The estimation of marginal time preference in a UK-wide sample (TEMPUS) project: a review

John Alexander Cairns, Marjon van der Pol

Research output: Contribution to journalArticle

32 Citations (Scopus)


Generally, any individual would prefer to receive a benefit today rather than in the future and to incur a cost later rather than sooner. Economists call these time preferences. Such preferences are relevant in two ways in the context of health care. First, how individuals view future costs and benefits influences health-affecting behaviour like smoking, exercising and following dietary restrictions. Information on peoples' time preferences could help us to understand health-affecting behaviour and therefore be valuable with respect to the design of policies for the promotion of health. Second, because timing matters, and because different interventions have different time profiles of costs and benefits, methods are required to take into account the timing of costs and benefits when undertaking economic evaluation of healthcare interventions. This is achieved by discounting future costs and benefits to present values by attaching smaller weights to future events the further into the future they occur.

To derive implied discount rates for future health benefits for a sample of the general public in the UK.
To establish whether individual inter-temporal preferences with respect to their own health differ from those with respect to the health of others.
To investigate the effect of different ways of asking questions on apparent inter-temporal preferences (specifically closed-ended and open-ended methods are compared).
To establish whether individuals value future health benefits in line with the traditional discounted utility model and to investigate, in addition, how well the hyperbolic discounting models explain individual responses.
Stated preference techniques comprising a series of health-related choices were used to elicit the time preferences of a random sample of adults. Two methods were used: an open-ended method and a discrete choice experiment (closed-ended method). Preferences were elicited for non-fatal changes in own health and others' health. Four different postal questionnaires were sent to a random sample of 5120 adults in England, Scotland and Wales. The data were analysed using a number of forms of regression analysis.

Results and conclusions
The median implied discount rates were 6.1% for own health and 6.2% for others' health using the open-ended method and, in the discrete choice experiment, 5.0%, 4.6%, 3.8% (5-, 8- and 13-year delay, respectively) for own health and 6.4%, 5.7%, 3.8% for others' health.

The results suggest that the implied discount rates for own and others' health are broadly similar. There are some differences but the similarities are much more striking, certainly in the case of the open-ended method.

The implied discount rates and the distribution of the implied discount were very similar for the open-ended method and the discrete choice experiment. The discrete choice experiment had a higher response rate and respondents considered that the discrete choices questions were easier to answer.

The results provide evidence against the discounted utility model. The key axiom of the discounted utility model, stationarity, was violated. The alternative, the hyperbolic discounting models, fitted the data better than the discounted utility model.

The implied discount rates elicited in this study should not be over-emed because of the unrepresentativeness of the study sample. However, it is notable how close the estimated median rates are to the rates advocated for use in economic evaluation in a range of countries (for example, 3% in the USA, 5% in Australia and Canada). The estimated implied discount rates in this study fall comfortably within the range of estimates from previous empirical studies.

Research recommendations
A single, albeit multifaceted, project such as TEMPUS adds significantly to our understanding but cannot by itself resolve the outstanding research issues, particularly as this is the first study in which a number of these issues have been addressed systematically. Three areas should be highlighted.

Continued refinement of the methods of eliciting time preferences is required. Relevant topics include the use of self-completed questionnaires versus interviews (face-to-face and telephonic) and the presence and impact of framing effects.
Further research is required on alternative models of time preference, in particular, models which allow for decreasing timing aversion. Also, the implications of using alternative models for policy making need to be investigated.
There is considerable scope for research to investigate the role played by time preference in explaining health-affecting behaviour. To what extent are individuals willing to incur short-term costs in order to secure longer-term benefits Ð for example, in the successful control of blood sugar levels by patients with diabetes or by participation in screening programmes?
Original languageEnglish
JournalHealth Technology Assessment
Issue number1
Publication statusPublished - Jan 2000

Fingerprint Dive into the research topics of 'The estimation of marginal time preference in a UK-wide sample (TEMPUS) project: a review'. Together they form a unique fingerprint.

  • Cite this