Estimating Individual Discount Rates in Denmark: A Field Experiment

Discount rates are often used in cost-benefit analysis. Whenever costs and benefits for a household or individual are spread over time, it is essential that one calculate present-value equivalents in order to undertake meaningful comparisons. In most cases welfare analysts use market rates as the basis for these present-value calculations. Sensitivity analysis often consists of varying the scalar discount rate up or down in relation to market interest rates. Since discount rates are a reflection of subjective time preferences, one would expect a priori that they could differ across different individuals. However, standard practice in intertemporal welfare analyses is to assume that those rates are (i) the same across households, and (ii) the same for all time horizons. We elicit individual discount rates from subjects in order to test these two hypotheses. The first hypothesis is that discount rates for a given time horizon do not differ with respect to sociodemographic characteristics that characterize households in our sample. The second hypothesis is that discount rates for a given individual do not differ across time horizons. We use survey questions with real monetary rewards to elicit individual discount rates and demonstrate the methodological complementarity between lab and field experiments. The survey questions are designed by Manbeth Coller and Williams (1999), who elicit nominal individual discount rates for university students using controlled laboratory experiments. We apply their experimental procedures, but employ subjects that are normally encountered in field surveys. Our experiments were carried out across Denmark for the Danish government, using a nationally representative sample of 268 people between 19 and 75 years of age. Our results indicate that nominal discount rates are constant over the one-year to threeyear horizons used in these experiments, and that discount rates vary significantly with respect to several sociodemographic variables. On the basis of these results one can assume constant discount rates for specific household types, but not the same rates across all households. In Section I we review the logic of our experimental design. Section II explains the field * Harrison: Department of Economics, Moore School of Business, University of South Carolina, Columbia, SC (e-mail: [email protected]); Lau: Centre for Economic and Business Research, Langelinie Alle 17, 2100 Copenhagen, Denmark (e-mail: [email protected]); Williams: National Center for Environmental Economics, U.S. Environmental Protection Agency, 1200 Pennsylvania Ave. NW (MC 1809T), Washington, DC 20460 (e-mail: williams. [email protected]). We are grateful to the Danish Ministry of Business and Industry for funding this study, to Maribeth Coller, Ben Heijdra, and two referees for comments. None of our employers or overlapping generations are responsible for our conclusions. 1 We elicit discount rates for individuals. To the extent that the characteristics of individuals are used to define “representative households,” we can refer to the individual and the household interchangeably. However, we remain agnostic concerning the way in which the individual discount rates of individual household members are aggregated into one household discount rate, akin to a social discount rate for the household as a small society. 2 Coller and Williams (1999) explain how their design relates to findings in the extant experimental literature. We review this discussion in our working paper, available at http://dmsweb.badm.sc.edu/glenn/idr/dkidr.htm . This web page also contains links to all experimental instructions, data, and software to replicate our results. For the convenience of Danish-challenged readers we also provide on this web site an English translation of all instructions and questionnaires. 3 At the time of the experiments the inflation rate in Denmark was just under 2 percent per annum, and had been steady for several years. It rose to 3 percent per annum by the end of the longest horizon used in our experiments. The realized rates of inflation, taking the front-end delay into account, were 0.3 percent, 1.2 percent, 3.2 percent, and 6.1 percent for the 6-, 12-, 24-, and 36-month horizons, respectively.