The effects of economic sanctions on targeted countries’ stock markets

ABSTRACT

Although much previous research has investigated the impact of sanctions on trade and global capital, few academic studies have explored the effect of sanctions on stock markets in targeted countries. The lack of research is surprising as a frequent goal of sanctions is to inflict pain on financial markets in targeted countries to promote policy change. Using monthly market data for 66 countries from 1990 to 2005, we find that the introduction of import sanctions by countries with developed economies, such as those with membership in the G20, has a significantly negative impact on stock market valuation in targeted countries. However, this effect only occurs when targeted states are not already subject to multiple sanctions. Our study suggests that sanctions can have a negative effect on stock market value in targeted countries, but that their effectiveness is relatively limited in practice due to the overuse of sanctions. This finding is supported by the marginal decrease in the negative effect on the target’s stock market as the number of sanctions increases.

KEYWORDS:

• Sanctions
• stock market
• CPE

Supplementary Material

Supplemental data for this article can be accessed on the publisher’s website.

Notes

¹ See Shin , Choi, and Luo (2016), who argue that targeted countries increase trade with regional trade bloc members to compensate for lost commerce with sanctioning countries who are nontrade bloc members.

² See Drezner (1999), Early (2009, 2015), Kaempfer and Lowenberg (1999), Lektzian and Biglaiser (2013), Miers and Morgan (2002), and Shin et al. (2016).

³ See Caruso (2003); Early (2009, 2015); Hufbauer et al. (1997).

⁴ See Biglaiser and Lektzian (2011); Colin and Kleinberg (2015); Lektzian and Biglaiser (2013); Lektzian and Souva (2001). For the effects of sanctions on currency collapses, see Peksen and Son (2015). See also Shin et al. (2016) for the impact of sanctions on foreign portfolio investment.

⁵ See also Dreger et al. (2016), who argue that economic sanctions and oil prices affected Russia’s ruble.

⁶ From a theoretical and empirical standpoint, we focus on the short-term effects of sanctions on stock markets. A long-term buy-and-hold stock strategy is possible, but in the context of sanctions, we expect investors to respond quickly in a risky milieu and thus our study focuses on short-term decision-making.

⁷ In this paper we deal exclusively with imposed sanctions and leave a robust treatment of the effect of threats to future research. We did conduct a preliminary exploration of the effect of threats, but that investigation revealed no correlation between the threat of sanctions and monthly changes in market capitalization in targeted countries.

⁸ See the lists of specifically designated sanctioned individuals developed by the United States and the United Nations Security Council at https://www.treasury.gov/resource-center/sanctions/Pages/default.aspx. and https://scsanctions.un.org/fop/fop?xml=htdocs/resources/xml/en/consolidated.xml&xslt=htdocs/resources/xsl/en/consolidated.xsl., respectively.

⁹ Annual data would allow for a longer-term view but would introduce many potential intervening factors that occur in an entire year. Daily stock market fluctuations are notoriously volatile, and it is difficult to attach substantive meaning to the rapid swings in daily stock market trading. Monthly data are the sweet spot.

¹⁰ Market capitalization is frequently used as a proxy of stock market performance (Demirgüç-Kunt, Feyen, and Levine 2013; Larrain 2010; Pinto, Weymouth, and Gourevitch 2010; Rajan and Zingales 2003) because it refers to the value of the listed shares on a country’s stock exchange (Demirgüç-Kunt , Feyen, and Levine 2013).

¹¹ Because there may be concerns about using data in current U.S. dollars, we include the consumer price index as a control variable, a key indicator of inflation. Additionally, for the period under study (1990–2005), the U.S. inflation rate averaged 2.83% (median 2.70%) and never exceeded 6.10%, underscoring the minimal inflation level. We also control for exchange rates, limiting apprehensions about differences in exchange rates affecting the measure. We also perform an empirical diagnostic whereby we interact sanctions with exchange rates to ensure that the effect of sanctions is not conditional on the value of exchange rate. The interaction term in this regression is insignificant.

¹² See http://data.worldbank.org/data-catalog/global-economic-monitor. In the Supplemental Appendix we list the countries along with a brief description of the source index for that country (Table SA1). The World Bank has stock market data on 80 markets covering the years 1990 to 2018. Thirteen indices are unusable either because they came into existence after 2005, when this study ends, or because they represent markets in non-independent entities. This leaves a population of 67 countries with stock markets during the years under observation. Personal e-mails with representatives from the World Bank confirmed that these 67 countries represent the population of stock markets with daily averages in Haver and Bloomberg databases during the years of observation. We use 55 countries in the models estimated with minimal control variables. In the models with a full set of control variables, the number of countries falls to 52, due to missing data on other variables.

¹³ Another argument is that a stock market rally may occur if “the use of military force propels international traders to buy stocks instead of alternatives such as gold or government bonds” (Schneider and Troeger et al. 2006, 628).

¹⁴ The consumer price index reflects the change in prices for the average consumer of a constant basket of consumer goods. Total Reserves are measured in current U.S. Dollars. The World Bank defines Total Reserves as “Total reserves minus gold comprise special drawing rights, reserves of IMF members held by the IMF, and holdings of foreign exchange under the control of monetary authorities.” The data come from the World Bank (2016a).

¹⁵ Consistent estimation of the Bewley ECM requires an instrumental regression with $\mathrm{\Delta }{Y}_t$ regressed on lagged values of$Y_t$, contemporaneous values of $X_t$, and$\mathrm{\Delta }{X}_t$.$\mathrm{\Delta }{X}_t$is then regressed on instrumented values of $\mathrm{\Delta }{Y}_t$ along with contemporaneous and differenced values of $X_t$ (De Boef and Keele 2008).

¹⁶ Additionally, because we are interested in estimating changes occurring within a country’s stock market, we estimate our models with a within group estimator using country and year/month fixed effects (Bailey, Strezhnev, and Voeten 2017; Rabe-Hesketh and Skrondal 2012, 96). To test for robustness to different estimators, we also use a random effects regression with robust standard errors, a multi-level modeling approach with random coefficients, a General Estimating Equations (GEE) model with robust standard errors and an AR1 correlation, and Panel Corrected Standard Errors (with a common and panel specific AR1 term). The results, which are available in a supplemental appendix, are robust to these different estimators.

¹⁷ Similar to an interaction or quadratic equation, the first order coefficient provides the slope of a tangent line to the cubic function at the value of zero. The second order indicates the curvature (concave up or concave down) at the value zero (Cohen et al. 2003, 209).

¹⁸ We use Imai , Kim, and Wang's (2019) PanelMatch R package for estimation in this section. It can be found here: https://github.com/insongkim/PanelMatch.

¹⁹ We chose 5 sanctions as the cutoff because the sample mean is 4.5 and because in Table 1 we found that the effect of new sanctions is no longer statistically significant when the number of existing sanctions exceeds 4. Conceptually, the treatment tests the effect of a new sanction when the existing number of sanctions is low.

²⁰ In this refinement method, units are first assigned a propensity score and weights are then assigned to control units based on the size of the differences in scores between treatment and control observations. Higher weights are given to units that have smaller differences (Kim et al. 2020).

²¹ Chaudoin , Hays, and Hicks (2018, 912) found that “better balance was not associated with a decreased false positive rate.”

²² We modeled our covariate balance graphs on syntax from Kim et al. (2020) and Allen (2020).

²³ Financial and export sanctions showed the worst levels of covariate imbalance. Even after matching and refinement, two variables remained above.5 in covariate imbalance in the export sanction graphs and one variable remained above .5 in the financial sanctions graph.