Using Structural Modeling in Experimental Design
A presentation delivered at the “Culture Matters” seminar on March, 13. Alexis Belianin, Ph. D., is the Head of Laboratory for Experimental and Behavioral Economics.
Alexis Belianin, Ph. D., is the Head of Laboratory for Experimental and Behavioral Economics.
Dr. Belianin presented on the use of structural equation modeling (SEM) in experimental design in psychology.
SEM describes interactions between objects (students, schools, firms, and other agents/players) as choices made between available strategies. Available strategies are defined by the researcher based on their theoretical understanding of the interaction.
Structural models allow for rigid operationalization of variables and mathematical formulation of the hypothesis, which makes these hypotheses easy to test. A drawback of the approach is that it has to draw from theory – in other words, to describe an interaction with SEM, one has to have a clear theoretical understanding of the interaction beforehand. That makes SEM different from reduced-forms models that are data-driven and explore statistical relationships (comparison of the means, analysis of variance, regressions, etc.).
Dr. Belianin concluded his lecture by presenting two SEM experiments. Andreoni and Miller have studied prosocial choices in the Dictator Game – a game in experimental economics that allows on player to unilaterally determine the distribution of rewards. Drawing on the assumption that individuals are rationally self-interested, one would assume that the Dictator would allocate all of the resources to themselves, however, experimental data shows that the Dictator only takes 70 to 80%. The Dictator’s behavior was modeled with an SEM that evaluated the altruism of the player and the costs of prosocial behavior.
The second experiment (Cappelan et al., 2007) has studied the standards of honesty. Participants were asked to make investments and then were asked to share the common returns. The model used to describe the decisions included three strategies reflecting different ideas of honesty: egalitarian (returns shared equally), libertarian (returns from my investments are mine), and mixed (returns shares are in proportion to the initial investment). The solution of the model predicted the participants’ behavior based on their investment, the profit, personal importance of honesty and the idea of honesty.