Aversion to loss and inequality in the evaluation of policies
OPUS 20, dr hab. Martyna Kobus, prof. INE PAN
The aim of this project is to develop testable criteria for evaluating political interventions in terms of loss aversion and inequality aversion. Inequality aversion is a standard property used to assess the effects of policies (e.g. aversion to income inequalities that policies may cause), but loss aversion has so far not been considered in the literature on policy impact assessment. Recent work by the project manager and co-authors brings loss aversion to the field of policy evaluation. This project is a continuation of this work incorporating both loss aversion and inequality aversion. Loss aversion, i.e. the fact that losses hurt more than gains (of the same magnitude), has been identified as a feature of individual preference in a wide variety of contexts. Many researchers have long noted the importance of loss aversion in political processes. For example, it plays a role in setting trade policy and helps explain why industries that suffer losses are more likely to be protected than industries with similar growth. Other examples of how the electorate's loss aversion drives policy makers to action is the repeal of Obamacare. Given this literature and evidence of the importance of loss aversion in policymaking, the project manager and co-authors address the topic of incorporating loss aversion into the evaluation of policy interventions. They are developing new, testable criteria and econometric methods to rank the distributions of individual policy effects taking into account loss aversion. These criteria are referred to as Loss Aversion Sensitive Dominance (LASD). It differs from standard first-order stochastic dominance (FOSD) in that the dominant distribution can be above the dominated distribution in the gains region as long as it lies sufficiently below the dominated distribution in the loss region. However, usually the policymaker is not only interested in loss aversion. Usually, he also wants to take into account that the final effects of a given policy do not generate too large income disparities. However, this essentially requires a two-dimensional model, which poses many technical and interpretation challenges. In such an environment, the decision maker is at the same time averse to losses and income inequality. The basic welfare function is expected value, with the value function having the property of loss aversion and inequality aversion. Our goal is to prove equivalence theorems, that is, to relate a class of welfare functions that cannot be observed in practice to distribution criteria of policy effects that can often be identified. We begin this problem by considering the case of additively separable preferences proposed in the loss aversion literature. Individuals derive utility from both income and its changes, and their utility function is additively separable in them. Preliminary research in this direction has already begun. However, in this case the relationship between levels and changes is neglected, while the social planner may be more inclined towards a given policy depending on whether losses are concentrated among the rich or the poor. To address this general problem, we will refer to the literature on multidimensional well-being, inequality and poverty. However, the insights we get from this will be limited because in our case these two dimensions are not treated symmetrically, which is usually the case in this literature. In addition, we deal with the case of both point and partial identification of the distribution of policy effects and prove our results in both settings. This project is a continuation of a research topic recently started by the project leader and co-authors of the inclusion of loss aversion in policy impact assessment. This is a new research topic that relates to and contributes to various strands of literature, namely the literature on stochastic dominance, loss aversion, econometric policy impact assessment, and multidimensional measures of inequality, well-being and poverty.