Schedule - Parallel Session 3 - Gender Differences

There are at least two possible reasons for why people demand redistribution: immediate self-interest, and concerns for distributional justice. A voter with a high income should oppose more redistribution from high to low income individuals, while a voter with low income should support it. Pure concern for distributional justice however does not depend on self-interest. While many voters may favor more equal distributions per se, some might worry that it leads to deadweight loss, reducing overall wealth. They might prefer a higher level of wealth even if it comes at the cost of inequality. This preference for efficiency over equality is an example of a notion of justice. Interestingly, women tend to be more in favor of redistribution than men (Edlund and Pande 2002). Is this based on self-interest, or do they have different ideas about distributional justice? In Becker et al. (2013a,b), we proposed a theory to explain giving behaviour in dictator games by a combination of self-interest and preferences over just distributions. We used a dictator game with varying transfer rates as in Andreoni and Vesterlund (2001) and Andreoni and Miller (2002) to measure self-interest, and a social planner (Dickinson and Tiefenthaler 2002) and a veil of ignorance game (Rawls 1971, Schildberg-Hörisch 2010) to learn about individual notions of justice. In this paper, we analyze the results of this experiment with a specific focus on gender differences. In our dictator game, we replicate the result of Andreoni and Vesterlund (2001): we find that men react more strongly to changes in the transfer rates. However, while Andreoni and Vesterlund (2001) find that men are somewhat more selfish as they give less on average, we find that men give more on average, although the difference is insignificant. We find that gender differences in the social planner and veil of ignorance games are even stronger than in the dictator game; in particular, women are far more concerned with equality than men. While this result may not be surprising per se, the combination of the two experiments allows us to test if differences in giving behavior in the dictator game are due to differences in self-interest between men and women. We find strong evidence that this is not the case: Differences in the dictator game can be explained by differences in the notion of justice. In situations where women give more money than men, this is not because men are more selfish, but because men feel that the situation does not warrant passing a great amount as it would be inefficient to do so. In situations where men give more money than women, this is not because women are more selfish, but because women feel that passing a great amount would not be just as they prefer more equal outcomes. We employ both parametric (CES estimates) and novel non-parametric techniques developed in Becker et al. (2013a) and partially inspired by Karni and Safra (2002b), both of which confirm the results.

How heterogeneous are individual preferences? Are men’s preferences more heterogeneous than women’s? Do preferences become less diverse with age? We use revealed preference analysis to answer these questions. In particular, we investigate the nature of the revealed different relation: two people are revealed different if their individual decisions considered jointly are less rational than their individual decisions considered separately. We show that the revealed different relation is useful not only in characterizing the heterogeneity of preferences non-parametrically, but also in evaluating the potential biases generated by imposing additional assumptions on preferences. Using data from a random sample of Dutch population we find that: 1) heterogeneity is a persistent feature of the data, 2) men have more diverse preferences than women and younger people while different from older people are not more diverse, 3) Stricter assumptions on preferences lead to an underestimation of the heterogeneity of preferences. The revealed different relation as defined above is complete and symmetric and it is well defined whether preferences are consistent with rationality or not. This definition is most appealing when the measure of distance to rationality is monotonic on the number of individual decisions (as in Afriat’s critical cost efficiency index). Since the revealed different relation is a summary statistics that does not require making parametric assumption about preferences, the relation can be used to directly test if observe heterogeneity in preferences is due to observed demographic differences. We find little evidence that heterogeneity in preferences is due to observed individual characteristics. Finally, we find that the revealed different approach is most informative when multiple decisions per individual are available. When only cross-sectional data is available (i.e. each individual trivially satisfies rationality) lack of differences between individuals might be due to lack of power to detect deviations from rational behavior.