Social influence and position effects

Abstract

A wide range of personal choices rely on the opinions or ratings of other individuals. This information has recently become a convenient way of simplifying the decision process. For instance, in online purchases of products and services, the possible choices or alternatives are often characterized by their position in a certain presentation order (or list) and their popularity, derived from an aggregate signal of the behavior of others. We have performed a laboratory experiment to quantify and compare popularity (or social influence) and position effects in a stylized setting of homogeneous preferences, with a small number of alternatives but considerable time constraints. Our design allows for the distinction between two phases in the decision process: (1) how agents search (i.e., not only which alternatives are analyzed but also in which order) and (2) how they ultimately choose. We find that in this process there are significant popularity and position effects. Position effects are stronger than social influence effects for predicting the searching behavior, however, social influence determines to a larger extent the actual choice. The reason is that social influence generates a double effect; it directly affects the final choice (independently on what alternative has been searched more thoroughly) and indirectly alters choice through the searching behavior which, in turn, is also affected by popularity. A novelty of our approach is that we account for personal traits and provide an individual analysis of sensitivity to both social influence and position effects. Surprisingly, we find that overconfident individuals are more influenceable, whereas other personal characteristics (e.g., gender and risk aversion) do not play a significant role in this context.