (I used 1000 draws, but larger numbers of draws equally generated a 15

17 To strengthen the claim of a causal link between electoral turnover and the probability of regime change, I next explore the underlying causes of the panel-variation in electoral turnover, that drives the results in table 4. This paper s argument is that the increase in political competition measured by electoral turnover was linked to freedmen being increasingly successful at getting their votes registered, and, further, that the degree to which political competition increased was determined by how easily freedmen could obtain the land needed to register their vote. Based on this causal chain, the historical record strongly suggests the introduction of the Caribbean Incumbered Estate Act (IEA) as an exogenous shock that increased political competition by increasing land transactions (Lowes [1994]). The historical record further suggests the differential impact of the IEA on different islands as an underlying driver of the differential increase in electoral competition. Before the IEA was implemented across the Caribbean in 1854, many estates were already bankrupt because of increased labor costs imposed by abolition. However, until then, a bankrupt plantation s debt was inseparable from its owner s other assets. If an estate was worth 20,000 pounds but indebted to the amount of 30,000, its owner still had to cover the remainder after the estate s sale. Consequently, many estate owners held on to their bankrupt estates, and, while there was pervasive squatting by freedmen on abandoned plantations, this did not give the legal title required for registering a vote. The IEA resolved this bottleneck by clearing estates owners of any remaining liabilities after selling incumbered estates. 28 However, this effect of the IEA varied by islands, its impact depending largely on the profitability of estates. Where plantations remained profitable after abolition, the IEA had little impact (Lowes [1994]). The key cross-sectional characteristic which determined plantations profitability was the cost of labor, which in turn depended on freedmen s options outside of estate work. Where there was little land, there were limited outside options, labor remained cheap and estates profitable (Dippel et al. [2014]). 29 The importance of density is well-illustrated by the case of Barbados. Barbados was the only one of the 12 Caribbean plantation colonies that did not abolish its parliament, it had the lowest average electoral turnover of all the colonies, and it was more than twice as dense as any other Caribbean colony. (See Figure 1 and table 1.) p-value of Essentially the same legal arrangement as in those U.S. states that have non-recourse bankruptcy laws today. 29 Dippel et al. [2014] study Caribbean labor markets. They provide a model of coercion in which plantation workers outside options and plantations profitability a jointly determined by underlying geographic factors, and they provide detailed historical accounts of the functioning of these labor markets and the Caribbean plantations. 16

18 Table 5: Exogenous Variation Underlying the Changes in Electoral Turnover and Political Regime (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Panel A: Did the Introduction of the IEA cause Regime Change, but less so Where Elites had More Control? (The "Reduced Form") D(IEA) 0.098*** ** ** 0.108*** * *** (5.256) (1.772) (2.647) (1.757) (2.305) (4.493) (1.379) (2.149) (1.602) (3.464) D(IEA)*1836-density *** *** *** *** ** ** *** *** ** * (-3.469) (-5.631) (-4.774) (-3.474) (-2.305) (-2.604) (-4.089) (-3.716) (-2.651) (-2.052) R Panel B: Did the IEA Increase Electoral Turnover, but Less so where Elites had More Control? (The "First Stage") D(IEA) * *** (1.326) (0.741) (1.351) (0.928) (1.862) (1.304) (0.128) (1.096) (0.227) (4.566) D(IEA)*1836-density *** *** *** *** *** * ** (-6.988) (-4.957) (-4.535) (-7.549) (-5.840) (-1.142) (-2.129) (-1.417) (-1.323) (-2.692) R Panel C: IV estimates Electoral Turnover 0.376*** 0.491*** 0.570*** 0.409*** 0.408*** ** ** (5.222) (7.526) (6.090) (4.786) (3.612) (1.189) (2.440) (1.630) (1.321) (2.408) Weak-Instrument F-Test Controls: year No.(Switched) D(post-1857) year-fe year No.(Switched) D(post-1857) year-fe Colony Fixed Effects Yes Yes Yes Yes Yes Note: This tables presents the results from the instrumental variable exercise, divided into a "reduced form" regression of regime change on the instruments (Panel A), the "first stage" regression of turnover on the instruments (Panel B), and the IV regression of regime change on instrumened turnover (Panel C). Panels D to F replicate panels A to C, with colony fixed effects included. Although the coefficient on "D(IEA)" is reported as a matter of separate interest, only the interaction term "D(IEA)*1836 density" is treated as an excludeable instrument in Panel C and F. The number of observations is 337 in all specifications. All s.e. are clustered at the colony level. 17

19 Motivated by this background, table 5 shows how the introduction of the IEA related to the timing and instance of the regime changes, as well as to electoral turnover, and how this relationship depended on density. Across columns, columns 1 5 are organized like table 4, incrementally introducing the same controls, while columns 6 10 do the same but further add colony fixed effects as in Panel B of table 4. In table 5, Panel A reports the effect of D(IEA) (time varying) and D(IEA) 1836-density (time-and-colony-varying) on the likelihood of regime change. Panel B reports the effect of D(IEA) and D(IEA) 1836-density on electoral turnover. If one is willing to assume that D(IEA) and D(IEA) 1836-density impacted regime change only through their effect on electoral turnover, then one can instrument electoral turnover with D(IEA) and D(IEA) 1836-density. A somewhat weaker assumption, and the one on which I report here, is that D(IEA), which is essentially a post-1854 break, may have correlated with other unobservable factors influencing regime change, but that the interaction D(IEA) 1836-density impacted regime change only through electoral turnover. (This implies D(IEA) is treated as a control variable and D(IEA) 1836-density as the instrument.) Under this assumption, I also report IV results in Panel C of table 5. Panel A shows that the introduction of the IEA was indeed associated with a higher probability of regime change, and that this was more true in islands that were less dense, i.e., where more land was freed up by the IEA. The results for D(IEA) are weakest in columns 4 and 9, which allows for a trend break after the Indian Mutiny. This is not surprising because columns 4 and 9 amount to a horserace between a post-1854 and a post-1857 indicator. Overall, more credence should be given to the interactions term D(IEA) 1836-density, which remains significant to all specifications. Panel B shows that the introduction of the IEA was also associated with higher electoral turnover, although these results are not very strong with colony fixed effects. The IV results in panel C also broadly confirm the previous results, although the results with colony fixed effects are again much weaker. While the IV results should be taken with a grain of salt because of the relatively strong assumptions needed to interpret them literally, it is worth noting that the IV results are about twice as large as those in table 4, suggesting that 10 percentage-points higher electoral turnover increased the likelihood of abolishing parliament by between 4 and 20 % in a given year. A plausible interpretation of these larger IV estimates is that the IV better isolates the part of electoral turnover that truly represents political competition. 18

20 Overall, the results in all three panels of table 5 are consistent with the logic of the paper s argument that increasing electoral turnover caused preemptive regime changes, and with the historical record which suggests that the rate at which electoral turnover increased depended at least in part on external factors like changing legal conditions and their interaction with islands geography. 4.3 The Consequences of Regime Change Next, I investigate the consequences of abolishing parliament. Because of the endogeneity of the regime-switch, the time-window just around it is of separate interest here. For example, it is possible that elites strategically delayed utilizing the regime change in order to secure a smooth transition to the new regime. To pick up such transition-patterns in addition to the level effect of autocracy, I regress Policies it = βautocracy it + 2 k= 3 λ k Regime-Switch it+k + φ t + θ i + ɛ it, (4) where Policies it are the two measures of expenditure, on education or health. The main coefficient of interest is on autocracy it, which estimates the permanent effect of the regime switch; it turns from 0 to 1 on the year of the regime change, and stays at 1 afterward. 2 k= 3 λ kregime-switch it+k is a set indicators for the 6 years around the year of the change. 30 Lastly, θ i are colony fixed effects, and the counter-factual time-trend φ t is modeled with either year fixed effects only or year fixed effects plus a colony-specific linear trend. Table 6 shows the results of estimating specification (4). The first two columns consider educational expenditure, the next two expenditure on healthprovision. All columns include colony and year fixed effects and columns 2 and 4 add colonyspecific linear trends. The permanent effect of abolishing parliament appears to have reduced educational expenditure by about 50 percent and reduced expenditure for health-related public goods by about 30 percent while the estimates in Table 6 suggest that it increased by about 20 per- 30 I show a window of six years because the only significant movement is in the four years around the change, and coefficients get monotonically less significant away from the regime switch. 19

21 Table 6: The Consequences of Regime Change Dependent: log(educational expenditure) log(sanitation & health expenditure) (1) (2) (3) (4) autocracy it ** ** ** ** (-2.877) (-2.241) (-2.260) (-2.221) p-val [cluster(colony)] [0.045] [0.048] [0.015] [0.046] p-val [wild bootstrap] [0.084] [0.064] [0.104] [0.046] regime-switch it (1.352) (0.371) (0.693) (1.043) regime-switch it * (1.931) (1.651) (0.955) (0.530) regime-switch it *** 0.582** (3.733) (2.712) (1.367) (0.993) regime-switch it * (1.223) (1.974) (1.783) (1.577) regime-switch it ** 0.916** (3.045) (2.816) (1.451) (0.988) regime-switch it (1.587) (1.748) (0.602) (1.172) colony-specific linear trend: Y Y Observations R Note: All regressions control for colony fixed effects, year fixed effects and a log(total expenditure). The main regressor autocracy it turns to 1 the year of the regime switch and stays at 1 after that; it estimates the one time level effect of the regime change. Each of the six regime switch indicators turns to 1 only in a single year (e.g., regime switch it 3 turns to 1 only in the single year 3 years before the regime switch); they trace out how the dependent changes around the time of the regime change. The regime switch indicators show that there was systematically higher expenditure on public goods just around the regime switch. Columns 1 and 3 have year fixed effects and columns 2 and 4 add colony specific linear trends. Data for the Tables of Revenue and Expenditure runs from 1838 to 1900 for both variables. However, in early years, education expenditure was often not separately reported so that N is smaller for educational expenditure. All s.e. clustered at the colony level. cent in an election year. 31,32 Overall, the estimates clearly suggest that autocracy helped the elites. For the main regressor autocracy it, table 7 also reports p-values for wild-bootstrapped standard errors in addition to those clustered by colony. Unlike in Table 3 (where the 163 electoral cycles seemed like a more reasonable cluster-dimension) and in tables 4 and 5 (where the data was set up as duration data), here the wild-bootstrapped p-values are bigger than the conventional ones, although the results do remain largely significant at conventional levels. Especially in columns 2 and 4, which include colony-specific time trends, the wild bootstrap makes little difference. The coefficient-estimates λ k are of separate interest from the main result. They display a very 31 This makes sense because educational expenditure was arguably even less in the interest of elites. While there were some positive externalities from better health even if health was a private good to elites, education have have even had negative externalities. An obedient and stable workforce was key for the profitability of elites plantations in the Caribbean (Dippel et al. [2014]), and higher education would have raised the outside options of plantation workers, a pattern observed in other agricultural labor markets as well (Bobonis and Morrow [2014]). 32 Obviously, if expenditure on public goods shrank as a relative share of total expenditure, some other expenditures must have increased. I found some evidence that aggregate salaries for the civil administrators increased. This is consistent with anecdotal evidence that black leaders were co-opted by the planters (Honychurch [1995], Brizan [1984]). However, this evidence was weak because aggregate salaries were reported in inconsistent ways across colonies and over time. 20

22 specific time-path of public-goods provision around the regime change. There was systematically higher expenditure on public goods from about two years before the change to about two years after. These results are more clear for education, but also come close to 10% significance for the other public good category. Educational expenditure, was between 50% and 90% higher in the three years around the regime switch. This at least strongly suggests that elites strategically delayed utilizing the regime change in their favor, in order to secure a smooth transition to the new regime. 4.4 Did Elites Maintain Insider Access A possible explanation for the findings in Table 7 is that the switch to Crown rule shielded elites from popular pressure but in fact preserved their insider access and influence on the colonial autocrat s policies. Since elites access would have depended on them actually staffing critical administrative positions, I look for direct evidence of this in the data of appointed legislators under Crown rule. I link all individuals into families based on last name, and calculate the share of appointed legislators whose families ever held elected office under the old Assembly system. 33 I pool all appointed legislators into a common appointed legislature defined by event time, i.e. year 1 pools the eleven legislatures after each one s switch to Crown rule. I then calculate the share of legislators whose families ever held elected office under the old Assembly system. 34 Table 7 shows that this share is both very high and very stable over time. Twenty-one years after parliaments were abolished, almost 90 % of the appointed positions continued to be held by locals whose families had at one time held elected office in their colony. A shortcoming of this measure is that I cannot distinguish between elites last names and the last names of politicians who represented freedmen. Therefore, while this share displays a high degree of persistence, it is not clear whether this persistence differentially favored the planter elites. To better gauge this, I calculate a measure of the degree to which appointed legislators represented the plantation economy. To do this, I first take the 1834 slave-density of each parish from Higman [1995], then average over all parishes that each family had represented in elected parliament to get a family s 33 Under Crown rule, legislative chambers were appointed by the governor. Legislators names and appointment dates continued to be reported in the Blue Books. 34 A family is specific to a colony in this calculation, although some influential families owned plantations in several of the colonies (Taylor [2002]). 21

23 Table 7: Measured Elite Persistence Years Since Switch to Autocracy (Crown Rule): Share Appointees from Elected Families: Slave-Density of Appointee's Former Elective Parishes: (relative to equal representation) Note: This table reports two measures of persistence, in 3 year steps after the switch to Crown rule. Both measures are calculated by pooling all appointed legislators after the switch to Crown rule into one legislature and calculating shares so that all appointed legislators of all 11 colonies are represented in each year. "Appointees from Elected Families" are those whose family had held an assembly seat at any time before a colony's abolition of its assembly. To construct the second measure, I calculate for each parish its slave density (#slaves/area) in 1836, relative to the colony average. This proxies for the dominance of the plantation system in a parish, and is divided by the colony average to make it comparable across colonies. The table reports the average of this relative slave density across all appointed legislatures. implied slave-density. 35 For each year since the regime change, I then average this familyspecific measure over all appointed legislators across colonies. If this average is larger than 1, then plantation districts are systematically over-represented across colonies. Table 7 shows that this measure remains very stable at around 2.5 over the two decades after parliaments were abolished. This evidence suggests that the planters continued to wield disproportionate influence over policy making long after the regime changes. 5 Discussion 5.1 Alternative Responses While the evidence in sections 4.3 and 4.4 suggests that the regime switches did help the planter elites, it is nonetheless important to ask whether they could not have pursued other options to limit political competition. An obvious choice may have been changes in franchise rules, like those adopted in the U.S. South. 36 Caribbean planters did try this to some extent: Craton [1997, p. 392] argues that there were organized efforts to evict peasants from the land throughout the Caribbean, and McLewin [1987, p. 189] argues that assemblies brought into law an umbrella of coercive acts with the purpose of creating a landless peasantry. 37 There is also some evidence of electoral engineering: Holt [1991, p. 244] recounts how in Jamaica, in 1844, governor Elgin called for early elections to blunt the registration drive. Yet, when the new assemblymen convened, there 35 I normalize this measure by a colony s total slave-density to create a measure that is comparable across colonies. 36 See Kousser [1999] for a discussion of disenfranchisement in the post-reconstruction US South. 37 For example, crown land was priced to encourage labor for wages and was chiefly in remote locations and of poor quality (Bolland [1981]) and parochial land taxes pressed hard on small proprietors (McLewin [1987, p. 184]). 22

24 were five new colored faces among them. 38 In the end, however, Caribbean planters were constrained in their use both of overt coercion and of coercive legislation. For one, they did not have the sort of influence in London that Southern planters enjoyed in Washington. The colonial office was much more influenced by abolitionists in London than by Caribbean planters. As a result, the Crown would overrule discriminatory local acts with so-called orders-in-council (Craig-James [2000, p. 65]). 39 Moreover, planters could scarcely use brute force because their numbers were very low compared to, for example, the U.S. South. 40 The Caribbean lacked the white manpower for an equivalent of the Ku Klux Klan The Role of the Freedmen Why did the freedmen accept the regime change to autocracy? The historical record suggests that black political leaders were aware that abolishing parliament would stunt their political influence. 42 One possibility is that the freedmen s leaders were at least in part bought off with administrative positions (Honychurch [1995]). Another possibility is that the freedmen s leaders could not prevent it, because the majority of freedmen had genuinely positive expectations of Crown rule. This was reasonable given that the Crown had imposed the abolition of slavery. 43 Perhaps reflecting this, the switch to Crown rule did not trigger any reported unrest, although rioting was overall a fairly common form of resistance to oppression (Morrell and Parker [1969, p. 396], Dookhan [1977, p. 114]). 44 Given the evidence presented, why did freedmen not mobilize against Crown rule once it failed to generate favorable outcomes for them? One answer is provided by the social science literature on mobilization and the coordination of beliefs (Schelling [1980]). Because 38 To the extent that such efforts at election-timing were successful, measured electoral competition will underestimate true political competition, which would also lead to the OLS coefficients being smaller than the IV coefficients. 39 Lowes [1994, ch. 5] writes that because of pressures from the Colonial Office, a comfortable translation of preemancipation legal distinctions into distinctions of skin color was not possible. 40 In the South, a black county may have been 50 % black. By contrast, all of the Caribbean plantation colonies were more than 90 % black. 41 Another mechanism for elite control in many rural settings is the paternalism of landlord-tenant relationships (Baland and Robinson [2008]). This mechanism had no bite in the Caribbean precisely because the vote was tied to legal title to land, so that tenants were excluded from the electorate. 42 Hall [1959, p. 262] quotes a black assemblyman in Jamaica: You and I have been equals, but what will be the respective position of our children? Yours will hardly speak to mine. 43 Further, the three Caribbean colonies that had always been governed under Crown colony rule were perceived as being run more in the interests of the freedmen. Laurence [1971, p. 53] writes that conditions were much better there, as planters never enjoyed the same influence over local government. 44 The 1865 Morant Bay Rebellion was caused by the imprisonment of squatters on an abandoned plantation. It led directly to Jamaica s switch to Crown Colony rule, which actually calmed the unrest (Dookhan [1977, p. 65]). 23

25 slavery had been specifically designed to undermine such coordination, post-abolition freedmen did not have many social ties that reached past the individual plantation, making mobilization difficult (Chay and Munshi [2011]). While this obstacle may have been overcome in the face of overt oppression, the mere absence of positive change may not have provided the necessary impetus required for mobilization (Hacker and Pierson [2011]). Furthermore, freedmen may not have associated the slow stagnation in public-good provision with the regime switch at all, given the evidence in section 4.3 that there was a temporary increase in public good provision around the time of the regime change. 5.3 The Role of the Crown The identification strategy does not allow me to rule out entirely the possibility that the Crown was a driving factor behind the abolition of parliament. The Crown clearly did want more control over its colonies, and it is possible that,it put differential pressure for Crown rule on colonies where local elites were losing political control. However, the Crown also clearly had objectives that ran counter to those of the elites. It had imposed abolition against planters protests, and before the regime switches, many planters expressed grave concern that the Crown would side with the interests of the freedmen by spending on public goods (Will [1970, p. 289], Dookhan [1977, pp , 80]). These concerns were echoed in the Crown s stated intentions to improve local public goods and to develop an independent smallholder society (Wrong [1923, pp ]; Burroughs [1999, pp ]). 45 Yet, the evidence in Table 7 clearly shows that the regime changes did not have these effects. Overall, the historical record suggests that where the Crown wanted policies that ran counter to the elites interests, its ability to implement these policies was curtailed by the reality on the ground. This could simply mean for colonial administrators to be socially ostracized. Lewis [2004, p. 104] writes of the governors incentives that to join with local white society meant a pleasant tour of duty, to fight them meant political conflict and social ostracism. Inevitably, the governor passed smoothly into the union, political and social, of government and vested interests. Or it could mean more overt push-back. Craig-James [2000, p. 252] recounts a confrontation between 45 Henry Taylor, the colonial office s supervisor of West Indian affairs, publicly chastised the assemblies for being eminently disqualified for the great task of educating and improving a people newly born to freedom (Wrong [1923]). 24

26 Table 8: Variation in Democracy across the 19th Century British Empire (1) (3) (2) (4) Dependent: Repr. Government 1832 Repr. Government 1882 populations-share slaves (1834) 0.004* *** (2.037) (-8.731) Distance-to-Equator *** (-0.648) (7.927) Constant 0.468*** 0.791*** 1.025*** ** (3.496) (4.042) (13.166) (-2.645) Observations R-squared Representative Govt. North American Settler Colonies: 5/6 6/6 Representative Govt. Australasian Settler Colonies: 0/5 5/5 Representative Govt. Caribbean: 14/17 3/17 Note: The 6 North American colonies are Lower and Upper Canada, New Brunswick, New Foundland, Novia Scotia, Prince Edward Island. The 5 Australasian colonies are New Zealand, Tasmania, West Australia, South Australia and New South Wales. The 17 Caribbean colonies are Antigua, the Bahamas, Bermuda, Barbados, Br. Honduras, Jamaica, Montserrat, Nevis, St. Kitts, the Virgin Islands, Dominica, Tobago, St. Vincent, Grenada, Trinidad, St Lucia and Br. Guyana. Tobago governor and the locally appointed administrators. The administrators stopped the governor s planned progressive land tax reform by tieing it to a salary cut to the governor. Whatever the Crown s motivation, the fact is that the 11 regime changes studied in this paper were explicitly initiated by and needed the support of the incumbent local elites to become law. 5.4 Changing Colonial Institutions and the Colonial Origins Literature The basic argument of the colonial origins literature is that initial conditions (population densities, settler mortality, geographic suitability for certain crops) determined the institutions that European settlers set up, and that these initial institutional differences persisted to determine both present-day institutions and income levels (Acemoglu et al. [2001, 2002], Engerman and Sokoloff [2002]). Figure 2 shows the basic argument. It compares the British Caribbean colonies with Britain s settler colonies in North America and Australasia in the core samples of Acemoglu et al. [2001, 2002]. The Caribbean colonies were denser and had higher settler mortalities, and they consequently have lower incomes and (not shown in Figure 2) worse institutions today. Despite this argument s influence on economists thinking about long-run economic development (Nunn [2009]), data on concrete colonial institutions has remained elusive. Table 8 extends the data on colonial representative institutions to the settler colonies of North 25

27 Figure 2: British Plantation vs Settler Colonies in the AJR 2001 and 2002 data American and Australasia, and reveals substantial changes in colonial institutions in the second half of the 19th century. In 1832, the year in which the Reform Act extended the franchise to the poor in England, 14 of the 17 Caribbean colonies, but only 5 of the 11 British settler colonies, had elected parliaments. 46 Fifty years later, only 3 Caribbean colonies had retained their parliaments, while all 11 settler colonies now had parliaments. Table 9 further illustrates this variation by showing that in 1832, having representative institutions actually correlated positively with the intensity of slavery and negatively with the distance to the equator, two variables that commonly correlate in the opposite way with good institutions. Fifty years later, these correlations had reversed. This variation highlights the fact that the colonial origins argument is really about the bundle of de jure and de facto institutions (Acemoglu and Robinson [2008]). While the Caribbean colonies may have de jure appeared more democratic than the settler colonies at some point in time, these democratic institutions served only a narrow elite and the majority of the population most likely enjoyed less de facto representation than in the settler colonies, even in I consider the Bahamas and Bermuda as part of the Caribbean here, although they were not plantation colonies and had white population majorities. 26

28 6 Conclusion This paper presents evidence that elites in non-inclusive democracies may prefer to initiate a regime change to autocracy rather than seeing their political influence become eroded as democracy becomes more broad-based. Because the coups that drive most regime changes are illegal, who supported them is usually difficult to observe even ex post. This paper s approach is therefore to focus on data on a series of legal regime changes from democracy to autocracy, which occurred in in the 19th-century British Empire, where local parliaments could vote themselves out of existence and replace themselves with legislative bodies that would be appointed autocratically by the colonial administration. The paper s key findings are that (i) increasing political competition from freed slaves explained the instance and timing of these voluntary switches to autocracy, and (ii) that autocracy shielded entrenched elites from the political pressures they were exposed to under democratic rule. Overall, this paper highlights an unexplored mechanism for why democracies become autocracies and shows that elites that emerged in very limited democracies may be better able to secure their interests by switching to autocracy. References Acemoglu, D., S. Johnson, and J. Robinson (2001). The Colonial Origins of Comparative Development: An Empirical Investigation. The American Economic Review. Acemoglu, D., S. Johnson, and J. Robinson (2002). Reversal of Fortune: Geography and Institutions in the Making of the Modern World Income Distribution. The Quarterly Journal of Economics 117(4), Acemoglu, D. and J. Robinson (2000). Why did the West Extend the Franchise? Democracy, Inequality, and Growth in Historical Perspective. Quarterly Journal of Economics 115(4), Acemoglu, D. and J. Robinson (2001). A Theory of Political Transitions. The American Economic Review 91(4), Acemoglu, D. and J. Robinson (2006). Economic Origins of Dictatorship and Democracy. Cambridge University Press. Acemoglu, D. and J. Robinson (2008). Persistence of Power, Elites, and Institutions. American Economic Review 98(1), Ashdown, P. (1979). Caribbean History in Maps. Addison-Wesley Longman Ltd. 27

29 Baland, J.-M. and J. A. Robinson (2008). Land and Power: Theory and Evidence from Chile. The American Economic Review, Banerjee, A. and L. Iyer (2005). History, Institutions, and Economic Performance: The Legacy of Colonial Land Tenure Systems in India. The American Economic Review 95(4), Besley, T. and A. Case (1995). Does Electoral Accountability Affect Economic Policy Choices? Evidence from Gubernatorial Term Limits. The Quarterly Journal of Economics 110(3), Besley, T., T. Persson, and M. Reynal-Querol (2012). Political Instability and Institutional Reform: Theory and Evidence. London School of Economics (unpublished). Bobonis, G. and P. Morrow (2014). Labor Coercion and the Accumulation of Human Capital. Conditionally accepted, Journal of Development Economics. Boix, C. (2003). Democracy and Redistribution. Cambridge University Press. Bolland, O. (1981). Systems of Domination After Slavery: The Control of Land and Labor in the British West Indies after Comparative Studies in Society and History 23(04), Britain, C. O. G. (1879). The Colonial Office List. HM Stationery Office. Brizan, G. (1984). Grenada, Island of conflict: from Amerindians to People s Revolution, Zed Books London. Brückner, M. and A. Ciccone (2011). Rain and the Democratic Window of Opportunity. Econometrica 79(3), Burroughs, P. (1999). Imperial Institutions and the Government of Empire. The Oxford History of the British Empire 3, Cameron, A. C. and D. L. Miller (2013). A Practitioner s Guide to Cluster-Robust Inference. Cartwright, J. (1970). Politics in Sierra Leone University of Toronto Press Toronto. Chay, K. and K. Munshi (2011). Slavery s Legacy: Black Mobilization in the Postbellum South. Manuscript, Brown Univ. Craig-James, S. (2000). The Social and Economic History of Tobago ( ). Craton, M. (1997). Empire, Enslavement, and Freedom in the Caribbean. James Currey. Dell, M. (2010). The Persistent Effects of Peru s Mining Mita. Econometrica 78(6), Dippel, C., A. Greif, and D. Trefler (2014). The Rents from Trade: Removing the Sugar Coating. Working Paper, UCLA. Dookhan, I. (1977). A Post-Emancipation History of the West Indies. Collins. 28

30 Dunn, R. and J. Parker (1972). Sugar and Slaves: The Rise of the Planter Class in the English West Indies, History: Reviews of New Books 1(1), 4 4. Dutt, P. and D. Mitra (2008). Inequality and the Instability of Polity and Policy. The Economic Journal 118(531), Engerman, S. (1982). Economic Adjustments to Emancipation in the United States and British West Indies. Journal of Interdisciplinary History 13(2), Engerman, S. (1984). Economic Change and Contract Labor in the British Caribbean: The End of Slavery and the Adjustment to Emancipation. Explorations in Economic History 21(2), Engerman, S. and K. Sokoloff (2002). Factor Endowments, Inequality, and Paths of Development Among New World Economics. National Bureau of Economic Research. Engerman, S. L. and K. L. Sokoloff (2005). The Evolution of Suffrage Institutions in the New World. The Journal of Economic History 65(04), Feroz, A. (2002). The Making of Modern Turkey. Routledge. Feyrer, J. and B. Sacerdote (2009). Colonialism and modern income: islands as natural experiments. The Review of Economics and Statistics 91(2), Fuller, T. (2014). Thai Army Declares Coup, Citing Need to Reform Nation. New York Times May 22nd. Hacker, J. S. and P. Pierson (2011). Winner-Take-All Politics: How Washington Made the Rich Richer and Turned Its Back on the Middle Class. Simon and Schuster. Haggard, S. and R. R. Kaufman (1997). The Political Economy of Democratic Transitions. Comparative Politics, Haggard, S. and R. R. Kaufman (2012). Inequality and Regime Change: Democratic Transitions and the Stability of Democratic Rule. American Political Science Review 1(1), Hall, D. (1959). Free Jamaica, : an economic history, Volume 1. Yale University Press. Henry, P. B. and C. Miller (2009). Institutions versus Policies: A Tale of Two Islands. American Economic Review Papers & Proceedings. Higman, B. (1995). Slave Populations of the British Caribbean, Univ of West Indies Pr. Holt, T. (1991). The Problem of Freedom: Race, Labor, and Politics in Jamaica and Britain, The Johns Hopkins University Press. Honychurch, L. (1995). The Dominica Story: A History of the Island. Macmillan Pub Limited. 29

31 Husted, T. A. and L. W. Kenny (1997). The Effect of the Expansion of the Voting Franchise on the Size of Government. Journal of Political Economy, Iyer, L. (2010). Direct versus Indirect Colonial Rule in India: Long-Term Consequences. The Review of Economics and Statistics 92(4), Kassimeris, C. (2006). Causes of the 1967 Greek Coup. Democracy and Security 2(1), Kousser, J. (1974). The Shaping of Southern Politics: Suffrage Restriction and the Establishment of the One-Party South, Yale University Press New Haven. Kousser, J. (1999). Colorblind Injustice: Minority Voting Rights and the Undoing of the Second Reconstruction. University of North Carolina Press. Lange, M. (2004). British Colonial Legacies and Political Development. World Development 32(6), Laurence, K. (1971). Immigration into the West Indies in the 19th Century. Caribbean Universities Press (Barbados and Aylesbury, Bucks, UK). Lewis, G. (2004). The growth of the Modern West Indies. Ian Randle Publishers. Lipset, S. (1959). Some Social Requisites of Democracy: Economic Development and Political Legitimacy. American Political Science Review 53(1), Lizzeri, A. and N. Persico (2004). Why Did the Elites Extend the Suffrage? Democracy and the Scope Of Government, with an Application to Britain s Age Of Reform. Quarterly Journal of Economics 119(2), Lowes, S. (1994). The Peculiar Class: The Formation, Collapse, and Reformation of the Middle Class in Antigua, West Indies, Columbia University. Margo, R. A. (1990). Race and Schooling in the South, : an Economic History. University of Chicago Press. McLewin, P. (1987). Power and Economic Change: The Response to Emancipation in Jamaica and British Guiana, Garland. Meltzer, A. and S. Richard (1981). A Rational Theory of the Size of Government. The Journal of Political Economy, Meyersson, E. (2014). Political Man on Horseback. Military Coups and Development. working paper, SITE. Morrell, W. and W. Parker (1969). British Colonial Policy in the Mid-Victorian Age: South Africa, New Zealand, the West Indies. Clarendon P. 30

32 Naidu, S. (2009). Suffrage, Schooling, and Sorting in the Post-Bellum US South. working paper, Columbia University. North, D. C. and B. R. Weingast (1989). Constitutions and Commitment: the Evolution of Institutions Governing Public Choice in Seventeenth-Century England. The journal of economic history 49(04), Nunn, N. (2009). The Importance of History for Economic Development. Annual Review of Economics 1, Przeworski, A., M. Alvarez, J. A. Cheibub, and F. Limongi (2000). Democracy and Development: Political Regimes and Economic Well-Being in the World, New York: Cambridge University Press. Ragatz, L. (1928). The Fall of the Planter Class in the British Caribbean, : a Study in Social and Economic History. The Century Co. Roby, J. (1831). Members of the Assembly of Jamaica from the Institution of that Branch of the Legislature to the Present Time, Arranged in Parochial Lists. Published by Alex holmes, Montego-bay, Jamaica. Schelling, T. C. (1980). The Strategy of Conflict. Harvard university press. Sewell, W. (1861). The Ordeal of Free Labor in the British West Indies. Harper & brothers. Taylor, A. (2002). The American Colonies, Volume 1. Penguin Group USA. Trebbi, F., P. Aghion, and A. Alesina (2008). Electoral Rules and Minority Representation in U.S. Cities. The Quarterly Journal of Economics 123(1), Will, H. (1970). Constitutional Change in the British West Indies, : with Special Reference to Jamaica, British Guiana, and Trinidad. Clarendon press. Wrong, H. (1923). Government of the West Indies. Clarendon Press. 31

33 APPENDIX TABLE 1 Colony & Year & Autocracy & Election & Elect. Turnover & log(voters) Antigua & 1850 & 0 & & &. Antigua & 1851 & 0 & & &. Antigua & 1852 & 0 & & &. Antigua & 1853 & 0 & Yes & & Antigua & 1854 & 0 & & & Antigua & 1855 & 0 & & & Antigua & 1856 & 0 & & & Antigua & 1857 & 0 & & & Antigua & 1858 & 0 & & & Antigua & 1859 & 0 & & & Antigua & 1860 & 0 & Yes & & Antigua & 1861 & 0 & & & Antigua & 1862 & 0 & & & Antigua & 1863 & 0 & & & Antigua & 1864 & 0 & & & Antigua & 1865 & 0 & & & Antigua & 1866 & 0 & & & Antigua & 1867 & 1 & Yes & & Figure Appendix 1: The Blue Books Assembly Data (Barbados 1847) 32