Risk Preferences Under Extreme Poverty: A Field Experiment

Abstract

The puzzle between poverty and risk aversion is still incomplete as there is no theory that can reconcile all t he behaviors document ed so far. This paper emerges from t he fact t hat , recent experimental evidence, found in psychology and economics, suggests that some decisions made under risk respond to reference points (status quo and others) challenging the theory of Expected Utility. To explain such behaviors, several hypotheses have been raised, of which the Cumulative Prospect Theory of Kahneman and Tversky stands out. Yet, even this theory still presents some limitations. This paper develops a theory where individuals respond to income references, which result of diverse income related exclusion mechanisms, so referenced behaviors can be consistent with rational choices. According to this theory some income references affects the behavior, either by being near them or by deciding over high stakes that involves them. T he extremely poor have the reference of the income that guarantees the consumption of calories to ensure a healthy and longer life. While the poor have the reference of the minimum income to avoid income determined exclusion. To test the reference of the extremely poor, an experimental exercise was conducted with 92 individuals from households living under poverty and extreme poverty in Bogotá, Colombia. Experiments on Baseline Risk and Loss Aversion were conducted with high stakes monetary payments; the highest was 23% of the country's monthly minimum wage. We fi rst analyze the selection of risky options according to an order according to the risk that each option implied. Additionally an alternative method is proposed that uses an average measure of risk aversion instead of ordered options to capture the fact that there are differences in the steps. We use an analysis of a Regression Discontinuity Design where, if there is a treatment like being undernourished determined by a critical value of single variable as income then, if all the other variables does not change around the critical value, any change in a dependent variable, for example risk aversion can be explained by the incidence of the treatment. In the Base Risk exercise some households were offered the opportunity to achieve the income enough to avoid undernourishment for one month while in Loss Aversion some framed options would result in entering undernourishment. The results suggest that being below and near the reference income decreases significantly the households risk aversion while the contrary holds when being near and above. T his can be understood as if the incentive to leave undernourishment significantly decreases risk aversion while avoiding it increases risk aversion.

Poverty is associated with both risk avoidance and risk taking: an empirical test of the desperation threshold model.

“…That should translate into a V-shape between risk taking and resources (Figure 1B). Most previous real-world studies only searched for an increase in risk taking when poverty increased (28,29,49,54). In our study, we aimed to simultaneously test the increase and the decrease.…”

“…In anthropology, (48) presented evidence of a U-shape between herd value and risk taking -but the small size of the sample (23 Andean farmers) makes it hard to draw conclusions. (49) tried to estimate a subsistence threshold in extremely poor neighbourhoods of Bogota, and found suggestive evidence of a jump in risk taking at that point -but again, the sample size was not sufficient to draw firm conclusions. In principle, though, any dataset containing resources and risk taking measures could be used to test our hypothesis, if it contains enough participants above and below the desperation threshold.…”

A decision‐making model with utility from anticipation and disappointment

“…Among these four cases, Cases i and iii are of special interest, as they describe two seemingly conflicting empirical observations that people with lower levels of wealth can be either more risk averse or more risk seeking (Caballero, ; Vieider, Chmura, & Martinsson, ). For Case i , it is straightforward to show that a ≤ CE ≤ Eμ implies Eμ ≥ (1 − γ ) CE + γa .…”