Suppose you are an executive in charge of marketing decisions for your firm. Your R&D people have developed a new product, and you need to decide what price to charge. Do you feel confident predicting what people will pay? Shane Frederick, a professor at Yale School of Management, is confident you’ll overestimate how much the typical person will pay – and overestimate it by a lot.
Professor Frederick conducted a variety of studies where he asked participants to engage in a simple exercise: he selected some object (e.g., premium chocolates, DVDs, smoked salmon, fossilized ammonites, magnetic sculptures, scented oils, a gift certificate for Pizza Hut) and asked respondents to predict what a specified person (or what a typical person) would pay. In every case, estimates were too high – much too high. He found the bias for every good examined, and the overestimates averaged around 40%!
His research also reveals an extraordinary lack of appreciation for where in the distribution of valuations one lies. It was not uncommon for the person with the highest value in a group to predict that the average value was higher still. There are interesting interactions between this “false consensus effect” and the “X effect” (Frederick’s agnostic moniker for the consistent pattern of overestimates).
First, for people who have comparatively high valuations, both the false consensus effect and the X effect work to inflate estimates. Second, for people with comparatively low valuations, the two effects work to cancel each other. Thus, he finds that the most accurate estimates are rendered by those with atypically low valuations: those around the 25th percentile in the distribution. In some situations, he finds that the X effect dominates the false consensus effect; that even those who would pay nothing for a good tend to overestimate the average valuation.
Unlike other biases, mere knowledge of the X effect reduces it. As part of a seminar on decision making, Professor Frederick conducted such a study with MBA students and revealed the results at the end. Several months later, as part of a lecture on pricing in his Consumer Behavior class some of the participants encountered the task again, though with a different set of products. The bias persisted, but was much smaller. A single exposure over time was sufficient to de-bias them.