Does democracy promote capital account liberalization?

1 As Houle, Kayser, and Xiang (Citation2016, p. 695) point out, ‘the actors in international diffusion processes are most often elites who observe and then adopt or eschew policies from abroad’.

2 Linos (Citation2013) and Pacheco (Citation2012) also argue that societal preferences help account for policy diffusion, though their studies focus on social policy issues.

3 The concentrated interest groups that typically dominate economic policy-making in autocracies are unlikely to have unified preferences vis-à-vis capital controls. Many politically connected firms can benefit from capital controls because, in a closed financial system, the government showers them with subsidized credit and provides other rents (Giannetti & Ongena, Citation2009; Johnson & Mitton, Citation2003). However, other organized interests, such as the banking sector, are more likely to support the liberalization of some types of cross-border financial flows (Gallagher, Citation2015; Mukherjee, Yada, & Béjar, Citation2014; Pepinsky, Citation2013).

4 Our argument does not, however, imply that the overall welfare effects of capital account liberalization are large and positive. It is likely that citizens pay more attention to the microeconomic benefits of liberalization, some of which are more direct, more tangible and more concentrated (e.g. access to cheap credit), than to the macroeconomic costs of liberalization (e.g. reduced monetary policy autonomy), which are less immediate, less well-publicized and tend to be spread more widely across the public.

5 We focus on microeconomic rather than macroeconomic consequences of capital controls for two main reasons. First, voters have difficulty connecting their personal interests to the macroeconomic effects of capital controls (Brooks & Kurtz, Citation2007, p. 705; Helleiner, Citation1994, p. 205). Second, while we have strong reasons to expect that liberalization abroad will intensify the microeconomic costs of capital controls, the macroeconomic benefits are less likely to change substantially. Rising international capital mobility increases the need for regulations over cross-border capital flows while at the same time reducing these regulations’ ability to stabilize the macroeconomy.

6 Small firms do not have large internal capital markets to tap into, and they rarely have the political connections necessary to reliably obtain cheap credit in closed financial systems. Large firms, by contrast, often benefit from capital controls because such controls make it easier for governments to channel subsidized credit toward ‘crony capitalists’ (Giannetti & Ongena, Citation2009; Johnson & Mitton, Citation2003).

7 Summary statistics for the variables included in this dataset can be found in Table A1 of the Supplementary Appendix. The appendix also lists all the countries included in our main model specification.

8 Importantly, however, we do use data on developed countries’ capital account policies when constructing the spatial lag variables. Note also that including these countries in the main analyses does not alter our main findings (see Table A2 in the appendix).

9 A Dickey-Fuller test rejects the null of a unit root (p < .001). We also obtain similar results when the dependent variable is the first-difference of ckaopen (see Table A18).

10 ckaopen is a modified version of Chinn and Ito’s (Citation2006) kaopen measure, which uses a five-year moving average of capital account restrictions. By including previous years’ openness in the measure, kaopen can generate biased estimates (see Karcher & Steinberg, Citation2013).

11 Our measurement strategy is similar to that of Simmons and Elkins (Citation2004), Brooks and Kurtz (Citation2012), and Guisinger and Brune (Citation2017), who test the effect of spatial lag variables on capital account liberalization. None of these studies, however, examines the interaction between the diffusion variables and democracy, as we do here.

12 Bilateral distances are drawn from the CEPII gravity database, and are calculated as the distance between each country’s largest cities, weighted by the population shares of those cities relative to their country’s total population. For a more in-depth description see Mayer and Zignago (Citation2011). The calculation of the spatial lag follows the procedure outlined by Attfield, Cannon, Demery, and Duck (Citation2000).

13 By contrast, endogeneity problems potentially plague some of most widely used alternative weighting matrices. For example, capital account policies influence whether a country is a major investment destination and whether it receives a strong credit rating, making it more problematic to use these variables to determine country weights.

14 One limitation of the geographic distance-weighted indicator is that it does not enable us to assess the relative importance of economic partners and cultural peers, but doing so is not our main objective. Instead, we aim to evaluate whether the international capital policy environment, defined broadly, moderates the relationship between domestic institutions and capital account policy.

15 All negative values of inflation were set to 0 plus 1 prior to the log transformation.

16 Lagging the spatial variables helps reduce simultaneity biases in OLS estimates of spatial effects (Franzese & Hays, Citation2008, p. 758). Simultaneity biases, however, are not likely to be much of an issue for our analyses because the estimated spatial effects are not especially strong, the spatial lag is only weakly correlated with the other explanatory variables, and the non-spatial variables have strong explanatory power. Franzese and Hays (Citation2007) show that simultaneity biases from spatial OLS estimates tend to be small under these conditions. Moreover, our interest is not in evaluating the relative importance of spatial versus domestic variables, which is where these biases are most likely to arise.

17 The choice of standard error does not affect the interpretation of the results. PCSEs and country-clustered standard errors were similar in size to the OLS standard errors (see Tables A6 and A7).

18 We obtain substantively similar results when using Hainmueller, Mummolo, and Xu’s (Citation2017) binning estimator and kernel estimator (see Figures A2 and A3) instead of the linear estimator.

19 For more details on the construction of these variables, see the Supplementary Appendix.

20 In Table A19, we test for the possibility that the direction of influence runs from capital policy liberalization to the level of democracy rather than the other way around using the difference-in-differences approach from Giuliano et al. (Citation2013). The results suggest that capital account liberalization does not produce improvements in countries’ democracy scores.

21 The median firm in the sample employs 23 workers. Firms of this small size are unlikely to have access to policy-makers in a dictatorship, and are more likely to be able to influence policy in democracies. Consistent with this intuition, Weymouth (Citation2012) shows that firms are more likely to lobby and believe that they influence policy in more democratic countries.

22 We use the logarithmic form of this variable because the interest rate series is highly skewed (the average interest is 15%, with a standard deviation of 207%).

23 The Supplementary Appendix provides more detailed descriptions of these variables.

24 Although some of our outcome variables are ordinal, we focus on linear models because the substantive effects are easier to interpret. As shown in Tables A24 and A25, we obtain similar results using hierarchical logit and OLS with clustered standard errors.

25 The range of values on the spatial lag variable is much narrower, and higher on average, than in the country-level dataset because the firm-level dataset only includes data for years in which the global financial system was largely open.

26 As a placebo test, we also examined the other 16 ‘business obstacle’ questions that the World Bank asked firms. The results, shown in Tables A21–A23, indicate that capital account policies are not significantly related to business problem questions that have no theoretical link to capital controls.

27 Argentina briefly deviated from its policy of financial openness when it imposed intense capital controls during a severe financial crisis in 2001–2002. However, Argentina’s overall level of financial openness was high in the 1990–2010 period; for example, the mean ckaopen score for this period is 0.2. This compares to a mean score of −1.3 for Argentina’s democratic governments between 1966 and 1989 – a period in which the global context was far more closed – and a mean of −0.4 for the country’s military regimes. These broad patterns over time within Argentina are consistent with our findings in the cross-national statistical analyses, which makes this a useful case for in-depth analysis (Lieberman, Citation2005).

28 See, for example, Cronista Comercial, Sep 12, 2012, http://www.cronista.com/economiapolitica/Miles-de-personas-protestaron-en-Plaza-de-Mayo-y-varios-puntos-del-pais-20120913-0133.html (accessed Jan. 31, 2017); La Nación, Apr. 3, 2013 http://www.lanacion.com.ar/1572356-cristina-viajara-a-venezuela-horas-antes-del-cacerolazo-del-18-a (accessed Jan. 31, 2017); La Nación, Nov 9, 2014 http://www.lanacion.com.ar/1742470-convocan-por-las-redes-a-una-protesta (accessed Jan. 31, 2017).

29 Specifically, the question asked whether respondents agreed with the policy that ‘citizens need to obtain permission from the national government and pay taxes to purchase dollars’.

30 The difference in proportion of respondents supporting the capital control compared to the trade restriction variables is statistically significant (p < .01).

31 The dollar is the most popular foreign currency in Argentina for a number of reasons: the dollar is the world’s number one reserve currency, Argentina is physically closer to the United States than it is to any other major reserve currency (e.g. Euro or British Pound), and the United States is one of Argentina’s leading trade and investment partners.

32 Consistent with this assumption, there is a very strong correlation between bank account ownership and savings in Demirguc-Kunt et al.’s (Citation2015) database, both in Argentina and globally.

33 Using this dataset, Steinberg and Nelson (Citation2016) shows that ‘financialized’ individuals (those with bank accounts and credit cards) were more opposed to capital controls than other Argentines, after controlling for other factors, such as income, ideology and partisanship.

34 Clarín, Nov 3, 2015 http://www.clarin.com/ultimo-momento/macri-scioli-dolar-devaluacion-precios_0_SJ4eYTgtPQg.html (accessed Jan. 24, 2017); Clarín Nov 19, 2015 http://www.clarin.com/politica/elecciones_2015-entrevista-daniel_scioli-mauricio_macri-balotaje_0_B16SmgFvQl.html (accessed Jan. 24, 2017).

35 La Nación Sep. 16, 2014. http://www.lanacion.com.ar/1727589-un-incomodo-espejo-para-la-argentina (accessed Jan. 24, 2017).

36 La Nación Oct. 24, 2012. http://www.lanacion.com.ar/1519922-cepo-cambiario-el-fracaso-disfrazado-de-exito (accessed Jan. 24, 2017).

37 La Nación, Aug. 3, 2014, www.lanacion.com.ar/1715318-verdades-ocultas-detras-del-festin-de-los-patriotas (accessed Jan. 31, 2017).

38 Clarín Nov. 15, 2013. http://www.clarin.com/mundo/dolar-inversiones-tendencias-chile-argentino_0_ByGNfjwme.html (accessed Jan. 24, 2017).

39 La Nación, Mar. 24, 2015, http://www.lanacion.com.ar/1778603-el-gran-dilema-politica-de-shock-o-gradualismo (accessed Jan. 31, 2017).

40 Bloomberg, Nov. 24, 2015, https://www.bloomberg.com/news/articles/2015-11-25/argentina-s-president-elect-playing-dangerous-game-with-the-peso (accessed Jan. 31, 2017).

41 Ambito Financiero, Nov. 23, 2015 http://www.ambito.com/817132-ahora-macri-no-le-pone-fecha-al-fin-del-cepo (accessed Jan. 31, 2017).

42 The Economist, Nov. 28 2015, http://www.economist.com/news/americas/21679249-mauricio-macris-victory-could-transform-his-country-and-region-end-populism (accessed Jan. 31, 2017).