Some issues of methods, theories, and experimental designs

• Published in: Journal of economic behavior & organization 2010, Vol.73 (1), p.24-28
• Main Author: Cox, James C.
• Format: Article
• Language: English
• Subjects: Economic behaviour; Economic theories; Experiment design; Experimental economics; Experiments; Games; Laboratories
• ISSN: 0167-2681; 1879-1751
• DOI: 10.1016/j.jebo.2009.01.014

Smith (2010) reports an example that illustrates how easily procedures can introduce a future into an experiment. The example is provided by the central result reported in Cox (2009). That paper explores a behavioral phenomenon that has not been included in theories of reciprocity or models of unconditional social preferences: the effects of social context on reciprocity, trust, and altruism. The weak social context treatment includes the investment game and two dictator control games in a triadic design (Cox, 2004) with double blind payoff protocol. The strong social context treatment includes the same three games and double blind protocol and two added features: (a) an unexplained second task; and (b) a coin flip random selection of the second task or the first task (investment game or one of the two dictator games) for money payoff.2 Introduction of the second task had no significant effect on first movers behavior in the investment game. But it significantly shifted behavior of second movers in the investment game and dictators in both the first-mover dictator-control game and the second-mover dictator-control game towards more generous giving: information about the existence of the second task caused subjects in these roles to behave as though they were more altruistic. The result is striking because the double blind protocol makes it impossible for a subject to have a reputation with other subjects or the experimenter and, hence, the random decision selection procedure should make behavior in the first task independent of information about the existence of the second task. This example involves two of the themes explored in Smith (2010), independence of history and future and context irrelevance, that can arise when subjects multiple decisions within an experiment are not independent. In the following pages, I will explore related questions that arise in theory and experiments when subjects make multiple decisions. These are questions of experimental design, theoretical interpretation of data, and behavioral impacts that arise from possible nonseparability of utilities: (a) among decision tasks in an experiment; (b) among risky choice options in an experiment; and (c) between decision tasks within an experiment and decision tasks outside the experiment. From a theoretical perspective, the issues appear to be somewhat different when the experiment involves decisions on a lottery space than when it involves decisions on a commodity space. [Copyright El