Preference-Shifting Violence: Evidence From The Marikana Massacre

Abstract

Objective: This study examines how the risk and time preferences of a sample of South African adults were affected by the Marikana massacre, a prominent episode of police violence that took place in South Africa on August 16, 2012. Previous studies have found that exposure to violence affects risk and time preferences (e.g., Voors et al., 2012; Jakiela & Ozier, 2019), but the direction of the effect is unclear. Our study contributes to this discussion and highlights a channel through which police violence – a sporadic but persistent phenomenon in many societies – can affect human decision making.

Methodology: The Marikana massacre intersected fieldwork for the second wave of a large longitudinal survey being conducted in the Gauteng region, creating plausibly exogenous variation in respondents’ exposure to the event. We exploit this natural variation to estimate the causal effect of the exposure to violence on individual risk and time preferences using Pooled OLS regression models. To corroborate our findings, we examine how risk and time preferences relate to a proxy measure of interest in the Marikana massacre that is based on the intensity of internet searches for the term “Marikana”. We further test the robustness of our results by controlling for other contemporaneous events in South Africa that may act as confounding factors, and we address potential unobserved heterogeneity by re-estimating our models with individual and area-level fixed-effects.

Originality: To the best of our knowledge, this study is the first to examine the impact of witnessing the Marikana massacre on risk and time preferences among South African adults. In contrast with prior research that examines long-run effects of violence on preferences or relies on retrospective recall, our design captures real-time behavioural consequences as the violent event and its aftermath unfold. We further contribute methodologically by using a measure of public salience of the event that relies on high-frequency internet search data to validate the timing-based identification strategy inherent in our natural experiment framework.

Results: We find a statistically significant difference in risk and time preferences between individuals surveyed before and after the Marikana massacre, with individuals surveyed after the massacre being more risk seeking and having higher discount rates. Our pooled OLS estimates indicate that among respondents interviewed after the massacre, risk aversion and the discount rate both are higher of each standard deviation. Estimates from fixed-effects models that account for unobserved heterogeneity suggest that immediately after the massacre these effects are substantially larger.

Practical implications: Out study provides strong evidence that witnessing a violent event alters risk and time preferences, two key determinants of a wide range of decisions – such as saving, investment, and health behaviour – that ultimately affect long-term welfare and development outcomes. If witnessing violence induces individuals to become more risk-averse or short-term oriented, this could reduce human capital accumulation, discourage entrepreneurship, or lead to unhealthy behaviour. From a policy perspective, our study thus highlights the broader developmental costs of police violence, and raises the possibility that such violence may contribute to widen inequality where minorities are disproportionately affected by police violence.