On the Difference between Social and Private Goods

Abstract

Standard economic models have long been applied to choices over private consumption goods, but have recently been extended to incorporate social situations as well. We challenge the applicability of standard decision-theoretic models to social settings. In an experiment where choices affect the payoffs of someone else, we find that a large fraction of subjects prefer randomization over any of the deterministic outcomes. This tendency prevails whether the other party knows about the choice situation or not. Such randomization violates standard decision theory axioms that require that lotteries are never better than their best deterministic component. For conceptually similar choices in classical non-social situations, we do not find much evidence for such violations, suggesting the need for theories of uncertainty that are targeted to social settings.

Revealed Preferences for Randomization: An Overview

“…Similar methods have been employed in other domains. For social preferences, subjects in Sandroni, Ludwig, and Kircher (2013) choose between two allocations (€7.5 only to themself or €5 to both) and a coin flip. About a third of participants choose to randomize, while this percentage drops to 6 percent when the 2 allocations involve only consumption goods for the self (money and a mug).…”

Ex-ante commitments to “give if you win” exceed donations after a win

“…to such contexts (e.g., DellaVigna et al 2012;Sandroni et al 2013;Vesterlund 2003, as well as in numerous laboratory experiments) using the expected utility framework. 10 For any model that can be expressed in terms of expected utilities over outcomes the timing and uncertainty of the decision (i.e., whether it is a sure thing or a prospect) is irrelevant to the individual's choice.…”

“…obtained in a given state of the world may depend on unrealised donation commitments from the past, as such commitments may still have an impact on beliefs. Also, the findings of Sandroni et al, 2013 could be understood as evidence for the presence of (self-)signaling or other non-outcome based motives. 11 To analyse this point, we offer a simple signaling model (see Appendix A.2 for derivations and extensions to additional settings) with two types of agents (or two types of self): one who gets an inherent benefit from donating to the charity (a "good type"), and one who does not (a "bad type").…”

“…13 11. Nearly a third of Sandroni et al (2013)'s dictator subjects demonstrated a strict preference for a coin flip between increasing their own and another's payoff, preferring this to getting either alternative with certainty. This choice may be driven by a diminishing returns to both private consumption and the benefits of the pro-social choice, (e.g., self-signaling, impact, or warm glow), if a commitment to donate with some probability itself yields these benefits.…”

“…According to Reinstein "if individuals are not [expected utility] maximizers over outcomes but gain warm glow utility from unrealized commitments, this decline could be attributed to satiation of warm glow". In laboratory dictator games Brock et al (2013) and Sandroni et al (2013) each found that social preferences and fairness concerns appear to depend on a combination of ex post and ex ante concerns. Smith (2011) found that giving (to other subjects who had incurred an income loss) was higher when giving decisions were made using the strategy method than when subjects were asked ex-post.…”