Measuring Reciprocity: Do Survey and Experimental Data Correlate?

This paper investigates whether the answers to commonly used attitudinal survey questions are significantly correlated with actual behavior and checks whether survey questions do exist that provide satisfactory measures of willingness to reciprocate. To address this issue I analyze the data from a questionnaire which subjects filled in after participating in a gift exchange experiment. I proceed in two steps. First, I perform a factor analysis to investigate whether different questions can be grouped into one factor. I then check whether the factors extracted correlate with the type of worker, as measured by the estimated slope parameter of the individual wage-effort schedule, and the employer’s wage choice. I extract four factors that correspond exactly with the a priori classification of the survey questions, i.e. reciprocal attitudes, trusting attitudes, lending behavior, and tipping attitudes. The results suggest that survey questions that ask respondents directly about their reciprocal attitude are the best predictor of actual reciprocal attitudes. Additionally, questions about trust appear to be a good predictor of employers’ wage offers.