Determinants of risk behaviour: three laboratory experiments on peer effects, group identity and incentive schemes

Abstract

Risk is inherent in many social and economic decisions, such as the choice of pathway in secondary school, the choice of major at university, job decisions, health-related behaviour, marriage, parenthood, migration and the allocation of financial assets. Investigating the determinants of attitudes towards risk is therefore essential to fully understand how people make such decisions. Recent research has shown that individual risk attitudes are not immutable personality traits, but are influenced by external factors with the potential to change them in more or less enduring ways, such as the characteristics of the environment, emotional states, life experiences such as poverty, job loss or violence, and social relationships. This thesis studies external factors that play a role in shaping risk attitudes. Specifically, it focuses on two important environmental factors: social relationships and the incentive structure that individuals face (e.g., competition or teamwork). It is composed of three chapters. Each chapter of the thesis presents the results of a different laboratory experiment, in which individual risk behaviour is always measured using the Bomb Risk Elicitation Task - BRET (Crosetto and Filippin, 2013). This task asks participants to choose how many boxes to collect out of 100, knowing that 99 boxes contain £0.10 while one contains a bomb, but without knowing in which box the bomb is located. They can therefore choose their preferred lottery among 100 lotteries whose outcomes and probabilities are fully described only by one parameter, i.e., the number of collected boxes. Earnings increase linearly with the number of boxes collected, but they are all lost if the bomb lies in one of the collected boxes. In the first two chapters, risk behaviour is measured both before and after the treatment manipulation, and feedback on the peers’ ex-ante risk behaviour is used as a channel to study peer influence on the subjects’ ex-post risk behaviour. The first two chapters provide new evidence that individual risk behaviour is influenced by the risk behaviour of the peer group and offer one explanation for why peer effects are not always present and vary in intensity. This is due to the fact that individuals are more influenced by those peers with whom they feel more bonded. Specifically, in the first chapter I study how group identity (that is, the portion of an individual’s self-concept derived from the sense of belonging to the social group) affects peer effects on risk behaviour. I induce different levels of group identity through different matching protocols (random or based on individual painting preferences) and the possibility of interacting with group members via an online chat in a group task. I find that subjects are affected by their peers when taking decisions and that a stronger group identity amplifies the influence of peers: painting preferences matching significantly reduces the heterogeneity of risk behaviour compared with random matching. On the other hand, introducing a group task has no significant effect on behaviour, possibly because this interaction does not always contribute to enhancing group identity. The second chapter digs deeper into this evidence by investigating the role of the incentive structure that characterizes the individuals’ environment. Since the first chapter shows that peer effects vary in intensity, I hypothesize that different types of incentive schemes may have different effects on peer relationships and, therefore, affect peer effects on risk behaviour. Using a real effort task, which consists of recognizing the value and the country of origin of a random sequence of Euro coins, I compare piece-rate compensation first with a cooperation-based and then with a competition-based incentive scheme. I find that competition significantly reduces attachment to peers and more than halves peer influence on risk behaviour compared with piece-rate compensation, despite the fact that the latter effect is not statistically significant. Such findings suggest that, when designing and evaluating an optimal compensation scheme, it may be important to also consider how peer effects on subsequent risk behaviour will in turn affect future decisions involving risk. For example, in research and development, competition may improve the results of current projects, but risk attitudes will shape the types of future projects that are attempted. The third chapter restricts the attention to competition and enquires whether this type of incentive scheme has a direct effect on risk-taking behaviour, beyond any social comparison, and whether its impact on subsequent risk behaviour is heterogeneous according to gender. Risk behaviour is measured after the performance of a real effort task, consisting of recognizing the value and country of origin of Euro coins, incentivized either as a tournament with fixed rewards or as a random draw with the same monetary payoffs. The data show that competition does not significantly affect subsequent risk-taking behaviour when considering the full sample. However, there is a positive relationship between competition and risk aversion for males, who become significantly more risk-averse after losing a competition than after randomly earning the same low payoff. In contrast, males do not become more risk-seeking after winning the tournament, while the average risk-taking behaviour of females is unaffected by tournament participation and outcomes. The reaction of males to negative outcomes might be driven by intrinsic motives, such as emotions or a shift in the locus of control from internal to external. Overall, the evidence presented here shows that risk attitudes are not immutable but may be shaped by external factors. Of particular importance is the role played by the risk behaviour of peers, which begins to emerge even when bonds are weak and becomes stronger as the social link intensifies. Any policy that aims to change risk attitudes (or that does so indirectly) will thus see its effects spread to the target subjects’ peers, and may amplify its success if the peer group is chosen wisely. Changing the characteristics of the subjects’ environment by introducing competition weakens their attachment to the competing peers and may attenuate peer effects on risk behaviour. In addition, competition per se has no impact on subsequent risk behaviour, except for males who become more risk-averse after losing.