Right here is a new and very important paper by Victor Stango and Jonathan Zinman, listed here are a few of the essential outcomes, noting that each paragraph is necessary:
Our first discovering is that biases are extra rule than exception. The median shopper displays 10 of 17 potential biases. Nobody displays all 17, however virtually everybody displays a number of biases; e.g., the fifth percentile is 6.
Our second discovering is that cross-consumer heterogeneity in biases is substantial. The usual deviation of the variety of biases exhibited is about 20% of its imply, and several other outcomes beneath counsel that this variance is economically significant and never considerably inflated by measurement error.
Our third discovering is that cross-consumer heterogeneity in biases is poorly defined by even a “kitchen sink” of different shopper traits, together with classical determination inputs, demographics, and measures of survey effort. Most strikingly, we discover extra bias variance inside classical sub-groups extensively thought to proxy for behavioral biases than throughout them. E.g., we discover extra bias variation with the highest-education group than throughout the highest- and lowest-education teams.
Our fourth discovering is that our 17 biases are positively correlated with one another within-consumer, particularly after accounting for measurement error following Gillen et al. (2019).1Across all biases, the common pairwise correlation is 0.13, and 18% have p-values < 0.001. Inside six theoretically-related teams of biases (present-biased discounting, inconsistent and/or dominated decisions, threat biases, overconfidence, math biases, and restricted consideration/reminiscence), the common pairwise correlation is 0.25 and 29% have p < 0.001.
Our fifth discovering is that there are additionally some necessary correlations between biases and classical inputs. Classical inputs and demographics might not clarify a lot of the variance in biases (per discovering #3), however a few of them are correlated with biases in patterns that align with prior work. Most notably, the common pairwise correlation between cognitive expertise and biases is -0.25. Cognitive expertise are strongly negatively correlated with most biases, however positively correlated with loss aversion and ambiguity aversion. Different classical inputs are comparatively weakly correlated with biases, aside from a couple of anticipated hyperlinks between endurance and current bias, threat aversion and aversion to uncertainty and losses, and threat aversion and math biases that may result in undervaluation of returns to risk-taking.
General not encouraging! However maybe a few of that can be what makes life extra significant, at a excessive price admittedly.