Exploring differences in empirical time preference rates for health: an application of meta-regression

Estimated time preference rates are extremely varied, with many rates being extremely high. Reviewing empirical studies without quantitative synthesis of their findings is largely unhelpful in determining how rates vary according to different factors. This study therefore explores the use of meta-regression techniques to combine available evidence to draw reliable conclusions about the factors influencing empirical time preference rates. Papers reporting empirically derived time preference rates related to health and health-care programmes were selected. Included were papers presenting all of: a mean time preference rate; information allowing derivation of standard errors; and one or more covariates. Appropriate data were derived from only eight of the 16 papers reporting empirical time preference rates. Meta-regression indicated that there were statistically significant relationships between mean time preference rates and: (a) delay period on a log scale; (b) whether the outcome question related to a gain or a loss. There were a number of limitations related to the use of meta-regression in this area, including difficulties in extracting appropriate data from the original studies, and the extent to which the original studies provide fully deliberated estimates of time preference.