Aggregation and the External Validity of Time Preference Measures
with Daniel Navarro-Martínez
Presented at SEET 2026; FUR Conference 2026; and the BSE Summer Forum Workshop on Advancing Methods in Behavioral and Experimental Economics, 2026.
Measures of time preferences are often only weakly associated to real-world intertemporal behaviors. We investigate the extent to which the external validity of these measures can be increased by combining observations to reduce measurement error. Using a novel day reconstruction approach, we tracked participants’ real-life intertemporal decisions over seven consecutive days and measured time preferences twice using six different measures. We show that both combining daily behaviors and combining preference measurements substantially increases lab–field correlations. This indicates that time preference measures can capture meaningful intertemporal preferences that predict real-world behaviors, but only once measurement error is sufficiently reduced.
External Validity of Preference Measures: The Role of Measurement Error
with Daniel Navarro-Martínez, Florian H. Schneider, and Xinghua Wang
Manuscript in preparation
Presented at ESA 2026
Preference measures often exhibit weak correlations with related real-world behaviors, raising concerns about their external validity. We investigate whether aggregating observations to reduce measurement error strengthens these correlations. Using data from 37 papers published in the top five economics journals, we examine 131 study cases in which preference measures and related field behaviors are observed for the same participants, with multiple observations available on at least one side. Within each case, we compare lab–field correlations based on individual observations with those obtained after aggregating available measures and behaviors. We find that aggregation increases the average absolute correlation from 0.075 to 0.141, an increase of 88%, with substantially larger gains in some cases. Combining observations on either side contributes to these improvements, which hold across social, risk, and time preferences. Our findings highlight the importance of aggregating related field behaviors to avoid underestimating the external validity of elicited preference measures. They also show that combining different elicitation tasks can yield more externally valid preference measures.