Measuring natural source dependence.

Summary The consequences of most economic decisions are uncertain: they are conditional on events with unknown probabilities that decision makers evaluate based on their beliefs. Beyond consequences and beliefs, decision makers' preferences for the source of uncertainty generating events can affect decisions, entailing a source-dependence of preferences. However, there is, to date, no direct way to measure and interpret source dependence for natural sources, preventing researchers from comparing source preference across decision makers or across natural sources. This paper presents a simple and general method to measure source dependence by introducing a source-dependence function that maps the subjective probabilities of events generated by two sources. The method allows to characterize source dependence from a limited number of observations and is compatible with commonly used, choice-based data. We implement the method on four datasets and show that it consistently captures clear, albeit heterogeneous, patterns of source dependence between natural sources.