![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
This work examines how U.S. economic sanctions affect social welfare spending in authoritarian countries. U.S. economic sanctions play a role of leading autocratic targets to change social policy through two theoretical channels. First, U.S. economic sanctions may reduce autocrats’ resources to buy off supports from ruling elite groups and so force autocrats to reallocate government expenditure in favor of their supporting groups. Consequently, autocrats facing longer U.S. sanctions are likely to cut spending on public goods and services, especially on education and health care spending. Second, the impacts of U.S. sanction duration on social spending vary according to political variables such as autocrats’ pseudo-democratic institutions. The empirical findings show that, even when U.S. sanctions last a long time, autocrats under nominal democratic institutions cut spending on education and health to a lesser degree than do autocrats with no such institutions. In contrast, autocrats relying on pseudo-democratic institutions reduce social security spending a little more than did non-institutionalized autocrats.
Acknowledgements
The author thanks Steven Redd, Uk Heo, Shale Horowitz, and David Armstrong for helpful comments. Sukwon Lee and anonymous reviewers also provided very helpful comments and suggestions.
Notes
No potential conflict of interest was reported by the author.
Supplemental data for this article can be accessed at https://doi.org/10.1080/10242694.2017.1392832.
1 For a democratic leader, continuous sanctions can threaten her political survival. Democratic audiences or constituencies punish their leaders through (re-) election in the sense that voters in democracies judge governments’ commitments to public goods and services that may be aggravated by foreign economic sanctions (Bueno de Mesquita et al. Citation2003; Fearon Citation1994). Thus, democratic regimes become the sanction targets during a shorter time compared to autocracies.
2 In robustness check, I examine the interactive effect of US sanction duration and pseudo-democratic institutions on autocracies’ social welfare expenditure further based different model specifications. Findings (Appendix Tables) show to be substantial and robust to a variety of model specifications and a different sanction dataset except a few additional tests using polynomial regressions. See the Appendix for details.
3 According to sanction studies, there is a selection effect in the sense that countries tend to concede in the sanction threat stage. For instance, existing studies on sanction initiation show that ‘democratic targets are more likely to be held responsible for economic failings and are more likely to view the threatened costs of sanctions as sufficiently severe. So, while democracies use sanctions more frequently, they tend not to use them against other democracies as often as they use them against autocracies’ (Lektzian and Souva Citation2007, 856). Peksen (Citation2017) recently found that there is a variation in sanctions success across different types of authoritarianism. In particular, he showed that military or single-party dictatorships are more likely to resist foreign sanctions than personalist autocracy.
4 Escribà-Folch (Citation2012) analyzed the impact of economic sanctions on government expenditure using the data with 70 countries from 1970–2000. The number of observations in his dataset is around 800–1000. Almost half of his original data are missing. Alternatively, increasing number of scholars in international relations and comparative politics employ the multiple imputation technique to avoid missing data problems (e.g. Ross Citation2006).
5 There are two reasons to use each social spending as percentage of total government spending. On the one hand, social expenditure as ‘a share of total government spending’ reflects government policy priorities better than ‘as a share of GDP’ even if many studies employ the latter measure (Rudra and Haggard Citation2005, 1022–1023). GFS yearbooks, on the other hand, do not consistently provide information of government expenditures on social spending as a share of GDP across time-periods. GFS yearbooks are more useful than the World Development Indicators in the sense of data availability for specific sectors of social spending - social protection, education and health. But, I still acknowledge that data from some developing countries may be unreliable while it is hard to find a solution to solve this.
6 In addition, I try to include social welfare spending in aggregating all three expenditures on social security, education and health. However, I do not analyze whether the U.S. sanctions influence this aggregated social spending due to missing data across three expenditures in an irregular way.
7 I report how many countries faced U.S. sanctions and how many countries targeted by U.S. sanctions are not included in the sample due to missing values in the dependent variables. See the detailed information in the Appendix, Table IV.
8 I exclude authoritarian countries that have no data in each social spending category. A large portion of those excluded countries have a very small size in population.
9 I focus on the impact of the U.S. sanctions because the U.S. has created and maintained U.S.-led international economic system as ‘the economic hegemon’ (Shin, Choi, and Luo Citation2016).
10 I also run the regressions using the TIES dataset compiled by Morgan, Clifton, and Kobayashi (Citation2013) for the robustness of the results. The TIES data distinguish the US sanction cases in which the US is the primary sender with cases where the US is a part of sanction senders. In the analysis, I deal with the sanction cases in which the US participate. The regression outcomes is reported in the appendix.
11 In addition to US sanction duration variable, I run regression models using several types of sanctions which might hurt target economies. First, the effect of sanctions on social spending might be conditioned by the degree to which economic sanctions inflict major damage on target macroeconomies (Peksen and Son Citation2015; Hatipoglu and Peksen Citation2016), increase the level of poverty (Neuenkirch and Neumeier Citation2016) and income inequality (Afesorgbor and Mahadevan Citation2016). I use high cost sanctions from the HSEO dataset and severe sanctions from the TIES dataset based on Lektzian and Biglaiser (Citation2013)’ and Lektzian and Patterson (Citation2015)’ operationalizations. The sanction cost variable was not statistically significant in the model. Second, another sanction variable is US sanctions through international organizations. Scholars of economic sanctions argue that sanctions through international organizations should resolve enforcement problems such as sanction-busting behaviors of third parties to a considerable extent. Drezner (Citation2000), for instance, contends that ‘sanctions with reasonably high levels of international cooperation should impose greater costs on the target country because of the inability to find alternative markets and suppliers.’ International institutions possess enforcement powers to prevent other countries from defecting and reduce the probability of backsliding. Bapat and Morgan (Citation2009) also find that ‘multilateral sanctions do appear to work more frequently than do unilateral sanctions’ because they have ‘the potential to create more coercive power than unilateral sanctions.’ This does not change the main results.
12 In response to a reviewer’s comments, I investigate the impact of the sender’s sanctions on the target’s economy using the trade dependence on the sender based on its share of GDP. I employ the indicators of both one percent and five percent of GDP. The results are reported in the Appendix, Tables I–5, II-4 and III-4.
13 Following Escribà-Folch (Citation2012)’s findings, I also control for the effect of different autocratic regimes on social spending, using Geddes, Wright, and Frantz (Citation2014)’s new dataset on autocratic regimes. The results are reported in the Appendix (Tables I-3, II-2 and III-3). Despite the control for the three autocratic regime types (i.e. single-party, personal and military regimes), my main findings still uphold while sanctions through IOs or multilateral sanctions do not significantly influence social spending in autocratic regimes. I also use Cheibub, Gandhi, and Vreeland (Citation2010) dataset on dictatorship (i.e. civilian, military and royal dictatorship). The findings are reported in the Appendix (Tables I-2, I-5, II-1, II-4, III-1 and III-4).
14 This variable is logged to make positively skewed distribution of data more normalized.
15 Haber and Menaldo (Citation2011, 14) describe in their supplementary appendix that total oil income per capita is calculated as ‘crude oil production times the price of crude oil in 2007 dollars, divided by population.’
17 As an anonymous reviewer observed, sanctions may have the largest effect a year or two after imposition. Thus, to capture a nonlinear effect of US sanction duration, I include a quadratic term and a squared interaction term of US sanction duration. The statistical significance of this squared term examines whether there is a quadratic effect of US sanction duration on the change in social welfare spending. I report such a regression in the Appendix V, finding that coefficients of the interactive term are similar to those in the main body of the paper regarding the change in social security spending (main results in Tables and ). Interaction terms predicting the change in health spending have the hypothesized signs, but their effects still become insignificant. Strikingly, the results for the change in education expenditure show that the interaction term in the long-run equation is negatively signed, indicating that lagged US sanction duration may reduce education spending in autocracies with institutions in the long run, but the squared term’s positive sign indicates a quadratic effect which makes the curve convex.
18 I generate missing map to detect how many data are missing. Among variables, social spending variables are missing for more than 40 percent. I make plots to show how similar imputed data are to observed data (see Appendix). Overall, imputed data closely fit the distribution of observed values.
19 For instance, let us estimate the effect of democracy on economic development using the case of Chile in 1985 (Kohli et al. Citation1995). We actually observed 1985 Chile as an authoritarian regime and know the GDP per capita income of 1985. To estimate the impact of democracy on development, we also need to observe 1985 Chile as a democracy at the same time. Unfortunately, we can seldom find democratic Chile in 1985 in observed data. Thus, we need to rely on a quasi-experimental method ‘to look for a case that is exactly like Chile in all aspects other than its regime’ (Kohli et al. Citation1995, 16–17).
20 There is a variety of matching techniques. Among them, I employ ‘genetic matching’ because it is nonparametric and so does not predefine the distribution. In addition, genetic matching ‘automatically finds the set of matches that minimizes the discrepancy between the distribution of potential confounders (extraneous variables) in the treated and control groups’ (Sekhon Citation2009, 499).
21 Based on a reviewer’s suggestion, I run regressions jackknifing each country in the models which has been a target of U.S. sanctions for a considerable amount of years. Unfortunately, I should exclude some cases such as Cuba, Libya and North Korea due to their missing values in the dependent variables (social welfare expenditure). Regarding social security spending, the main findings except the model excluding Iran still hold (Appendix, Table I-1). About education spending, the main findings also hold except models excluding Ethiopia, South Africa, Sudan, Iran or Myanmar (Appendix, Table II-5). Regarding health expenditure, unlike the main results, models excluding Uganda or Iran obtain statistical significance. The results are reported in the Appendix (Table III-5).
22 Most public policies retain the characteristics of both private and public goods (Bueno de Mesquita and Smith Citation2010). But, some policies may favor a specific segment of the population as considered as a part of private goods. Here I regard social security spending in autocracies as private rather than public goods for a few reasons. Government expenditure on social security tends to be restrictive in the developing world. A number of scholars argue that social security spending or transfer is mainly associated with pensions. In particular, Kaufman and Segura-Ubiergo (Citation2001) posit that each subtype of social welfare spending has its own ‘political logics.’ According to Kaufman and Segura-Ubiergo (Citation2001, 576), the general social security includes ‘antipoverty programs and targeted assistance to the poor’ but pensions mainly target ‘the middle class’ and ‘workers in the formal sector’ in Latin American countries. Park and Estrada (Citation2013) argue that pension systems in many Asian countries tend to be ‘skewed toward urban areas and the formal sector’ and ‘initially covered only government workers.’