A meta-analysis of extremeness aversion

Abstract

Using a meta-analysis of 142 experimental observations, this work examines the influence of different research design and outcome measures on extremeness aversion (i.e., the propensity to avoid extreme options in choice situations). The results indicate that extremeness aversion is a robust phenomenon: middle options are significantly more often selected than other options. However, the strength of this behavioral effect exhibits substantial variation (up to three times the average magnitude) across methodological decisions: employing price-quality tradeoffs, nondurable categories, or binary-trinary choice-set comparisons tend to reduce the probability of extremeness aversion among consumers, whereas using a larger number of tradeoff dimensions, non-numeric attributes, high-quality extensions, or utilitarian products increase its likelihood. Because extremeness aversion has been assessed using three different measurement paradigms (absolute-share changes, relative-share shifts, and middle-option proportions), we discuss their characteristics and investigate their degree of agreement. We find that the three measures can lead to rather different effect magnitudes and even contradictory conclusions about the effect of moderators.