سرریز نوع دوستی: آیا رفتارها در آزمایش های مستقل از متن پیش بینی کننده نوع دوستی به سوی یک کالای عمومی به طور طبیعی است؟

This paper addresses the external validity of experiments investigating the characteristics of altruism in the voluntary provision of public goods. We conduct two related experiments that allow us to examine whether individuals who act more altruistically in the context-free environment are also more likely to act altruistically toward a naturally occurring public good. We find that laboratory behavior can be predictive of contributions toward naturally occurring goods, but not in a uniform way. In fact, parametric measures of altruism do a poor job of predicting which subjects are most likely to contribute to a naturally occurring public good.

A common criticism of using experiments to test economic hypotheses is that the laboratory environment is so sterile as to lose its relevancy for more complex “naturally occurring” markets and behaviors. For example, experiments investigating the characteristics of altruism in the voluntary provision of public goods typically use induced-values with neutral-contexts in which laboratory contributions to the “public good” are framed as an “investment” decision, with the internal and external benefits specified, rather than as a “contribution” decision. Because the external validity of this type of experiment has not been addressed, the question remains: are decisions that are made in the sterile laboratory environment using tokens, anonymous partners, and benefits that only accrue to a small number of subjects (those who are in the experiment concurrently) predictive of altruistic behavior toward naturally occurring public goods that involve benefits that accrue to the general public? We address this issue directly by conducting two related experiments. First, we conduct a context-free laboratory public goods experiment that is used to estimate the level of altruism exhibited by each subject. Subjects are paid their earnings from these induced-value experiments, and a follow-up experiment is immediately conducted in which subjects are given the opportunity to contribute to a naturally occurring local public good. Contributions to this local public good can be made in cash (either from money brought into the laboratory or experiment-earnings) or paid by check. To address potential ordering effects, we also reverse the sequence of the experiments, allowing subjects to contribute to the local public good first, followed by the context-free laboratory public goods. Our experimental design allows us to examine whether those individuals who are estimated to be more altruistic in the context-free environment are also more likely to contribute to a naturally occurring public good—i.e., to test for altruism spillovers.1 Results indicate that altruistic behavior in the laboratory can be predictive of contributions to naturally occurring public goods, but not in a uniform manner. While there is some degree of correlation between some simple, non-parametric measures of altruism estimated with the induced-value laboratory game (such as average tokens contributed across rounds) and the likelihood of exhibiting altruistic behavior toward the naturally occurring public good, other non-parametric measures are inconsistent with our priors. For instance, the more often a subject acts as a weak free-rider in the induced-value public goods game, the more likely she is to contribute to the naturally occurring public good. Furthermore, as describe in Section 3, our parametric estimates of altruism do a poor job of predicting giving to the naturally occurring public good. The next section presents the experimental design. Section 2 provides a parametric model that formally incorporates altruism and error in the public goods decision. Results follow in Section 3. The final section concludes.