Measuring the Mere Measurement Effect in Non-Experimental Field Settings
|The mere measurement effect is a well-documented effect in laboratory studies and has recently been documented in several field surveys. In this paper, we identify two major obstacles to measuring the mere measurement effect in a non-experimental field, field setting. These are: Who is asked to participate in a survey? Who agrees to respond to a survey? Firms often target surveys at particular consumers and this induces an effect we refer to as targeting bias. In addition, consumers who respond to a survey may not be representative of the overall population and this may create response bias.In this paper, we present a methodology for measuring the mere measurement effect while simultaneously controlling for both targeting and response bias.
We illustrate our methodology on a dataset from a direct marketing firm that regularly surveys its customers. The results show that failure to control for targeting and response bias leads to substantial overestimates of the impact of survey response on purchase frequency. Without controlling for these biases, our model predicts a mere measurement effect of 9% on purchase frequency and 4% on spending. After controlling for these biases, these effects decrease to -2% and 0%, respectively. We find evidence of a small, positive mere measurement effect among customers with greater tenure with the company.
|Dr. S. Puntoni|