Christian von Soest ab & Michael Wahman c a Institute of African Affairs, GIGA German Institute of

Transcription

1 This article was downloaded by: [Harvard Library], [Christian von Soest] On: 02 April 2014, At: 06:58 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK Democratization Publication details, including instructions for authors and subscription information: Are democratic sanctions really counterproductive? Christian von Soest ab & Michael Wahman c a Institute of African Affairs, GIGA German Institute of Global and Area Studies, Hamburg, Germany b Weatherhead Center for International Affairs, Harvard University, Cambridge, MA, USA c Department of Government, London School of Economics and Political Science, London, UK Published online: 31 Mar To cite this article: Christian von Soest & Michael Wahman (2014): Are democratic sanctions really counterproductive?, Democratization, DOI: / To link to this article: PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content ) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-

2 licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at

3 Democratization, Are democratic sanctions really counterproductive? Christian von Soest a,b and Michael Wahman c a Institute of African Affairs, GIGA German Institute of Global and Area Studies, Hamburg, Germany; b Weatherhead Center for International Affairs, Harvard University, Cambridge, MA, USA; c Department of Government, London School of Economics and Political Science, London, UK (Received 11 September 2013; final version received 20 January 2014) Previous research has shown that sanctions have a negative impact on the level of democracy in targeted authoritarian countries. This runs counter to substantive comparative literature on democratization which finds that economic stress is connected with regime collapse and democratic liberalization. To solve this puzzle, we focus on the effects of democratic sanctions (those that explicitly aim to promote democracy) which have become the most common type of sanction issued against authoritarian states. We introduce a new data set of imposed sanctions in the period that clearly separates sanctions according to the explicit goal of the sender. Our cross-sectional time-series analysis demonstrates that although sanctions as a whole do not generally increase the level of democracy, there is in fact a significant correlation between democratic sanctions and increased levels of democracy in targeted authoritarian countries. A fundamental mechanism leading to this outcome is the increased instability of authoritarian rule as democratic sanctions are significantly associated with a higher probability of regime and leadership change. Keywords: sanctions; authoritarianism; democratization; persistence; regime change; ruler exit Introduction Instigating democratization has been by far the most common goal of sanctions initiated by the United States, the European Union and the United Nations against authoritarian regimes in the post-cold War period. However, sanctions have a bad reputation, especially as an instrument for democratization. Not only have most sanction scholars concluded that sanctions are generally ineffective, 1 Peksen and Drury 2 have even argued that they have an adverse effect on the level of democracy in targeted countries. Further analyses have found that Corresponding author. christian.vonsoest@giga-hamburg.de # 2014 Taylor & Francis

4 2 C. von Soest and M. Wahman sanctions increase the pressure on autocratic rulers and consequently trigger the incentive to increase repression. 3 This background urges the critical question: Has one of the West s most important tools for promoting democracy externally been nothing but counterproductive? Our current knowledge on democratization and sanctions derives from two separate sets of research; a first strand of research that focused on the effectiveness of sanctions more broadly 4 and a second set of more quantitative literature that has focused exclusively on the relationship between sanctions and democratization. 5 This first set of literature was largely divorced from comparative democratization theory, and sanction success was measured as a discrete outcome on a case-to-case basis in relation to the explicit goal of the sender. In the second, more recent, strand of literature, a quantitative method with a continuous approach to measuring the dependent variable was adopted. Although this literature improved substantially on the measurement of the dependent variable, the democratic effect of sanctions was not related to the goal of the senders. Hence, the question of effectiveness was largely left aside and we still know very little about whether democratic sanctions actually have a positive effect on the level of democracy in targeted authoritarian states. Previous research on the democratic effects of sanctions have explained the adverse effects of sanctions with a tendency among targeted countries to increase repression to cope with higher levels of domestic pressure. However, we argue that the theoretical expectations derived from the literature on sanctions and comparative democratization are less clear-cut than previously suggested. As the comparative democratization literature has shown, economic stress is one of the most robust determinants of democratization and/or regime change in authoritarian regimes; 6 there are good reasons to believe that well designed sanctions aimed at instigating political change could potentially lead to policy concessions or regime destabilization and ultimately increase levels of democracy. In this study we empirically address the question of democratic sanction effectiveness, by looking especially at those sanctions designed to instigate democratization in the targeted authoritarian country. We adopt a continuous approach to democracy and go beyond the question of whether sanctions fail or succeed in instigating regime change. Additionally, we study the relationship between democratic sanctions and the stability of authoritarian institutions and leaders. To enable this analysis we introduce a new data set of EU, UN and US sanctions against nondemocratic regimes in the period The newly compiled data set improves on earlier sources by clearly distinguishing sanction episodes according to their explicit goals. In the analysis, we focus specifically on the post-cold War period as Western liberalism s triumph 7 in the 1990s meant that the US and its allies became far more assertive in actively promoting democracy at the global level. 8 The results emerging from this analysis are significantly more optimistic than prior assessments with regard to the prospects of achieving democratization from international sanctions. 9 Imposed sanctions, in general, are not associated with higher levels of

5 Democratization 3 democracy. However, when looking only at sanctions that aim to democratize in the post-cold War era, they do actually have a significantly positive effect on the target state s level of democracy. Moreover, supplementary tests show a significant relationship between democratic sanctions and an increased probability of regime change and leader exit. This article starts by discussing the importance of sanction goals for democratic outcomes. We then present the data, particularly the new data set on sanctions, and research design in the third section, followed by the empirical results and the supplementary test on democratic sanctions and regime stability in the fourth and fifth sections. The conclusion discusses the implications of our findings. Sanctions and their effect on democratization Research into the general effectiveness of sanctions has concluded that sanctions are rarely successful. Hufbauer et al., 10 who compiled the most widely used data set on economic sanctions, found that these coercive measures attained their goal only in one-third of all cases. 11 Sanctions with the specific aim of regime destabilization or democratization show a similar success score of 31%. 12 However, these findings are heavily contested. Most authors 13 maintain that even this low success rate of sanctions is overstated, as Hufbauer et al. s bivariate approach does not allow for the assessment of other plausible explanations for the changes in target states behaviour. 14 The quantitative literature on sanctions and its democratic consequences has improved on earlier research by studying democracy as a continuous rather than dichotomous concept. 15 In this way the measurement of the dependent variable improved drastically. As scholars of democratization have acknowledged, democratic transformation is a long-term process often achieved in small, incremental steps. 16 Measuring only the actual transition to democracy seriously truncates the dependent variable and entails the risk of missing improvements in the most autocratic contexts. Although a sanction episode might not lead to a fullyfledged transition to democracy, it may well contribute to liberalizing outcomes. 17 Although later quantitative contributions to the field improved on the measurement of the dependent variable (democratization), they abandoned the independent variable and research question of earlier sanction studies. In order to study the effectiveness of sanctions, sanction outcomes must be related to the goals they set out to achieve. However, when studying the democratic effect of sanctions, Peksen and Drury 18 do not separate the sanction episodes that actually aim to instigate democratization. For instance, no distinction is made between sanctions aimed at ending nuclear proliferation and those designed to pressure authoritarian states to hold elections. Consequently, the question of effectiveness is left unanswered. However, earlier studies have provided theoretically attractive reasons to believe that democratic sanctions might be counterproductive. Using a classic political economy understanding of authoritarian stability, 19 it has been argued that

6 4 C. von Soest and M. Wahman authoritarian governments, strained economically by sanctions, tend to resort to increased repression to maintain stability. 20 Although the repression-focused argument is highly feasible, it runs counter to some prominent arguments found in the democratization literature outside the specific field of sanctions research. It is now widely held that economic downturns tend to destabilize authoritarian regimes, by fusing regime splits and popular uprisings. 21 Authoritarian regimes tend to survive with a mixed strategy of repression and co-optation. 22 Without the financial ability to co-opt counter elites, authoritarian regimes often resort to a strategy of increased repression. However, repression is generally a less efficient tool for regime survival in the long term than co-optation, 23 as it is an imprecise instrument often leading to an increase in both opposition support and levels of dissent. 24 Hence, economic downturn often results in regime accommodation, or even more dramatically regime collapse. In a seminal article, Bueno de Mesquita and Smith 25 showed that regimes that rely on small winning coalitions (that is, most authoritarian regimes) have a significantly increased risk of regime collapse when growth is low. Moreover, regimes with small winning coalitions are also more likely to experience mass mobilization in connection with poor economic performance. 26 When economic downturns are caused by sanctions associated with democratic goals, we can expect sanctions to send particularly strong signals to opposition groups and positively affect their chances of mobilizing the masses in demand for political reform. Regime change does not necessarily lead to democratization, but as shown by Teorell, 27 economic decline is one of the most prominent determinants of short-term democratization. Given the ambiguous state of the literature on democratization and sanctions, as well as the empirical gap left by the earlier sanctions literature, we find it relevant to reassess the effectiveness of democratic sanctions. In doing so, we will combine the earlier sanction research s focus on effectiveness with the later quantitative research s understanding of democratic change as a continuous variable that can take on both negative and positive values. Although the current literature on comparative democratization makes it plausible that sanctions could have a positive effect on the level of democracy in the targeted country, this is not to say that all sanctions necessarily have this effect. As several authors have acknowledged, 28 senders have used a multitude of sanction designs to achieve their desired goals. Sanctions vary in intensity, whom they target, and their general comprehensiveness, and senders design sanctions in relation to the goals they set out to achieve. It is now widely accepted that the naïve theory of sanctions, suggesting that effectiveness is dependent on the total cost sanctions afflict on the target economy, was generally misguided. 29 In line with the general trend, sanctions which aim at instigating democratization have become more focused and targeted. 30 It is therefore reasonable to expect that democratic sanctions with their regular focus on autocratic regime elites are more effective in instigating democratization than sanctions designed to achieve

7 Democratization 5 alternative goals. It is also not at all evident that sanctions aimed at goals other than democratization would have the unexpected and in some cases even undesired effect 31 of causing regime change. Given this theoretical discussion, we hence formulate our hypotheses in conflict with the current state of the sanctions literature: H1a: Democratic sanctions increase the level of democracy in authoritarian regimes. H1b: Democratic sanctions increase the level of democracy in authoritarian regimes more than those sanctions aiming to achieve other goals. To complement the main analysis, we will also study how democratic sanctions correlate with leadership survival and institutional change in authoritarian states in order to corroborate the claim that democratic sanctions weaken authoritarian rulers and thereby strengthen democracy levels in targeted regimes in the long term. 32 If sanctions destabilize authoritarian regimes, they are also likely, on average, to increase the level of democracy. We therefore also test the following hypotheses: H2: Democratic sanctions increase the likelihood of an irregular change in leadership. H3: Democratic sanctions increase the likelihood of a change in regime type. Threats to use democratic sanctions So far we have talked about the effectiveness of actually imposed sanctions. A number of authors have suggested that studying imposed sanctions is inadequate and that research must also take into account episodes where there are mere threats of sanctions. Leaving out those episodes, where sanctions did not proceed to the imposed stage, could potentially underestimate the effectiveness. 33 From our perspective, it is important to acknowledge that different sanction goals are likely to be associated with varying success rates at the threat stage. Drury and Li 34 found that targets are more likely to give in to sanction threats when the goals are minor 35 and thus unlikely to threaten their rule. In contrast, democratic sanctions entail major goals associated with higher costs of concession for the target government in that they try to weaken the authoritarian grip on power. The data from the Threat and Imposition of Sanctions (TIES) data set 36 confirm the theoretical expectation that prospective targets seldom respond positively to sanction threats associated with democratic demands. 37 Our analysis using the TIES data demonstrates that democratic and human rights sanction threats 38 are rarely effective. Only 10% of sanctions related to democracy or human rights accomplished complete or partial concession by the target country at the threat stage. The corresponding number for sanctions with other goals is more than double (22%). Moreover, sanction senders are generally serious about their democratic sanctions. Where targets did not comply, 85% of all democratic sanction threats were carried out, compared to 79% for the rest (see Table 1).

8 6 C. von Soest and M. Wahman Table 1. Threatened democratic sanctions vs. other sanctions. Goal Successful threats (%) Non-credible threats (%) Non-democratic and non-hr Democratic and HR Note: Table shows data from the TIES data set. 91 Data include threats of sanctions against all countries classified as authoritarian by Hadenius, Teorell, and Wahman, 92 issued by the EU, UN, or US. The results in Table 1 are hardly surprising as there is no sanction category that introduces a higher cost of concession for authoritarian regimes than those demanding democratization. 39 The political costs are normally detrimental to a military regime that introduces civilian rule, a one-party regime that allows political opposition, or an electoral authoritarian regime that arranges free and fair elections. Our results also imply that autocratic rulers are regularly not successful in placating potential sanction senders and averting the imposition of threatened democratic sanctions through piecemeal reforms (for example, flawed elections, referendums, or assurances that civil rights will be protected in the future). Given that autocrats resist almost all democratic sanction threats and that senders generally pursue these sanctions, it follows that we see a high rate of democratic sanctions actually implemented. This insight stresses the importance of analysing imposed democratic sanctions with more refined data. Although democratic sanctions might be less effective at the threats stage than sanctions aimed at other goals, this is not necessarily the case when looking at imposed sanctions. Data and research design Methodologically, this study utilizes cross-sectional time-series analysis to investigate how sanctions affect democratization in authoritarian regimes. As discussed above, we understand democracy as a continuum rather than a dichotomy. Thus, our dependent variable measures the gradual change in the level of democracy. To account for these, we utilized a combined Freedom House and Polity IV score. Following Hadenius and Teorell, 40 the combined score compensates for the individual shortcomings of the two indices. 41 It is derived by computing the average Freedom House political rights and civil liberties score (reversed and transformed to a 0 10 scale) and combining it with the revised combined autocracy and democracy score from the Polity IV data (also converted to a 0 10 scale). The change in the level of democracy is calculated as demo t -demo t21. As we are also interested in the dynamics of democratic sanctions, we analysed in a supplementary test whether democratic sanctions destabilize authoritarian regimes, and whether they do so to a greater degree than other sanctions. This sheds light on how democratic sanctions work in authoritarian regimes and influence democratization. We ran a series of conditional fixed-effects logistic regressions. One

9 Democratization 7 set of these models uses leader exit 42 as the dependent variable, and one uses regime change as recorded in the Hadenius, Teorell, and Wahman 43 regime data set. 44 All authoritarian regimes that existed in the period are included in the sample. According to Hadenius and Teorell s 45 classification, a country is regarded as authoritarian in a certain year if its democracy score is lower than 7.5 on the 10-point index described above. The sample does not include any democratic regimes, except those that entered the democratic regime category during a sanction episode. 46 All in all, the sample consists of 2079 country years; sanctions against an authoritarian regime were imposed in 532 of these country years. Introducing a new data set on sanctions In order to assess the effects of democratic sanctions in comparison to measures pursuing other goals we had to clearly differentiate various demands connected to sanctions. To achieve this, we had to go beyond the commonly used data from Hufbauer et al. 47 and constructed a new data set. First and most fundamentally, our data set divides all goals into coherent categories and clearly separates democratic sanctions from other demands. Adapting the model proposed by Charron, 48 we classified the goals of sanctions according to broad categories (see Table 2). Table 2 also specifies the objectives in the imposing documents or senders statements which we classify as democratic demands. We acknowledge that as recurrently claimed by sanction targets these stated goals of senders might not always reflect the real motivations behind a sanction. However, taking the goals at face value makes for a conservative test of the effect of democratic sanctions. If we expect sanctions to be more effective Table 2. Democratic and non-democratic goals of sanctions. Democratization-related demands (1) Holding of new elections (2) Modification of the constitution or the electoral code (3) Admission of an international electoral observation mission (4) Restoration of a democratically elected leader (5) Recognition of electoral results (6) Recognition of rights and freedoms directly linked to the electoral process (such as freedom of assembly and expression) (7) Protection of human rights connected to electoral competition (8) Regime change Other demands (1) Support for human rights (all human rights which are not directly related to the democratic process) (2) Fighting of narcotics (3) Ending of the proliferation of weapons of mass destruction (WMD) (4) Termination of bellicosities or establishment of peace agreements (5) Fighting of terrorism Source: Authors compilation.

10 8 C. von Soest and M. Wahman in instigating democratization when they are aimed at this specific goal, the coefficient for this sanction category would only be smaller if a number of the sanctions within this category are actually designed to accomplish other non-explicit goals. In addition, we had to depart from Hufbauer et al. s coding of the ending of sanction episodes, as for our analysis it is important to demarcate the end of an episode as the point not only when the sanctions are lifted but also when the sender s goal changes substantially. Third, the data set runs all the way to 2010, whereas the Hufbauer et al. data set effectively stops assessing sanction cases as of Also, even the assessment for the cases before this date is often inconclusive. Finally, going beyond existing large-n research, which has focused almost exclusively on economic sanctions, all coercive measures that potentially exert economic and political effects on targeted autocracies are understood as sanctions. These are economic (for example, restrictions on exports and imports, aid sanctions), political (for example, restrictive measures to enforce other sanctions), military (for example, arms embargoes) and diplomatic (for example, the recalling of ambassadors) sanctions. 49 The data set is composed of the entire universe of sanction regimes imposed by the UN, the US, and the EU (the main sanction senders) in the period , including those sanction regimes that were already in place in 1990 and those episodes that are still ongoing (see Online Appendix). 50 Given that the investigation explores the impact of sanctions on authoritarian regimes, the data set only includes countries that the Hadenius, Teorell, and Wahman data set 51 has coded as being non-democratic regimes at the start of the sanction episode. We also excluded measures related to trade disputes. 52 The imposition of these measures follows a trade policy rationale and their use is for World Trade Organization (WTO) member states regulated under the WTO framework. Often, senders use these trade-related measures solely in retaliation for other trade measures initiated by the target. 53 Data were obtained from the websites of the United Nations, UN Security Council Sanctions Committees, US Treasury, Office of Foreign Assets Control, US State Department, Directorate of Defense Trade Controls, and the European Commission, External Relations. These data were complemented by the Hufbauer et al. data set 54 and the TIES data set. 55 Further major bibliographic sources were Charron, 56 Crawford, 57 Portela, 58 Reynolds and Wan 59 and Taylor, 60 as well as media reports. 61 As outlined, the coding of sanction episodes was done according to the coding rules laid out in the codebook, with specific focus on the differentiation of sanction goals and the fine-grained assessment of democratic sanction demands. In that, the categorization is similar but at times deviates from the two other data sets, particularly the TIES data set. Based on the laid out rules, coding was done for all sanction episodes by three coders; which was followed by a final round of adjustment in cases where coders disagreed (ca. 20% of cases). For our comprehensiveness measure, we differentiated dichotomously between comprehensive and targeted sanctions. We define comprehensive sanctions as those where individuals other than members of the autocratic ruling elite

11 Democratization 9 Table 3. Comprehensive and targeted sanctions. Comprehensive sanctions (1) Aid sanctions (2) Commodity embargoes (3) Flight bans (4) Financial sanctions (5) Comprehensive trade embargoes Targeted sanctions (1) Targeted sanctions, affecting only designated (2) individuals and entities (3) Sanctions directed towards the military Diplomatic sanctions Source: Authors compilation. ( innocents ) are affected. 62 Table 3 shows which sanction types are coded as comprehensive and which as targeted measures. We transformed the data set into a cross-sectional time-series format that includes all authoritarian regimes for the period The main independent variables in the analysis are (1) whether a country was subjected to sanctions in a given year and (2) whether sanctions with explicitly democratic goals were imposed. Several senders may sanction a target in the same year, sometimes with varying goals. All years where at least one sender sanctioned a country (demanding democratization) were coded as being exposed to democratic sanctions. Figure 1 shows the number of ongoing sanctions, according to goal, for every year of our data set. The figure suggests that the number of democratic sanctions increased dramatically short after the end of the Cold War, but declined somewhat in the mid-2000s. The number of democratic sanctions did, however, increase again in the last years of our sample and democratization remains the most common goal of international sanctions. Both the EU and the US have used democratic sanctions frequently. EU democratic sanctions were in place in 169 of the country-years in our data set and the US had issued democratic sanctions against the authoritarian regimes of our data set in 267 country-years. For 105 of the country-years a country was targeted by both EU and US democratic sanctions. Control variables The issuing of sanctions comes with costs not only for the target but also for the senders for instance, higher economic costs through export and import losses. Senders are thus less likely to introduce sanctions against richer and bigger countries and/or those well integrated into the global economy. 64 Moreover, there is a possibility that democratic sanctions are more likely to be targeted against unstable regimes, where they are expected to be effective. We therefore included a number of controls in the statistical models to account for this potential selection bias. Boix 65 argues that smaller countries are generally more likely to democratize. In smaller countries, capital holders are less afraid of democratization since fixed

12 10 C. von Soest and M. Wahman Figure 1. Number of sanctions according to sanction goals. Source: Representation based on own data set. assets can more easily be transferred out of the country. Our models thus include a measure of logged population size. 66 The economic integration or linkage of countries with the world market 67 might be a powerful predictor of how likely it is that a country will be subjected to sanctions in the first place, 68 as sanctions on countries with high foreign direct investment (FDI) inflows and trade levels are more costly for sanction senders than those on less economically integrated countries. FDI inflows and trade also point towards the stability of a country, with higher levels of FDI and trade expected to indicate institutional strength and high levels of regime stability, 69 which in turn might constitute a disincentive to impose democratic sanctions. We used the data from the World Development Indicators (WDI) to account for FDI and trade in the models. In accordance with classical modernization theory, we expect a positive relationship between the level of economic development and the level of democracy. 70 The level of development may also influence the sanction sender s behaviour as they may be strategically more inclined to issue sanctions against targets where they expect the chances of future democratization to be high and those of fallback to be low. We thus incorporated a measure for GDP/capita (WDI) in all models. 71 We also included a measure for civil wars, since sanctions are often used to pressure parties in civil wars to stop internal conflicts. Civil wars might influence

13 Democratization 11 sender behaviour in that a sender is more likely to impose democratic sanctions on civil war countries given that these conflicts constitute critical junctures for democratization processes. For this variable we used data from the UCDP/PRIO data set. 72 An even more immediate signal to potential senders is provided by the occurrence of mass protests. 73 In these cases, sanctioners can rationally expect that their measures will contribute to the further destabilization of autocrats. 74 We accounted for the number of popular protests by adding the number of anti-government demonstrations, strikes and riots recorded in the Banks and Wilson data set. 75 Finally, we also included a control for oil production due to the fact that rentier state theory predicts oil-producing regimes to be more stable. Senders may therefore be less prone to target these regimes with likely ineffective sanctions. Furthermore, senders are also often dependent on oil exports from authoritarian regimes. We measure oil production in millions of metric tons. 76 The democratizing effect of democratic sanctions To investigate the hypotheses laid out in the theory section, this section will introduce a number of statistical models. The main analyses are displayed in Table 4. Again, the dependent variable is the change in the combined Freedom House and Polity index. Hence, the dependent variable represents D i,t -D i,t21, where D is the level of democracy for country i and time t. All models use panel-corrected standard errors and a lagged dependent variable to account for serial correlation. 77 All independent variables, including sanctions, are lagged one year. Lagging the independent variables has important consequences for the results. Sanctions are often implemented as a result of drastic democratic decline, such as unfair elections or coup d états. Therefore, the democratic development in the first year of a sanction episode is often more a cause than a consequence of the sanction. All models feature the same controls, but they vary in how they account for sanctions. All models were checked for multicollinearity; the highest variance inflation factor observed for any coefficient in any model was 4.96, indicating that multicollinearity is not a pressing concern. The results are remarkable given that Peksen and Drury 78 have shown that all sanctions have a robust negative impact on democracy. Model 1 displays a positive, but insignificant, effect of all sanctions against authoritarian regimes on democratic development in targeted countries. When we look specifically at democratic sanctions in Model 2, we observe much clearer positive outcomes. The variable for democratic sanctions becomes highly significant and the expected yearly change in democracy scores is higher in authoritarian countries with democratic sanctions than in those without. The models utilize a combined Freedom House and Polity IV score to account for democratic change. But the results also remain robust if the indices are used separately to account for democratic change or if we extend the time lag of the democratic sanctions variable with one additional year. These results lend strong support to H1a and H1b.

14 Table 4. Regression models 1 4. Dependent variable: change in level of democracy (1) (2) (3) (4) LDV (0.067) (0.067) (0.067) (0.067) Sanctions.041 (.046) DM sanctions (0.069) (0.085) HR sanctions (0.107) PC sanctions (0.099) TR sanctions (0.096) WMD sanctions (0.123) OT sanctions (0.115) Nondemocratic (0.067) Log population (0.012) (0.013) (0.011) (0.013) GDP (0.000) (0.000) (0.000) (0.000) Trade (0.000) (0.000) (0.000) (0.000) FDI (0.002) (0.002) (0.002) (0.002) Civil war (0.059) (0.059) (0.059) (0.057) Oil production (0.000) (0.000) (0.000) (0.000) Popular protests (0.011) (0.011) (0.011) (0.011) Constant (0.142) (0.143) (0.134) (0.140) Observations Number of countries R squared Notes: Entries are unstandardized coefficients with panel-corrected standard errors within parentheses. All independent variables use a one-year lag. DM, democratization; HR, human rights; PC, peace; TR, terror; WMD, weapons of mass destruction; OT, other. Significant at the 0.10 level. Significant at the 0.05 level. Significant at the 0.01 level. 12 C. von Soest and M. Wahman

15 Democratization 13 Further emphasizing the importance of separating democratic sanctions from other sanction categories, Model 3 introduces a variable that accounts for nondemocratic sanctions (that is, all sanction categories except democratic sanctions). Not only does the coefficient become insignificant when the democratic sanctions are excluded, it also changes direction and becomes negative. Since many of the cases are targeted with sanctions that have multiple goals, Model 4 introduces controls for all the sanction categories. The results regarding democratic sanctions remain significant, and we can see that the democratic sanction category is the only one with a significantly positive effect. A potential problem with the results above could be that democratic sanctions are generally more comprehensive than sanctions imposed to obtain other goals. 79 In Table 5, we therefore make the distinction between comprehensive and targeted sanctions and observe whether the variable remains significant when we control for these different measures. Table 5. Regression models 5 7. Dependent variable: change in level of democracy (5) (6) (7) LDV (0.067) (0.067) (0.067) DM sanctions (0.098) HR sanctions (0.107) PC sanctions (0.111) TR sanctions (0.091) WMD sanctions (0.140) OT sanctions (0.132) Targeted sanctions (0.063) Comprehensive sanctions (0.055) (0.100) Log population (0.013) (0.013) (0.013) GDP (0.000) (0.000) (0.000) Trade (0.000) (0.000) (0.000) FDI (0.002) (0.002) (0.002) Civil war (0.059) (0.059) (0.058) Oil production (0.000) (0.000) (0.000) Popular protests (0.012) (0.011) (0.011) Constant (0.134) (0.144) (0.147) Observations Number of countries R squared Notes: Entries are unstandardized coefficients with panel-corrected standard errors within parentheses. All independent variables use a one-year lag. DM, democratization; HR, human rights; PC, peace; TR, terror; WMD, weapons of mass destruction; OT, other. Significant at the 0.10 level. Significant at the 0.05 level. Significant at the 0.01 level.

16 14 C. von Soest and M. Wahman Models 5 and 6 show the effects of targeted and comprehensive sanctions. The results suggest that comprehensive sanctions have a more positive effect than targeted sanctions, but the effect is still not significant. Model 7 replicates Model 4, but adds a control for comprehensive sanctions. We find no evidence in this Model 7 to suggest that the effect of democratic sanctions is due to their more comprehensive nature. Instead, the democratic sanctions variable remains significant, with a slightly larger coefficient than in Model 4. All in all, these analyses indicate that none of the imposed sanctions have been counterproductive to the democratic development of authoritarian regimes. Sanctions in general do not promote democratization, but democratic sanctions are significantly associated with higher levels of democracy in the targeted country. 80 Supplementary tests on sanctions and regime stability The previous section showed that democratic sanctions are associated with increased levels of democracy. We also wanted to test whether democratic sanctions are more often associated with regime change and ruler exit (H2 and H3). This would provide an insight into how implemented democratic sanctions actually work in targeted authoritarian regimes and how they contribute to improved democracy levels. Therefore, as a supplementary test, we ran a series of logistic regressions accounting for regime instability. The models below are logistic regressions with conditional fixed effects. The fixed-effects specification is a strategy that can be used to account for countryspecific omitted variables. 81 Following Marinov, 82 we used a fixed-effects model rather than an event-history model. 83 According to Beck, Katz, and Tucker, 84 when using cross-sectional time-series data with a dichotomous dependent variable of an event measured at discrete intervals, fixed-effects logit is technically equivalent to an event-history model. The models use largely the same control variables as the main regression models. However, because simply utilizing a lagged dependent variable to correct for autocorrelation is not sufficient in models with binary dependent variables, we also include a restricted spline based on the last event (that is, the last change in regime type or leader exit). 85 The models above confirm our expectations that democratic sanctions induce leadership exit and regime change (H2 and H3). Model 8 shows no general relationship between sanctions and a higher probability of ruler exit. However, looking at only democratic sanctions in Model 9, we see a significant positive relationship. Our findings reaffirm Marinov s 86 conclusion that sanctions generally increase the probability of leadership exit. Looking at the models for regime change, we once again see no significant effect of sanctions generally (Model 10), but we find a significant effect for imposed democratic sanctions (Model 11). A brief glimpse at some sanction cases reveals that different mechanisms may account for increased democracy levels in targeted authoritarian regimes, with the two major ones being (1) elite splits (and, in turn, regime changes and leadership exit) and (2) democratic concessions without regime and/or leader change. In

17 Democratization 15 Guatemala (1993), for instance, the military ousted President Serrano who had unconstitutionally dissolved parliament and the judiciary after the US and its allies imposed sanctions. An interim president took over, and the country s democratic institutions were restored. In Nicaragua (1996) and Thailand (1993), sanctions similarly contributed to regime change. However, democratic sanctions rarely create liberal democracies instantly; rather, they lead to multiparty autocracies. In addition, democratic sanctions have different effects on autocratic rulers and regimes. In Peru, sanctions contributed to democratization without ruler change. When President Fujimori suspended the legislature and introduced rule by decree in 1992, the US withheld military assistance and economic aid and blocked Peru s efforts to obtain loans from international financial institutions. Table 6. Conditional fixed-effects logistic regressions of regime change and leader exit (models 8 11). Ruler exit (8) Ruler exit (9) Regime change (10) Regime change (11) Sanction (0.598) (0.713) DM sanction (0.812) (0.755) Nondemocratic (0.625) (0.788) sanction Comprehensive (0.741) (0.621) (0.753) (0.729) Log population (1.173) (1.199) (0.936) (0.956) GDP/capita (0.000) (0.000) (0.000) (0.000) Trade (0.006) (0.006) (0.006) (0.006) FDI (0.016) (0.016) (0.017) (0.016) Civil war (0.327) (0.328) (0.334) (0.335) Oil production (0.032) (0.032) (0.020) (0.020) Popular protest (0.040) (0.039) (0.039) (0.039) Spline (0.123) (0.123) (0.116) (0.118) Spline (0.906) (0.828) (0.830) (0.900) Spline (3.270) (3.279) (1.609) (1.619) Observations Number of countries Log likelihood Notes: Conditional fixed-effects logistic regression. Entries are logistic coefficients with standard errors in parentheses. All independent variables are lagged one year. Significant at the 0.10 level. Significant at the 0.05 level. Significant at the 0.01 level.

18 16 C. von Soest and M. Wahman In response, Fujimori agreed to hold elections and to reinstate formally democratic institutions. Although his presidential dominance persisted until 2000, Peru s political system was liberalized to some extent for the remainder of his time in office. 87 In this particular case we could, hence, observe both institutional change and democratization, without change in leadership. Interestingly, Table 6 also shows that FDI decreases the chances of both ruler exit (Models 8 and 9) and regime change (Models 10 and 11). However, this might be due to international investors aversion to investing in politically unstable countries in the first place. Regarding sanctions, it is interesting to see that comprehensive sanctions decrease the likelihood of regime change in Models 9 and 11. We also see that (1) larger countries are more stable in terms of both leadership exit and institutions and (2), as expected, that richer countries experience fewer regime changes (Models 10 and 11). Reaffirming the idea of the resource curse, we also see more regime stability in regimes with larger oil production (Models 8 11). 88 There are several instances where regime change has been preceded by sanctions, although we would not argue that the relationship between sanctions and institutional change was causal in all these instances. The Online Appendix provides a full list of these instances. Changes in regime type rarely lead to full democracy, and countries sometimes transition from what could generally be perceived as a more democratic regime type to a less democratic one. However, most of the instances where change in regime type is preceded by democratic sanctions show signs of at least limited liberalization (such as the implementation of multiparty elections). Conclusions Since the end of the Cold War, democratization has motivated an overwhelming majority of all sanctions directed towards authoritarian regimes. Previous literature gives little support for continuing the practice of sanctioning authoritarian regimes, especially if the aim is to induce democratization. It is suggested that sanctions are not only generally ineffective, but also counterproductive, in increasing the level of democracy in the targeted countries. This study has provided evidence to the contrary. We argue that empirical research on the effectiveness of sanctions must take the explicit goal of sanctions into account. In order to properly distinguish sanctions according to their respective goals, this study introduced a new data set of sanctions imposed by the EU, the UN, and the US in the period By concentrating on only sanctions aimed at increasing the level of democracy in the targeted authoritarian state, we found that these sanctions are indeed associated with increasing levels of democracy. Moreover, authoritarian regimes targeted by democratic sanctions are more likely to experience institutional and leadership change. These findings are in tune with recent democratization research arguing the importance of economic decline for short-term democratization. 89

19 Democratization 17 Further research is needed with regard to the strategic decisions made by sanction senders. First, the question of how senders choose authoritarian targets warrants further analysis. An investigation into the factors influencing target selection would provide additional information on the criteria applied and, at the same time, contribute to the selection bias discussion in general. 90 Second, it is important to know more about what factors bilateral and multilateral sanction senders take into account when they decide which measures to impose on authoritarian regimes. Do they explicitly try to reduce the potential for rentseeking by for instance limiting the number of trade sanctions? Such knowledge would further contribute to the debate about sanction design, targeted sanctions and sanction effectiveness measured against their goal. Acknowledgments The authors contributed equally to this article. We would like to thank Jana Bischler and Sinja Hantscher for their excellent research assistance. Earlier versions of this article were prepared for the 2012 Annual Meeting of the American Political Science Association (New Orleans) and presented at the international GIGA Workshop Autocratic Regimes and the Effects of International Sanctions (Hamburg, 25 June 2012); Rice University, Houston; and Leuphana University, Lüneburg. We are grateful for all comments received on these occasions. We are especially grateful to Matthias Basedau, Julia Grauvogel, Viola Lucas, Manus Midlarsky, Clifton Morgan, Clara Portela, Jan Pierskalla and Milan Svolik, and two anonymous reviewers for their invaluable comments. Funding The research is a product of the research project Ineffective Sanctions? External Sanctions and the Persistence of Autocratic Regimes, funded by the Fritz Thyssen Foundation (grant number: ). Online Appendix available from the Supplemental tab at doi.org/ / and at soest. Notes 1. Drury, Revisiting Economic Sanctions Reconsidered ; Haass, Economic Sanctions and American Diplomacy; Pape, Why Economic Sanctions Do Not Work. 2. Peksen and Drury, Coercive or Corrosive. 3. Wood, Sanctions and State Repression. 4. For example, Drury, Revisiting Economic Sanctions Reconsidered ; Haass, Economic Sanctions and American Diplomacy; Pape, Why Economic Sanctions Do Not Work. 5. Peksen and Drury, Coercive or Corrosive ; Wood, Sanctions and State Repression. 6. For example, Geddes, Democratization after Twenty Years ; Bueno de Mesquita and Smith, Leader Survival, Revolutions ; Teorell, Determinants of Democratization. 7. Levitsky and Way, Rise of Competitive Authoritarianism, Burnell, Democracy Assistance; Carothers, Aiding Democracy Abroad. 9. Drury and Peksen, Women and Economic Statecraft ; Peksen and Drury, Coercive or Corrosive ; Peksen, Effect of Economic Sanctions ; Wood, Sanctions and State Repression.

20 18 C. von Soest and M. Wahman 10. Hufbauer et al., Economic Sanctions Reconsidered. 11. For a similar assessment see Cortright and Lopez, The Sanctions Decade. 12. Hufbauer et al., Economic Sanctions Reconsidered, Pape, Why Economic Sanctions Do Not Work ; Drury, Revisiting Economic Sanctions Reconsidered. 14. Bonetti, Sanctions and Statistics; Haass, Economic Sanctions and American Diplomacy; Jing, Kaempfer, and Lowenberg, Effectiveness of International Sanctions ; Kaempfer and Lowenberg, International Economic Sanctions. 15. Peksen and Drury, Coercive or Corrosive ; Wood, Sanctions and State Repression. 16. Collier and Adcock, Measurement Validity ; Elkins, Gradations of Democracy? ; Verkuilen and Munck, Conceptualizing and Measuring Democracy. 17. See for example, Teorell, Determinants of Democratization. 18. Peksen and Drury, Coercive or Corrosive. 19. For example, Wintrobe, Political Economy of Dictatorship. 20. Wood, Sanctions and State Repression ; Peksen, Effect of Economic Sanctions ; Escribà-Folch, Authoritarian Responses to Foreign Pressure ; Drury and Peksen, Women and Economic Statecraft. 21. Gasiorowski, Crisis and Political Regime Change ; Geddes, Democratization after Twenty Years ; Ulfelder and Lustik, Modelling Transitions. 22. Rotberg, Repressive, Aggressive, and Rogue Nation-States ; Wintrobe, Political Economy of Dictatorship. 23. Bueno de Mesquita and Smith, Leader Survival, Revolutions ; Gandhi and Przeworski, Authoritarian Institutions. 24. Lichbach, Deterrence or Escalation. 25. Bueno de Mesquita and Smith, Leader Survival, Revolutions. 26. Ibid. 27. Teorell, Determinants of Democratization. 28. Kirshner, Microfoundations of Economic Sanctions ; Allen, Determinants of Economic Sanctions ; Hufbauer et al., Economic Sanctions Reconsidered. 29. Lektzian and Souva, Institutional Theory of Sanctions ; and see early Galtung, Effects of International Economic Sanctions. 30. Drezner, Sanctions Sometimes Smart. 31. Sometimes regime destabilization is not at all in the interest of the sender country, especially if such political instability could have an adverse effect on other foreign policy goals such as national security. 32. For personalist regimes: Escribà-Folch and Wright, Dealing with Tyranny ; for sanctions in general: Marinov, Do Economic Sanctions Destabilize. 33. Drezner, The Sanction Paradox; Drezner, Economic Coercion ; Lacy and Niou, Theory of Economic Sanctions ; Miers and Morgan, Multilateral Sanctions and Foreign Policy Success ; Morgan, Bapat, and Krustev, Threat and Imposition ; Whang, McLean, and Kuberski, Sanction Threats. 34. Drury and Li, US Economic Sanction Threats. 35. See also Lacy and Niou, Theory of Economic Sanctions. 36. Morgan, Bapat, and Krustev, Threat and Imposition ; Morgan, Krustev, and Bapat, (TIES) Data Users Manual. 37. Data includes threats of sanctions against all countries classified as authoritarian by Hadenius, Teorell, and Wahman, Authoritarian Regimes Dataset, issued by the EU, UN or US. Due to small-n for the post-cold War period (TIES collects data from ), data for the Cold War period were also included. 38. In our own data we separate between democratic and human rights sanctions. Since this separation is not done in the TIES data, we present the categories jointly.