Can hypothetical measures of time preference predict actual and incentivised behaviour? Evidence from Senegal.

Abstract

Time preferences are an important determinant of decision-making and are widely measured through hypothetical survey questions. However, the extent to which they offer a good representation of time discounting remains largely unexplored. This paper estimates time preference parameters using a commonly-applied hypothetical elicitation method. We explore whether our estimated parameters correlate with actual and incentivized behaviours related to time preferences. First, we consider the correlation between our hypothetical measures and the result of an incentivised experiment using the unique reference numbers of banknotes as a means of determining an individual’s willingness to save money. Individuals are given a banknote and informed that if they chose to retain this specific note for a randomly assigned period of time (2, 7 or 14 days) they will receive a second banknote, in effect doubling their initial endowment. Second, we consider the correlation between hypothetical measures and an individual’s observable saving behaviour, including ownership of a savings account and participation in a Rotating Credit and Savings Association (ROSCA). Overall, our results show that hypothetically-derived time preference parameters are not significantly correlated with our measures of actual or incentivized behaviour. We explore the extent to which our results are due to limited power and find that a version of our results comparable to the relevant literature can detect effect sizes in line with similar studies. Furthermore, we recognise that our incentivized experiment will be a noisy reflection of time discounting and subject to confounding factors, such as the inherent fungibility of money. We provide ancillary evidence suggesting that our main results remain robust to these considerations and others.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 8 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.