Strategic voting in the laboratory
Résumé
The claim upon which most experiments in laboratories are conducted is that the random allocation of a treatment among various groups or individuals makes it possible to draw accurate inferences about causality. The laboratory is a context insulated from outside influences so that variations in behavior can be attributed to differences in treatment. Yet, the laboratory is not remote from all noises characterizing the ‘real world’. The issue of internal validity is probably overlooked too often because threats to external validity appear to be the most important (McDermott, 2002; Green and Gerber, 2004). Most experiments in political science are faced with potential threats to internal validity which are often unavoidable. The good news is that most of these problems have no significant impact on the results; the outcomes of an experiment prove to be rather resilient to marginal changes of core characteristics (as remarked earlier by Fiorina and Plott, 1978). This is our general argument, supported by a series of experiments on electoral systems and strategic voting.