The predictive accuracy of intertemporal‐choice models

Abstract

How do people choose between a smaller reward available sooner and a larger reward available later? Past research has evaluated models of intertemporal choice by measuring goodness of fit or identifying which decision-making anomalies they can accommodate. An alternative criterion for model quality, which is partly antithetical to these standard criteria, is predictive accuracy. We used cross-validation to examine how well 10 models of intertemporal choice could predict behaviour in a 100-trial binary-decision task. Many models achieved the apparent ceiling of 85% accuracy, even with smaller training sets. When noise was added to the training set, however, a simple logistic-regression model we call the difference model performed particularly well. In many situations, between-model differences in predictive accuracy may be small, contrary to long-standing controversy over the modelling question in research on intertemporal choice, but the simplicity and robustness of the difference model recommend it to future use.

Time-Preference Tests Fail to Predict Behavior Related to Self-control

“…They are useful as reassurance that we did not choose particularly poor tests, to which subjects responded mostly randomly. Arfer and Luhmann ( 2015 ) likewise showed that laboratory tests of time preferences can predict each other very accurately. It follows that the lack of predictive accuracy for CVs cannot be attributed solely to noise in the measurement of preferences or temporal instability in preferences.…”

“…The goal of this study was not to evaluate any one of the many models of intertemporal choice that have been proposed (Doyle, 2013 ); our focus was on tests rather than models. We have previously found evidence that a wide variety of models can successfully predict responses in a pure intertemporal-choice task (Arfer and Luhmann, 2015 ), but in that study, we did not try to predict other kinds of behavior, and prediction was within rather than between subjects.…”

“…Predictive accuracy can be quantified with some of the same statistics as association, such as RMSE. However, association is optimistically biased as a measure of predictive accuracy, due to overfitting (Wasserman, 2004 , Theorem 13.15; see also Arfer and Luhmann, 2015 ). Measuring predictive accuracy while avoiding this problem requires additional steps, such as cross-validation (Hastie et al, 2009 , p. 228), which splits the data into parts for separate model-fitting and model evaluation.…”

Does excessive use of smartphones and apps make us more impulsive? An approach from behavioural economics

“…Consumers face daily multiple-choice situations involving different rewards with different discounting intervals. In their simplest form, these situations entail choosing between an option that provides a small but sooner reward (SSR) and another that offers a larger but more delayed option (larger later reward or LLR) (Arfer and Luhmann, 2015). These situations can be routine decisions such as choosing between eating an ice cream (SSR) or staying on a diet (LLR), saving to pay for college tuition (LLR) or going out for an expensive meal with friends (SSR), taking a day off at the club or studying for an exam the next day.…”

Flexible Utility Function Approximation via Cubic Bezier Splines

“…However, achieving this predictive power requires considerably more data. In a dataset of about 100 choices, Arfer and Luhmann (2015) found that support vector machines, random forests, and k-nearest neighbor clustering algorithms do not have higher predictive capabilities than parametric models. More importantly, these non-parametric methods are agnostic, 'black-box' approaches that do not readily yield interpretable insights, and therefore have rarely been used in studies seeking to advance theories of the decision-making processes involved in ITC and RC.…”