Long-term Well-being among Survivors of the Rwandan and Cambodian Genocides

Abstract

This paper adds to the thin empirical literature estimating the long-term effects of exposure to conflict from in utero to adolescence on adult well-being. The effects through adolescence of the two worst genocides in recent history – those occurring in Rwanda (1994) and Cambodia (1975–79) – are examined. The Rwandan genocide is shown to have produced long-term health outcomes among women exposed to the conflict during adolescence. A further contribution is the analysis of gendered effects during adolescence, which is enabled by the availability of data on men’s height for Rwanda. The long-term effects are confirmed for men, however this appears to be the consequence of exposure during adolescence later than for women, a result that is consistent with the biological literature on the differential timing of the onset of puberty by gender. No significant effects are detected in the case of the Cambodian genocide and we discuss some issues that may influence this result. Although more research and better data are needed, our results are suggestive of adolescent-specific effects of the Rwandan genocide, which may be comparable or larger than those previously found for younger children.

Acknowledgments

The views expressed in this article are solely those of the authors. They do not necessarily reflect those of their institutions or their Boards. The authors would like to thank Richard Akresh, Goran Holmqvist, Marinella Leone and two anonymous referees for their constructive feedback and suggestions on earlier drafts. The authors also acknowledge participants in the joint UNICEF-CIFAR workshop on Adolescents' Wellbeing and Genocides, October 2017, held in Florence, Italy.

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/10.1080/00220388.2021.1919630

Notes

1. Citing Prentice et al. (2013): ‘adolescence represents an additional window of opportunity during which growth-promoting interventions, possibly initiated years before puberty, might yield substantial life cycle and intergenerational effects.’ (p. 911).

2. This question can only be answered for Rwanda due to the availability of data on men’s height for Rwanda and not for Cambodia.

3. Akresh et al. (2011) and Akresh, Lucchetti, et al. (2012), analyse the lasting impacts of shocks among children under five in Rwanda and Eritrea, respectively. They examine the impacts by gender; however they focus on young children, not adolescents. This is likely because surveys often include anthropometric measurements for young boys and girls.

4. Other studies analyse the effects of exposure to shocks at different ages on educational attainment, such as Akresh and de Walque (2008), Islam et al. (2016), León (2012), Caruso (2017), Caruso and Miller (2015), and Shah and Steinberg (2017). While focusing on the effects of the Cambodian civil war and genocide on educational attainment and fertility, Islam et al. (2016) also examine the effects on a few health outcomes (including height) for women exposed during primary school age (using older individuals who had already completed primary school during the conflicts as a control group) and finds no impact.

5. Akresh, Bhalotra, et al. (2012, 2021) use ethnicity to measure war exposure which mitigates issues related to migration since a person’s ethnicity does not change if she/he migrates.

6. Estimates of the casualties of the bombing campaign range from 150,000 to 500,000 (Moyano, 2018). Islam et al. (2016) report estimates of deaths during the first half of the 1970s ranging between 30,000 and 500,000.

7. DHS are available for Cambodia also in 2010 and 2015, and for Rwanda for the years 2000 and 2005 (a survey was also conducted in 1992 but did not record height measurements). The selection of DHS for our analysis is mainly motivated by the timing of the DHS waves in the two countries relative to the timing of the respective genocides and the fact that they report height for women aged 15 to 49 only. We use the earliest rounds of the Cambodia DHS which are the closest in time to the genocide in order to capture most survivors, as some of those exposed decades before the survey years may not have survived. In the case of Rwanda instead, we exclude earlier surveys because individuals exposed at a young age in 1994 would not have been surveyed. In fact, those exposed to the genocide aged 8 or younger would not have been sampled in the 2000 DHS, while those aged 4 or younger during the genocide would not have been sampled in 2005. We conduct robustness checks using all rounds of the Rwandan DHS and results are consistent with the ones obtained using the selected surveys (Supplementary materials Appendix 5, Table A5a).

8. See https://genodynamics.weebly.com/ (Accessed 09.02.2019) for more information on the project, data, methodology and sources.

9. See Supplementary materials Appendix 3 for the distribution of killings across provinces.

10. For more information on these datasets, see CGDB (Cambodian Genocide Databases), Genocide Studies Program, Yale University, New Haven, CT, https://gsp.yale.edu/cambodian-genocide-databases-cgdb; Databases, Documentation Center of Cambodia, Phnom Penh, Cambodia.

11. The estimates obtained using alternative definitions of killings intensity in Cambodia using data from Islam et al. (2016) and Etcheson (2000) separately rather than combined confirm the results obtained using the combined measure (that is no long-term effects on height). Results available upon request.

12. See Supplementary materials Appendix 1 for the methodology used to ensure full correspondence between the administrative divisions of 1975–79 and those of the survey years.

13. However, this variable does not capture how many times the household moved or whether it moved during the genocide, before or after (we only know how many years before the survey the household moved).

14. Akresh and de Walque (2008) document that about 88.5 per cent of individuals currently reside in their province of birth.

15. This check is shown in Supplementary materials Appendix 5.

16. We also run robustness checks with different age groupings and control group (see Section 6.2).

17. On the relationship between birth order and height, see for example Jayachandran and Pande (2017), and Myrskylä, Silventoinen, Jelenkovic, Tynelius, and Rasmussen (2013).

18. Using this alternative strategy, the share of women in the control group is 20 per cent (as opposed to 7 percent using the original strategy).

19. As described in Section 3, there were episodes of pre-genocide conflict also in the case of Rwanda, but these were limited geographically to the northern provinces of Byumba and Ruhengeri. Results are qualitatively similar when excluding those provinces from the sample (Supplementary materials Appendix A5).

20. Non-parametric results for both women and men are consistent with the regression results for Rwanda (see Supplementary materials Appendix 4).

21. For both countries, we cannot identify clear differences between the results obtained using the sample of movers and non-movers (the latter estimates shown in Supplementary materials Appendix 5, results for movers available upon request). In the case of Rwanda, we do not find differential results for the age groups 7–12 and 13–16 across movers/non-movers. For the age group 17–19, we see that non-movers achieve lower height, while this is not found among movers (however, this is the case only when using dead sibling as proxy for intensity of exposure, not when using killing shares). In the case of Cambodia, the estimates are not statistically different between movers/non-movers. Based on these results and considering the small samples and that we do not have information on the full migration history, we are not able to assess whether migrating or not had a positive or negative effect on health.

22. Adult mortality rate is the ‘the probability of dying between the ages of 15 and 60’ and it is expressed as number of deaths per 1000 individuals (World Bank WDI indicators 2019). Ten years before the start of the genocides, it was 480 and 540 for Cambodia, and 336 and 391 for Rwanda (for females and males, respectively). Mortality rates increased steeply during the genocides, reaching 796 and 921 in Cambodia (in 1977, the year with the highest indicators of mortality), and 551 and 607 for Rwanda (for females and males, respectively). Infant and under-5 mortality rates (per 1000 live births) depict a similar picture however these are available for Cambodia only starting in 1975 (177 and 308 for Cambodia in 1975, and 127 and 284 for Rwanda, respectively).

23. https://mics.unicef.org/methodological_work/5/POST-EMERGENCY.