Sanctioned to Death? The Impact of Economic Sanctions on Life Expectancy and its Gender Gap

Abstract

We empirically analyse the effect of UN and US economic sanctions on life expectancy and its gender gap in target countries. Our sample covers 98 less developed and newly industrialised countries over the period 1977–2012. We employ a matching approach to account for the endogeneity of sanctions. Our results indicate that an average episode of UN sanctions reduces life expectancy by about 1.2–1.4 years. The corresponding decrease of 0.4–0.5 years under US sanctions is significantly smaller. In addition, we find evidence that women are affected more severely by the imposition of sanctions. Sanctions not being ‘gender-blind’ indicates that they disproportionately affect (the life expectancy of) the more vulnerable members of society. We also detect effect heterogeneity, as the reduction in life expectancy accumulates over time and countries with a better political environment are less severely affected by economic sanctions. Finally, we provide some evidence that an increase in child mortality and Cholera deaths as well as a decrease in public spending on health care are transmission channels through which UN sanctions adversely affect life expectancy in the targeted countries.

Acknowledgements

We thank Elisa Aranda, Lukas Böker, Nehal Brain, Marek Endrich, Ines Kalai, Ekaterine Lomtatidze, Stephan Michel, Ahmed Mostafa, Hashem Nabas, Ana Odorovic, Konstantinos Pilpilidis, Thomas Plümper, Georg Ringe, Lamis Saleh, Julia Samwer, Gerald Schneider, Jawaher Skhiri, Stefan Voigt, Orlin Yalnazov, two anonymous referees, and participants of the 2017 Silvaplana Workshop on Political Economy, the 2018 Meeting of the European Public Choice Society, the 2018 European Political Science Association Conference, the 2018 Congress of the German Economic Association, the MACIE Research Seminar in Marburg, the faculty seminar at the Gutenberg School of Management & Economics in Mainz, and the economics research seminar at the British University in Egypt for helpful comments on earlier versions of this paper. Data and codes are available on request from the corresponding author. The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary Materials

Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2020.1746277

Notes

1. Studies on the human rights consequences of sanctions also show only very specific effects that are in no way reflective of the total negative impact on the population. See Gutmann, Neuenkirch, and Neumeier (2019) for empirical results and a survey of the literature.

2. The third dimension of the Human Development Index, education, is not well-suited for identifying a causal effect of sanctions on human well-being. Education is a highly persistent quantity that does not react quickly to adverse shocks due to sanctions. In addition, there is no broad and reliable database with annual education data that also covers countries subject to sanctions (see Barro & Lee, 2013). The same problem arises with the use of happiness data (see DiTella & MacCulloch, 2006).

3. A thorough discussion of the advantages of entropy balancing over other matching approaches can be found in Hainmueller (2012). He demonstrates, using Monte Carlo simulations as well as empirical applications, that entropy balancing outperforms other matching techniques, such as propensity score matching, nearest neighbour matching, and genetic matching, in terms of estimation bias and mean square error.

4. In that, entropy balancing is very similar to the synthetic control method pioneered in Abadie and Gardeazabal (2003) and Abadie, Diamond, and Hainmueller (2015). The main difference is that the latter method is applied when only one unit is subject to treatment. In such a case, the synthetic control method is used to construct a synthetic control group that resembles the sole treated unit. In contrast, entropy balancing applies to cases where the number of treated units is larger than one. The synthetic control group constructed using entropy balancing resembles the average unit exposed to treatment.

5. All variable definitions and data sources can be found in Table A1 in the Appendix.

6. The Census Bureau argues that ‘as a result of single-year age and calendar-year accounting, IDB data capture the timing and demographic impact of important events such as wars, famine, and natural disasters, with a precision exceeding that of other online resources for international demographic data.’

7. The list of countries in our sample can be found in Table A2 in the Appendix.

8. Information on the sanctioned countries can be found in Table A2 in the Appendix.

9. By controlling for the gender gap in schooling, that is, the total years of schooling for women of age 15 and older minus that for men of age 15 and older, we implicitly take into account de facto women’s rights. This is important because Neumayer and Plümper (2007) only find an effect on the gender gap in life expectancy that is conditional on women’s rights. Knowles, Lorgelly, and Owen (2002) link the gender gap in education directly to adverse health and development outcomes. An advantage of using this indicator as a proxy for women’s rights in general is its far superior country and time coverage compared to other indicators.

10. By controlling for overall life expectancy and its gender gap we implicitly control for life expectancy of men and women. If both genders constitute roughly 50 per cent of the total population, life expectancy of men (women) can be obtained by subtracting (adding) 50 per cent of the gender gap from (to) overall life expectancy. The same considerations apply to overall schooling and its gender gap.

11. Adding country-fixed effects in the first step of the matching algorithm is not feasible as using these to compute the vector of weights would imply that all countries that were never subject to sanctions would receive a weight of zero and thus be discarded.

12. Our sample contains 13 observations with UN and US sanctions in place in the same country and year.

13. Life expectancy of both men and women is increasing by roughly one year over the five years before a sanction episode starts. This implies that countries that are about to be sanctioned are experiencing an upward trend in life expectancy that is similar to their non-sanctioned counterparts.

14. Information on the countries in the weighted control group can be found in Table A2 in the Appendix.

15. The high level of significance of the tests for the gender gap in Column (3) is due to the SUR framework, which takes into account the covariance of the point estimates for men and women.

16. Accounting for potential non-linearity by adding an interaction term with the number of years squared in which sanctions are in place yields qualitatively similar results.

17. The costs to target variable is demeaned so that the baseline effect of UN and US sanctions is to be interpreted at the average value of the costs to target.

18. This variable is generated by applying a principal component analysis to the Polity2 indicator (loading: 0.18), the Political Terror Scale (loading: – 0.70), and an indicator variable for the occurrence of conflicts (loading: – 0.69), and explains 51 per cent of the total variation in these variables. Given these loadings, an increase in the variable implies a better political environment. To facilitate interpretation, we normalise the principal component to a mean of 0 and a standard deviation of 1. Adding the individual variables as interaction terms to the baseline model does not yield significant coefficients.

19. Employing separate indicators for UN sanction threats and US sanction threats is not feasible due to the very low frequency of UN sanction threats.

20. One reason for these differences might be the different composition of the treatment group and the resulting weighted control group in the placebo tests.

21. To conserve space, we do not report these results in detail. They are available on request.

22. To conserve space, we do not report these results here in detail. They are available on request. All UN sanction episodes in our dataset take place after 1991. There are at least two reasons for this. First, UN sanctions have to be enacted by the UNSC with its five veto powers. Hence, it presumably was difficult to agree on the imposition of a sanction during the Cold War. Second, for some of the countries important control variables became available only in the later sample period.

23. To conserve space, we do not report these results in detail. They are available on request.