Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, ECONOMICS AND FINANCE (economics.oxfordre.com). (c) Oxford University Press USA, 2018. All Rights Reserved. Personal use only; commercial use is strictly prohibited (for details see Privacy Policy and Legal Notice).

date: 14 November 2018

Time Preferences for Health

Summary and Keywords

The interest in eliciting time preferences for health has increased rapidly since the early 1990s. It has two main sources: a concern over the appropriate methods for taking timing into account in economics evaluations, and a desire to obtain a better understanding of individual health and healthcare behaviors. The literature on empirical time preferences for health has developed innovative elicitation methods in response to specific challenges that are due to the special nature of health. The health domain has also shown a willingness to explore a wider range of underlying models compared to the monetary domain. Consideration of time preferences for health raises a number of questions. Are time preferences for health similar to those for money? What are the additional challenges when measuring time preferences for health? How do individuals in time preference for health experiments make decisions? Is it possible or necessary to incentivize time preference for health experiments?

Keywords: time preference, health, stated preferences, elicitation methods, hyperbolic discounting

Introduction

Many decisions have consequences that extend over time. These intertemporal choices require individuals to make value comparisons between outcomes that occur at different points in time. Time preferences (i.e., preference for immediate utility over delayed utility) are key parameters in economic models of intertemporal choices. The first economists to examine intertemporal preferences in detail were Rae (1905), Senior (1836), and Jevons (1911). The interest in individuals’ time preferences increased after the discounted utility model was established in the 1930s and economists and psychologists started to elicit empirical estimates of individuals’ time preferences in the early 1980s. This empirical literature has since boomed. The interest in time preferences for health has been particularly strong. In part, this may be due to health economists’ interest in individual preferences. The ongoing discussion around the appropriate discount rate for health in the economic evaluation of health(care) intervention has also clearly stimulated an interest in time preferences for health, as policymakers seek to compare programs. Fuchs (1986; see also Fuchs & Zeckhauser, 1987) was among the first to connect time preferences (for money) to health behavior, and he found relatively weak associations. Since that initial study, there have been significant developments in the literature, both in the number of studies and the methods used.

The broader time—preference literature has not reached consensus on either the theoretical framework or the empirical methodology (Cohen, Ericson, Laibson, & White, 2016). However, as Cohen et al. (2016) note, “The good news is that conceptual discord is invigorating the intertemporal choice literature and opening the way for new ideas that we hope will resolve the multiple conceptual and methodological contradictions that have emerged.” This article focuses on the empirical time preferences for health literature, which has been innovative in terms of elicitation methods and which faces some specific challenges due to the special nature of health. It discusses some key issues and shows that many research questions remain. There is a particular focus on specific health issues, but general issues affecting both the health and monetary domains are also discussed. The article first explores why time preferences for health are of interest. It then focuses on the domain specificity of time preferences, demonstrating the need for eliciting time preferences for health. The underlying model of time preferences has received substantial interest in both the monetary and health domains, the health domain exploring a wider range of models. The article presents the wide range of elicitation methods in health; provides an overview of how these have been applied; and discusses the challenges that health poses. These include anticipation, the difficulty of using real incentives, and completeness of preferences. This is followed by a discussion of the most appropriate “type” of time preference when the interest is in predicting health behavior. Finally, future directions are identified.

Why Are Time Preferences for Health of Interest?

The interest in eliciting time preferences for health has increased rapidly since the early 1990s. Initially, the main source of interest was a concern with the appropriate methods for taking timing into account in economics evaluations of health(care) interventions. Discounting practices often play a central role in determining the relative cost-effectiveness of different interventions because the discounting assumptions reveal the relative weight put on future outcomes versus near-term costs. Debate concerning the choice of social discount rate in evaluating healthcare is ongoing (see, e.g., Jit & Mibei, 2015; Severens & Milne, 2004). One normative view is that the rate that best represents people’s preferences should be used to discount future health benefits. Whether this should be the social discount rate of the general population, paying group, affected group (patients), or some other subset is open for debate.

The estimation of health utilities to calculate quality-adjusted life years (QALYs) for use in the economic evaluation of healthcare interventions has also generated interest in time preferences for health. The most widely used method for estimating quality of life is the time trade-off (TTO), which requires individuals in hypothetical valuation exercises to give up time in good health to gain improvements in quality of life. These trade-offs were originally elicited through closed-ended binary choices, but there has been a move toward the use of discrete choice experiments (Bansback, Brazier, Tsuchiya, & Anis, 2012). The TTO method generally assumes a constant marginal utility of life years. The existence of individual time preferences violates this assumption (Dolan & Jones-Lee, 1997). This potential bias has stimulated further interest in eliciting time preferences for health and exploring the impact of individuals’ time preferences on TTO values.

Another source of interest in time preference for health is a desire to obtain a better understanding of individual health investments. Health investments, such as quitting smoking, engaging in physical activity, or having a cancer screening, are generally associated with short-term costs and longer term health benefits. An individual’s time preference is therefore likely to influence the level of investment (Fuchs, 1986). As several systematic reviews have demonstrated (Amlung, Vedelago, Acker, Balodis, & MacKillop, 2017; Barlow, McKee, Reeves, Galea, & Stuckler, 2017; Barlow, Reeves, McKee, Galea, & Stuckler, 2016) the interest in the relationship between time preferences and health behaviors has boomed since the 2000s. However, note that the majority of these studies elicit time preferences for monetary outcomes rather than for health outcomes. Domain specificity is an active area of research. Although a single discount rate is a convenient assumption, the specific differences between health outcomes and monetary outcomes means that the transfer could be inappropriate.

Related to individual and patient behavior is the interest in the relationship between time preference and clinical decision-making by health professionals. Many treatment decisions involve outcomes over time, and treatment recommendations may therefore depend on the health professionals’ time preferences. Insights into their time preferences may therefore provide insights into practice variations.

What Is the Appropriate Underlying Model?

For many years, the standard intertemporal model was the discounted utility model. Although the discounted utility model was introduced with caution, it was rapidly established as the framework for analyzing intertemporal decisions. The key axiom of the model is stationarity, which is a condition of dynamic consistency. Stationarity refers to the assumption that the trade-off between two periods (denoted by t and s) depends only on the absolute temporal distance (st). That is, if an individual is indifferent between outcomes (X and Y) in two periods (t and s), then she or he remains indifferent if t and s are incremented by a given constant amount (σ‎):

(X,t)(Y,s)and(X,t+σ)(Y,s+σ).
(1)

For example, if an individual decides to give up smoking tomorrow, then she will follow through with this plan when the next day arrives because the temporal distance between the costs of giving up and the future benefits has not changed with the passage of time.

The intertemporal utility function, which expresses satisfaction or utility as a function of outcomes at different points in time, can be expressed as

D(δ,t)=1(1+δ)t
(2)

where δ‎ is the time preference rate and D is the discount factor.

Whether individuals’ preferences satisfy the axiom of stationarity has been extensively tested (Frederick et al., 2002). A growing body of empirical literature shows that individuals’ time preference rates decrease as a function of delay (Cairns & van der Pol, 1997; Camerer, 1998; Frederick, Loewenstein, & O’Donoghue, 2002; for an example in the health domain see Cairns & van der Pol, 1997). This is referred to as decreasing timing aversion. It can lead to time inconsistency because an individual’s optimal plan from today’s perspective will not necessarily be the optimal plan from tomorrow’s perspective. Unless a pre-commitment strategy is used, individuals will not necessarily stick to a plan with long-term consequences. For example, the smoker who decides to give up cigarettes tomorrow will not stop smoking when the next day arrives. The distance in time between the costs and the future benefits of giving them up has become more important with the passage of time. Decreasing timing aversion in the first interval of time is also referred to as the immediacy effect or present bias and reflects that individuals may attach enhanced significance to outcomes that occur in the present.

This has led to the development of the quasi-hyperbolic model (Laibson, 1997), which can account for present bias

D(β,δ,t)={1ift=0β*1(1+δ)tift>0
(3)

where β‎ is a measure of present bias. When β‎ is less than one, the individual exhibits present bias. The quasi-hyperbolic model is used extensively in economics, although the extent of quasi-hyperbolic discounting has been questioned (Andersen, Harrison, Lau, & Rutstrom, 2014). Quasi-hyperbolic discounting has also been linked to uncertainty in that the immediacy effect may be the result of the certainty of present outcomes; whereas future outcomes are always uncertain (see, e.g., Andersen, Girolamo, Harrison, & Lau, 2014; Dasgupta & Maskin, 2005). This is interesting in the health context, where it can be argued that future health is associated with higher levels of uncertainty compared to money because the individual can only influence the probability of being in good health in the future.

The quasi-hyperbolic model incorporates the immediacy effect but assumes that further delays are discounted at a constant rate. However, further delays may also be discounted at a decreasing rate, and hyperbolic models that imply a monotonically decreasing rate of discount may be more appropriate than the quasi-hyperbolic model. For example, van der Pol and Cairns (2011) showed that individuals discounted both the initial delays between health outcomes and further delays between health outcomes at a decreasing rate. It is interesting to note that the most popular hyperbolic model in psychology, the Mazur model (Mazur, 1987), implies a monotonically decreasing rate. This model has not been applied to a great extent in economics, but the health literature has shown a greater interest in a wider range of models (see, e.g., Bleichrodt, Gao, & Rohde, 2016; Cairns & van der Pol, 1997).

It should be noted that a range of anomalies have been identified that are not accounted for in the quasi-hyperbolic model. These include the magnitude effect (larger outcomes are discounted at a lower rate than small outcomes), sign effect (gains are discounted at a higher rate than losses), and framing effects (including the delay-speedup asymmetry, where the amount required to compensate for delaying the receipt of a reward by a given interval is greater than the amount subjects are willing to pay to speed up consumption by the same interval (Cairns, 1992a; Chapman & Elstein, 1995; MacKeigan, 1993)). Although it possible to develop a model that accommodates several anomalies, there is clearly a trade-off between descriptive accuracy and tractability.

The question arises whether the best-fitting model for health is similar to that for money. Attema (2012) argues that individuals may be less likely to deviate from the discounted utility model with money compared to health because time preferences for money are influenced by market interest rates. Only a few studies have directly compared the descriptive validity of different discounting models across money and health. Bleichrodt et al. (2016) found that individuals deviated more from the constant discounting of the discounted utility model for health than for money. The hyperbolic models that imply a monotonically decreasing rate was the best-fitting model for both money and health. In contrast, Galizzi, Miraldo, Stavropoulou, and van der Pol (2016) show that individuals are more decreasing averse in money compared to health.

Are Time Preferences for Health Different?

Standard economics assumes that individuals exhibit a single time preference that governs their intertemporal choices in all contexts. If this assumption holds, then time preference rates for health and for money are the same and there is therefore no need to specifically consider time preferences for health. However, one of the key differences between money and health in the context of time preference is that health can only be substituted across time periods to a very limited extent. An individual cannot effectively determine his or her state of health. Heredity, environmental factors, and fate all influence an individual’s health state. This means that the individual can only influence the probability of being in good health in the future. Health is also nontradable because it cannot be sold to a third party.1 Although this is, strictly speaking, true, an individual has some opportunities to trade health “with him/herself” (Zweifel & Breyer, 1997). To the extent that exercising “now” is painful, individuals trade the pain now for the benefits of exercise that occur in the future. Health is traded against other valuable commodities, such as time. For example, Cawley (2015) finds weak evidence associating obesity with the time-cost of acquiring food. Given the limited tradability, we cannot assume time preferences for health to be the same as time preferences for money. Assuming a perfect capital market in standard economics, the individual’s time preference rate is equal to the market interest rate: anyone with an innate time preference that is different from the market rate could increase utility by lending or borrowing. Health is not so easily traded, so it can no longer be assumed that the time preference rate for health is equal to the market interest rate.

The domain specificity of both health and money has received substantial empirical interest. This literature shows that the mean implied discount rates for health and for wealth tend to be different from each other and that the differences are statistically significant (see, e.g., Cairns, 1992a; Chapman, 1996a, 1996b; Chapman & Elstein, 1995; Chapman, Nelson, & Hier, 1999; Cropper, Aydede, & Portney, 1994; Khwaja, Silverman, & Sloan, 2007; Lazaro, Barberan, & Rubio, 2001). However, the evidence is mixed with respect to the directions in which they differ. Higher correlations are generally found within domains compared to across domains (see, e.g., Khwaja et al., 2007).

The empirical literature has generated several hypotheses why estimates of time preferences for health and for money may differ. First, it was hypothesized that differences arise because some anomalies of the discounted utility model only affect intertemporal preferences for health, and not for wealth, or vice versa. This hypothesis was rejected by Chapman and Elstein (1995), who showed that two common anomalies of the discounted utility model, the magnitude effect and the common-difference effect, affected intertemporal preferences for both health and wealth. Chapman (1996b) also showed that gain/loss asymmetry affects intertemporal preferences for both health and wealth. Second, it was hypothesized that differences in intertemporal preferences for health and wealth are caused by differences in the utility functions for health and wealth. Chapman (1996b) estimated separate utility functions for health and wealth for each respondent and found that domain dependence is not caused by differences in the utility functions.

Third, Chapman et al. (1999) tested the hypothesis that the differences were the result of differences in familiarity with the wealth and health scenarios. Two patient groups (migraine patients and bowel-disease patients) were presented with three intertemporal choices. One concerned a one-time cash prize; one concerned medication that would reduce the number of migraines per month over the next five years; and one concerned medication that would reduce the number of days of annual hospitalization due to bowel disease. The period of delay was either one month or six months. The mean implied rates were higher for health than for wealth. However, the degree of familiarity did not systematically influence the implied rates of discount. Also, the implied rates for wealth were not correlated with the rates for familiar ill health. Fourth, the framing of the intertemporal choices may cause differences between health and wealth to arise. In previous studies, the units that outcomes were expressed in differed for health and wealth. In the case of wealth, the outcomes occurred at one single point in time, whereas health was expressed in terms of duration. Chapman et al. (1999) formulated the intertemporal choices concerning wealth similarly to those for health. The monetary outcome concerned receiving a tax rebate over the next five years that reduced annual housing expenses either immediately or after a delay of one, three, or six months. Reformulating the intertemporal choices concerning wealth did not change the results; that is, the mean implied rates of discount were still higher for health than for wealth.

Given the mixed evidence, uncertainty remains over whether and why time preferences for health differ from time preferences for money. A better understanding of the domain specificity is important because this will inform discounting practices in economic evaluations (e.g., should health benefits be discounted at a lower rate than money benefits?) and indicate whether choice of outcome is important if we want to improve our understanding of the relationship between time preference and health and clinical behavior.

Time Preference for What Type of Health Outcome?

When discussing time preferences for health, it is important to consider the type of health outcome and who is experiencing the outcome (self or other). Health is a multidimensional concept, and it incorporates both quantity of life (life years) and quality of life. If the assumption in standard economics that individuals exhibit a single time preference that governs intertemporal choice in all contexts holds, then an individual’s rate of time preference would be the same for quantity of life and quality of life (and, indeed, for monetary outcomes), and the choice of outcome is less relevant. As noted, time preferences have been shown to be domain specific. There is also evidence that time preferences may vary in the health domain (see, e.g., Olsen, 1993).

The discussion around the appropriate discounting rate in the economic evaluation of health(care) interventions has also generated an interest in individual time preferences for health versus social time preferences for health. Economic evaluations inform societal decision-making, and discounting practices usually apply the social time preference rate to discount future costs and benefits to present values. If it is assumed that the social welfare function is the sum of the utilities of the individuals in the society, the social time preference rate is equal to an aggregate measure of individuals’ time preference rates. However, it has been argued that individuals’ preferences for individual decisions may differ from those they have for social decisions in that they are willing to save more in the social context (Krahn & Gafni, 1993). It could be that individuals do not believe that a group should bear the same cost for delaying ill health that they as individuals would bear for themselves. Individuals’ time preference rates may also be different because they are affected by such factors as uncertainty, mostly about the time of death. The discussion around the social time preference rate for health has generated an interest in empirical studies that compare individual and social time preferences for health (see, e.g., Cairns & van der Pol, 1999; Gyrd-Hansen, 2002; Lazaro et al., 2001). Health time preferences were the same for individual and social contexts in Cairns and van der Pol (1999) and for lives saved in Gyrd-Hansen (2002). Individual and social time preferences were different for QoL in Gyrd-Hansen and for lives saved in Lazaro et al. (2001). This is an issue that has received only limited interest in the monetary domain.

How Are Time Preferences for Health Measured?

Both economists and psychologists started to elicit time preferences in the early 1980s, initially focusing on testing the assumptions of the discounted utility model, in particular, the axiom of stationarity. This early literature also focused on eliciting time preferences for health, and the first study was published by Lipscomb (1989). A wide range of methods have been used to elicit time preferences for health, and it can be argued that the time preference for health literature has explored a wider range of elicitation methods compared to the monetary domain. The reason may be that health economists tend to use stated preference methods to a greater extent because of the special nature of health.

There is a recognition in health economics that the special nature of health and healthcare results in many fewer opportunities to obtain valuations from observed behavior. This is especially so in the case of time preferences for health. Individuals can invest in their future health by adopting a healthy lifestyle now, or they can invest in health insurance. However, these opportunities to trade are limited, especially compared to the opportunities to trade wealth across time. Another limitation of using revealed preference methods is that the estimation of time preference rates is relatively indirect and quite complicated. This results partly from the difficulty of using data that were collected primarily for some other purpose and the many more factors out of the researcher’s control (compared with an experimental approach). Many confounding factors are often present. The literature has therefore relied on the use of stated preference methods to elicit time preferences for health.

A range of open- and closed-ended methods are used to elicit time preferences for health, including matching tasks and variations on choice (such as single binary choice, multiple price lists, and discrete choice experiments). Common to all these methods is the attempt to find an indifference point for the respondent. Using this indifference (whether in timing, size of outcome, or probability of outcome) makes it possible to identify the discount rate under assumptions that vary by method. What follows is an explanation of the different methods using the common design of trading off days of ill health over two points in time (days of ill health earlier or later). The underlying discounting model can be identified by varying the timing of the outcomes.

In an open-ended question, the individual is asked to return a value. An example of an open-ended question is:

This question identifies the individual’s indifference point between two durations of ill health occurring at two different points in time. A choice has to be made about which value the subject is asked to return. To make the question slightly easier to answer, a visual aid containing a range of possible responses can be provided to subjects. The advantage of an open-ended method is that it identifies indifference points for each individual, and it therefore produces relatively rich data based on fewer questions than a closed-ended method does. However, open-ended questions are more cognitively complex. Open-ended questions are used in the health domain but are rarely used in the monetary domain as they are thought to be incentive incompatible.2

In a closed-ended method, all the values are specified, and the individual is asked to indicate which scenario she or he prefers. This kind of question is also called binary choice, dichotomous choice, pairwise comparison, and discrete choice. The following is an example of a discrete choice:

If an individual prefers scenario B, his or her time preference rate is higher than the discount rate offered in the choice; and similarly, if the individual prefers scenario A, his or her time preference rate is lower than the rate offered. There are several closed-ended methods that differ mainly in terms of the number of choices they present the individual with. In the monetary domain in particular, the use of “multiple price lists,” which present individuals with a series of discrete choices, has become popular. The following is an example of a multiple price list where the respondent chooses when ill health will occur:

An individual with a low time preference rate would switch to option A relatively quickly, whereas an individual with a very high time preference rate would switch later. Closed-ended questions are cognitively less complex than open-ended ones. Their main disadvantage is that they identify only the interval within which the individual’s implied time preference rate falls. Moreover, only limited information is available for individuals who always choose option A or always choose option B. If an individual always chooses B, we only know that the implied time preference rate is larger than that offered in the choices. Closed-ended methods are also known to suffer from biases, such as starting-point bias, range bias, and midpoint bias (see, e.g., Andersen et al., 2014; Andersen, Harrison, Lau, & Rutstrom, 2008; Hermann & Musshoff, 2016, for framing effects within multiple price lists).

Other closed-ended methods are single discrete choice, discrete choice with follow-up, discrete choice with repeated follow-up, and discrete choice experiment. The single discrete choice method presents subjects with a single discrete choice only. The population’s underlying distribution of time preferences is identified by randomly assigning discrete choices implying a number of different discount rates, also called the “bid vector.” Since only limited information is available for each subject, it is only possible to estimate a time preference rate for the whole sample using regression analysis (see, e.g., Cameron, 1988, for details), and a relatively large sample is required. To increase statistical efficiency, a follow-up choice can be introduced. The discount rate offered in the follow-up choice depends on the answer to the initial choice. The discrete choice with repeated follow-up method is an extended version of the discrete-choice-with-follow-up method. Subjects are presented with additional follow-up choices until an indifference point is identified. This is also referred to as “titration” in the psychology literature. One concern regarding this method is error propagation. Subjects in a discrete choice experiment are presented with several discrete choices that imply a range of discount rates. The rates to be offered are identified using experimental design methods.

The long tradition in the measurement of health utility has led to the development of more indirect methods of eliciting time preferences for health. Instead of directly presenting individuals with trade-offs between periods of ill health over time, time preferences are elicited by obtaining utility scores for different individual health profiles that differ in terms of when the ill health occurs. The health utility measurement methods used are the time trade-off (TTO), visual analogue scale and standard gamble. Willingness to pay to avoid ill health at different points in time has also been used. The main disadvantage of the indirect methods is that they require a number of restrictive assumptions to identify the time preference rates.

To demonstrate the extent to which these different approaches have been applied in practice, Table 1 summarizes the empirical work on time preferences for health. The studies have been grouped into those using open-ended methods, those using closed-ended methods, and those using rating/pricing methods to elicit time preferences. The tables show the method and the sample used. There seems to be a move toward estimating individual discount rates rather than social discount rates. In the health domain, there remains a balance of choice and open-ended surveys over time; in the wider literature, especially in the money domain, there has been a stronger shift toward choice-type designs.

Table 1. Studies of Time Preferences for Health

Author (Year)

Method

Sample

Perspective

Outcome

Gains/Losses

Lipscomb (1989)

VAS

60 students

Social

QoL

Losses

Horowitz & Carson (1990)

Single DC

75 students

Social

Lives

Losses

Cropper, Aydede, & Portney (1991)

DC with follow-up

1600 general public

Social

Lives

Gains

J. Cairns (1992b)

Fully open-ended

29 students

Individual

QoL

Losses

MacKeigan (1993)

VAS

108 students & hospital volunteers

Individual

QoL

Both

Olsen (1993)

Fully open-ended

250 general public & 77 planners

Social

Lives & QoL

Gains

Redelmeier & Heller (1993)

SG & VAS

121 students & doctors

Individual

QoL

Losses

J. A. Cairns (1994)

Payment scale

223 general public

Social

Lives

Gains

Cropper et al. (1994)

DC with follow-up

3200 general public

Social

Lives

Gains

Olsen (1994)

TTO

90 students & 40 doctors

Individual

QoL

Gains

Chapman & Elstein (1995)

Fully open-ended

104 students

Individual

QoL

Gains

Dolan & Gudex (1995)

TTO

39 general public

Individual

QoL

Both

Chapman (1996b)

Fully open-ended

148 students

Individual

QoL

Losses

Johannesson & Johansson (1996)

Single DC

850 general public

Social

Lives

Gains

Cairns & van der Pol (1997; Cairns & van der Pol (1997)

Payment scale

473 general public

2 Individual/4 social

Lives

Gains

Johannesson & Johansson (1997b)

WTP

528 general public

Individual

Lives

Gains

Johannesson & Johansson (1997a)

WTP

2577 general public

Individual

Lives

Gains

Enemark, Lyttkens, Troeng, Weibull, & Ranstam (1998)

SG

25 surgeons

Social

Lives

Gains

Cairns & van der Pol (1999)

Payment scale

298 general public

Individual/social

QoL

Losses

Chapman & Coups (1999)

Payment scale

409 employees

Individual

QoL

Losses

Chapman et al. (1999)

DC repeated follow-up

79 patients & 77 students

Individual

QoL

Gains

van der Pol & Cairns (1999)

DC with follow-up

367 general public

Individual

QoL

Losses

Ganiats et al. (2000)

Single DC

169 patients

Individual

Lives & QoL

Both

Poulos & Whittington (2000)

Single DC

3127 general public

Social

Lives

Losses

Bleichrodt & Johannesson (2001)

DC experiment

172 students

Individual

QoL

Losses

Lazaro et al. (2001)

Payment scale

203 students

Individual

QoL

Gains

Lazaro et al. (2001)

Payment scale

203 students

Social

Lives

Gains

van der Pol & Cairns (2001)

DC experiment

787 general public

Individual/social

QoL days)

Losses

Cairns, van der Pol, & Lloyd (2002)

DC repeated follow-up

203 students

Individual

QoL

Losses

Gyrd-Hansen (2002)

Fully open-ended & SG

79 students

Individual

Years

Gains

Gyrd-Hansen (2002)

TTO, PTO

79 students

Social

Lives

Gains

Hojgard et al. (2002)

SG

66 doctors, 21 patients, 22 general public

Social

Life years

Gains

Lazaro, Barberan, & Rubio (2002)

Payment scale

427 general public

Individual

QoL

Gains

Lazaro et al. (2002)

Payment scale

427 general public

Social

QoL

Gains

Stavem, Kristiansen, & Olsen (2002)

TTO

59 patients

Individual

QoL

Gains

Robberstad (2005)

Open-ended

450 general public

Individual/social

QoL

Losses

Khwaja et al. (2007)

Open-ended

663 general public (oversample smokers)

Individual

QoL (days)

Gains

Hardisty & Weber (2009)

Choice

65 general public

Individual, Social

QoL

Both

Meerding, Bonsel, Brouwer, Stuifbergen, & Essink-Bot (2010)

Choice

173 healthcare, 34 policy makers

Social

Lives, QoL

Gains

Bobinac, Brouwer, & van Exel (2011)

Choice

656 general public

Social

QoL & Years

Gains

Attema, Bleichrodt, & Wakker (2012)

Choice

70 students

Individual

QoL (years)

Gains

Attema et al. (2012)

Open-ended (certainty equiv.)

70 students

Individual

QoL (years)

Gains

Attema & Versteegh (2013)

TTO

4,000 general public

Individual

QoL

Both

Attema & Brouwer (2013)

Choice

80 Students

Individual

QoL (years)

Gains

Attema & Brouwer (2013)

Choice

80 Students

Individual

QoL (years)

Losses

Attema & Brouwer (2013)

Open-ended

80 Students

Individual

QoL (years)

Gains

Attema & Brouwer (2013)

Open-ended

80 Students

Individual

QoL (years)

Losses

van der Pol, Walsh, & McCartney (2015)

DC repeated follow-up

3702 general public

Individual

QoL

Losses

Bleichrodt et al. (2016)

Choice with repeated follow-up

63 students

Individual

QoL

Gains

Galizzi et al. (2016)

MPL

300 patients, 67 doctors

Individual

QoL (days)

Gains

Note: VAS = visual analogue scale; DC = discrete choice; QoL = quality of life; SG = standard gamble; TTO = time trade-off; WTP = willingness to pay; PTO = ; MPL = multiple price list.

Most empirical studies have assumed a linear utility function when estimating a time preference rate. It has been recognized that any nonlinearity in the utility function may bias the time preference estimates (Frederick et al., 2002). Several alternative approaches have been developed to address this. One approach is to measure utility using risky choices. Andersen et al. (2008) proposed a joint estimation of time and risk preferences using multiple price lists, and this method has become popular in the monetary domain. However, this approach has been criticized because risk preferences are generally analyzed using expected utility, which is known to be frequently violated, and it is not clear that risky cardinal utility can be transferred to riskless intertemporal choice (Attema et al., 2016). Another approach is the convex time budget (CTB) method developed by Andreoni and Sprenger (2012). In a CTB design, the respondent has a set number of tokens to allocate between outcomes or tasks. Tokens are given values that vary by time. The allocation of tokens between periods is, then, the solution to a constrained optimization problem. The method can be applied to nonmonetary outcomes, as it was by Augenblick, Niederle, and Sprenger, (2015), but it has not yet been applied to health.

Health applications have been at the forefront of developing further alternative methods. Attema et al. (2012) developed an approach that makes it possible to estimate time preferences for health without requiring a knowledge of utility by treating outcomes as flow variables. They then adapted this approach to money (Attema, Bleichrodt, Gao, Huang, & Wakker, 2016). It is perhaps not surprising that this method, which originated in health as treating outcomes as flow variables, is more common in the literature for time preference for health (for example, days or months of ill health).

It could be argued that though methods such as the CBT and the direct method are more theoretically sound, they also increase the complexity of the questions. Very little is known about the decision-making process and levels of understanding of the individuals in time preference experiments. These issues have been explored in other stated preference methods, such as health utility measurement and discrete choice experiments, but they have not received much interest in the time preference literature. Despite the relatively large number of studies on time preference, we still know relatively little about the validity of these methods. The very wide range of estimates found in the literature suggests that framing is important.

What (Additional) Challenges Do Health Outcomes Pose?

There are several additional challenges when eliciting time preferences for health outcomes instead of monetary outcomes: the influence of anticipation, difficulty of using real incentives, and completeness of preferences.

Positive time preference is assumed universally. Olson and Bailey (1981) wrote that “the case for positive time preference is absolutely compelling” (p. 130). However, negative time preference is commonly found in studies eliciting time preference for health (van der Pol & Cairns, 2000). In this case, for example, individuals prefer to experience a spell of ill health sooner rather than later. It has been hypothesized that individuals may exhibit negative time preference because of the anticipation of future unpleasant consequences (dread; see Loewenstein, 1987; Loewenstein & Prelec, 1991). The choices may therefore reflect both the discounted value of the loss and dread (Harris, 2012). The majority of health studies have focused on trading off health losses, which is likely to introduce dread. However, little is known about the extent to which dread influences individuals’ time preferences. It is possible to frame the question in terms of health gains rather than health losses, but this means that respondents have to imagine being in a chronic health state and that they will experience a gain. Unless time preferences are elicited in a particular patient sample, this introduces cognitive complexity. Recent studies have used health states that are common and relatively easy to imagine, such as chronic back pain (Attema et al., 2012).

The use of real incentives in experiments has become standard practice. It is argued that incentives are essential for getting respondents to reveal their true preferences. The use of real incentives in time preference elicitation tasks is challenging even with monetary outcomes. For example, to ensure that incentives do not bias the choices between money amounts at different time points, it is crucial to hold the transaction costs and payment reliability constant (Cohen et al., 2016). In the case of health outcomes, there is of course a clear ethical dimension. Real incentives in time preference for health outcomes would mean reducing an individual’s health at some point in time. Even if ethics review boards were to deem this acceptable to, it is unlikely that individuals would be willing to participate in such studies. More recently, in the broader literature of health experiments, researchers have used incentives by converting health outcomes into monetary donations to a health charity (Brosig-Koch, Hennig-Schmidt, Kairies-Schwarz, & Wiesen, 2017; Hennig-Schmidt, Selten, & Wiesen, 2011). These have been used in experiments examining doctors’ decision-making. This is an option for time preference elicitation, though given their indirect nature, it is unclear how effective these incentives are. Respondents could just view the task as an opportunity to give money to a charity, biasing their results. This is important to note, as the even the evidence on the more direct incentives in monetary time preference elicitation tasks is relatively weak (Cohen et al., 2016).

Another issue to consider is that respondents are less likely to be familiar with trading-off health over time. Most individuals are used to trading money over time, such as taking out loans or investing, but not health. It is therefore possible they may have incomplete or imprecise time preferences for health and may “learn” their time preferences as they go through the experiments. They may also be more likely to use rules of thumb when faced with a decision task. The implication of this is that responses may be more liable to framing effects.

It should be noted that though eliciting time preferences for health outcomes involves additional issues, there are also some clear advantages compared to monetary outcomes. As Cohen et al. (2016) highlighted, in the monetary domain there are concerns about the influence of liquidity constraints and transaction costs in time preference experiments. These concerns do not apply in the health domain.

Predicting Health Behavior

The interest in time preferences for health has shifted from informing to discounting practices to improving our understanding of health behaviors since the early 2000s. This may have implications for what “type” of time preference we wish to elicit. There is an ongoing discussion about whether time preferences for health outcomes are better predictors of health behavior than time preferences for monetary outcomes are. If time preferences are domain specific, then time preferences for health should be better predictors of health behaviors, such as smoking, physical activity, and alcohol consumption. However, the review by Story, Vlaev, Seymour, Darzi, and Dolan (2014) highlights that time preferences for money are well correlated with health behavior, perhaps even more highly correlated than are time preferences for health. The question arises of whether this may be due to a higher level of measurement error in health time preferences. Although asking questions in the health domain might be expected to correlate more strongly with health behavior, the exact question in terms of framing (specific or generic health state, choice intervals and limits) and outcomes (size, timing) is either too precise for a generalizable result or potentially just as remote as a money-domain question. Will we find that preferences are as well-correlated to behavior with a more robust measure of time preference for health?

Related to this is that time preferences for health (and money) are usually elicited under the conditions of certainty. Given that the outcomes of health investments are highly uncertain, it is not clear that the trading involved in long-term behaviors, such as diet or exercise, can be meaningfully captured in current survey methods.

Future Directions

Interest in individuals’ time preferences for health is likely to remain high. To improve our understanding of the role time preference plays in health and healthcare behaviors, and health utility measurement to inform discounting practices, it is crucial to elicit time preferences using robust measures. While the literature on time preference for health has shown a willingness to explore a wide range of different stated preference methods, it can be argued that less attention has been given to testing the robustness of the different methods. Further studies exploring the validity of the different methods, including convergent validity and external validity, where possible, is warranted. This is also the case for the wider time preference literature. Health economists have extensive experience of stated preference methods due to the special nature of health and are therefore well placed to take this research forward.

Time preference questions are complex. Part of the problem is that most individuals are unused to thinking in the way required of them. A related difficulty is devising questions that are meaningful to the individual. This is particularly true for health. Little is known about individuals’ decision-making processes in time preference for health experiments and what factors might influence their responses. In particular, it is unclear whether negative time preferences for health are true preferences. There is a need for increased use of qualitative research to better understand individuals’ decision-making processes in time preference experiments. Domain specificity remains an important issue, in particular, with regard to the interest in understanding health and healthcare behavior. A shift toward gaining better understanding of why preferences may vary across the health and money domains and examining whether the underlying model varies across domains is warranted. Many economists consider the use of real incentives to be crucial. Further exploration of the feasibility and impact of alternative forms of incentivization, such as donation to health charities, is warranted.

Acknowledgments

The Chief Scientist Office of the Scottish Government Health and Social Care Directorates funds the Health Economics Research Unit. Alastair Irvine’s PhD studentship is funded by the Institute of Applied Health Sciences, University of Aberdeen. The views expressed in this article are those of the authors only and not those of the funding body.

Further Reading

Cairns, J., van der Pol, M., & Lloyd, A. (2002). Decision-making heuristics and the elicitation of preferences: Being fast and frugal about the future. Health Economics, 11, 655–658.Find this resource:

Cohen, J. D., Ericson, K. M., Laibson, D., & White, J. M. (2016). Measuring time preferences (Working Paper 22455). Cambridge, MA: National Bureau of Economic Research.Find this resource:

Frederick, S., Loewenstein, G., & O’Donoghue, T. (2002). Time discounting and time preference: A critical review. Journal of Economic Literature, 40, 351–401.Find this resource:

Fuchs, V. R. (1986). In Fuchs V. R. (Ed.), Time preference and health. Cambridge, MA: Harvard University Press.Find this resource:

Severens, J. L., & Milne, R. J. (2004). Discounting health outcomes in economic evaluation: The ongoing debate. Value in Health, 7, 397–401.Find this resource:

Story, G. W., Vlaev, I., Seymour, B., Darzi, A., & Dolan, R. J. (2014). Frontiers in Behavioral Neuroscience, 8, 76.Find this resource:

van der Pol, M. M., & Cairns, J. A. (2000). Negative and zero time preference for health. Health Economics, 9, 171–175.Find this resource:

References

Amlung, M., Vedelago, L., Acker, J., Balodis, I., & MacKillop, J. (2017). Steep delay discounting and addictive behavior: A meta-analysis of continuous associations. Addiction, 112, 51–62.Find this resource:

Andersen, S., Girolamo, A., Harrison, G. W., & Lau, M. I. (2014). Risk and time preferences of entrepreneurs: Evidence from a Danish field experiment. Theory and Decision, 77, 341–357.Find this resource:

Andersen, S., Harrison, G. W., Lau, M. I., & Rutstrom, E. E. (2008). Eliciting risk and time preferences. Econometrica, 76, 583–618.Find this resource:

Andersen, S., Harrison, G. W., Lau, M. I., & Rutstrom, E. E. (2014). Discounting behavior: A reconsideration. European Economic Review, 71, 15–33.Find this resource:

Andreoni, J., & Sprenger, C. (2012). Estimating time preferences from convex budgets. American Economic Review, 102(7), 3333–3356.Find this resource:

Attema, A. E. (2012). Developments in time preference and their implications for medical decision making. Opent extern. European Journal of Operational Research, 63(10), 1388–1399.Find this resource:

Attema, A. E., Bleichrodt, H., Gao, Y., Huang, Z., & Wakker, P. P. (2016). Measuring discounting without measuring utility. American Economic Review, 106, 1476–1494.Find this resource:

Attema, A. E., Bleichrodt, H., & Wakker, P. P. (2012). A direct method for measuring discounting and QALYs more easily and reliably. Medical Decision Making, 32, 583–593.Find this resource:

Attema, A. E., & Brouwer, W. B. (2013). In search of a preferred preference elicitation method: A test of the internal consistency of choice and matching tasks. Journal of Economic Psychology, 39, 126.Find this resource:

Attema, A. E., & Versteegh, M. M. (2013). Would you rather be ill now, or later?. Health Economics, 22, 1496–1506.Find this resource:

Augenblick, N., Niederle, M., & Sprenger, C. (2015). Working over time: Dynamic inconsistency in real effort tasks. Quarterly Journal of Economics, 130, 1067–1115.Find this resource:

Bansback, N., Brazier, J., Tsuchiya, A., & Anis, A. (2012). Using a discrete choice experiment to estimate health state utility values. Journal of Health Economics, 31, 306–318.Find this resource:

Barlow, P., McKee, M., Reeves, A., Stuckler, G., & Stuckler, D. (2017). Time-discounting and tobacco smoking: A systematic review and network analysis. International Journal of Epidemiology, 46, 860–869.Find this resource:

Barlow, P., Reeves, A., McKee, M., Galea, G., & Stuckler, D. (2016). Unhealthy diets, obesity and time discounting: A systematic literature review and network analysis. Obesity Reviews, 17, 810–819.Find this resource:

Bleichrodt, H., Gao, Y., & Rohde, K. I. M. (2016). A measurement of decreasing impatience for health and money. Journal of Risk and Uncertainty, 52, 213–231.Find this resource:

Bleichrodt, H., & Johannesson, M. (2001). Time preference for health: A test of stationarity versus decreasing timing aversion. Journal of Mathematical Psychology, 45, 265–282.Find this resource:

Bobinac, A., Brouwer, W., & van Exel, J. (2011). Discounting future health gains: An empirical enquiry into the influence of growing life expectancy. Health Economics, 20, 111–119.Find this resource:

Brosig-Koch, J., Hennig-Schmidt, H., Kairies-Schwarz, N., & Wiesen, D. (2017). The effects of introducing mixed payment systems for physicians: Experimental evidence. Health Economics, 26, 243–262.Find this resource:

Cameron, T. A. (1988). A new paradigm for valuing non-market goods using referendum data: maximum likelihood estimation by censored logistic regression. Journal of Environmental Economics and Management, 15, 355-379.Find this resource:

Cairns, J. A. (1992a). Discounting and health benefits: Another perspective. Health Economics, 1, 76–79.Find this resource:

Cairns, J. A. (1992b). Health, wealth and time preference. Project Appraisal, 7, 31–40.Find this resource:

Cairns, J. A. (1994). Valuing future benefits. Health Economics, 3, 221–229.Find this resource:

Cairns, J. A., & van der Pol, M. (1997a). Constant and decreasing timing aversion for saving lives. Social Science and Medicine, 45, 1653–1659.Find this resource:

Cairns, J. A., & van der Pol, M. M. (1997b). Saving future lives: A comparison of three discounting models. Health Economics, 6, 341–350.Find this resource:

Cairns, J., & van der Pol, M. (1999). Do people value their own future health differently from others’ future health? Medical Decision Making, 19, 466–472.Find this resource:

Cairns, J., van der Pol, M., & Lloyd, A. (2002). Decision-making heuristics and the elicitation of preferences: Being fast and frugal about the future. Health Economics, 11, 655–658.Find this resource:

Camerer, C. (1998). Bounded rationality in individual decision-making. Experimental Economics, 1, 163–183.Find this resource:

Cawley, J. (2015). An economy of scales: A selective review of obesity’s economics causes, consequences and solutions. Journal of Health Economics, 43, 244–268.Find this resource:

Chapman, G. B. (1996a). Expectations and preferences for sequences of health and money. Organizational Behavior and Human Decision Processes, 67, 59.Find this resource:

Chapman, G. B. (1996b). Temporal discounting and utility for health and money. Journal of Experimental Psychology: General, 22, 771–791.Find this resource:

Chapman, G. B., & Coups, E. J. (1999). Time preferences and preventive health behavior: Acceptance of the influenza vaccine. Medical Decision Making, 19, 307–314.Find this resource:

Chapman, G. B., & Elstein, A. S. (1995). Valuing the future: Temporal discounting of health and money. Medical Decision Making, 15, 373–386.Find this resource:

Chapman, G. B., Nelson, R., & Hier, D. D. (1999). Familiarity and time preferences: Decision-making about treatments for migraine headaches and Crohn’s disease. Journal of Experimental Psychology: Applied, 5, 17–34.Find this resource:

Cohen, J. D., Ericson, K. M., Laibson, D., & White, J. M. (2016). Measuring time preferences (Working Paper 22455). Cambridge, MA: National Bureau of Economic Research.Find this resource:

Cropper, M. L., Aydede, S. K., & Portney, P. R. (1991). Discounting human lives. American Journal of Agricultural Economics, 73, 1410–1415, 1428–1430.Find this resource:

Cropper, M. L., Aydede, S. K., & Portney, P. R. (1994). Preferences for life saving programs: How the public discounts time and age. Journal of Risk and Uncertainty, 8, 243–265.Find this resource:

Dasgupta, P., & Maskin, E. (2005). Uncertainty and hyperbolic discounting. American Economic Review, 95, 1290–1299.Find this resource:

Dolan, P., & Gudex, C. (1995). Time preference, duration and health state valuations. Health Economics, 4, 289–299.Find this resource:

Dolan, P., & Jones-Lee, M. (1997). The time trade-off: A note on the effect of lifetime reallocation of consumption and discounting. Journal of Health Economics, 16, 731–739.Find this resource:

Enemark, U., Lyttkens, C. H., Troeng, T., Weibull, H., & Ranstam, J. (1998). Implicit discount rates of vascular surgeons in the management of abdominal aortic aneurysms. Medical Decision Making, 18, 168–177.Find this resource:

Frederick, S., Loewenstein, G., & O’Donoghue, T. (2002). Time discounting and time preference: A critical review. Journal of Economic Literature, 40, 351–401.Find this resource:

Fuchs, V. R. (Ed.). (1986). Time preference and health. Cambridge, MA: Harvard University Press.Find this resource:

Fuchs, V. R., & Zeckhauser, R. (1987). Valuing health: A “priceless” commodity. American Economic Review, 77, 263–268.Find this resource:

Galizzi, M. M., Miraldo, M., Stavropoulou, C., & van der Pol, M. (2016). Doctor-patient differences in risk and time preferences: A field experiment. Journal of Health Economics, 50, 171–182.Find this resource:

Ganiats, T. G., Carson, R. T., Hamm, R. M., Cantor, S. B., Sumner, W., Spann, S. J., . . . Miller, C. (2000). Population-based time preferences for future health outcomes. Medical Decision Making, 20, 263–270.Find this resource:

Gyrd-Hansen, D. (2002). Comparing the results of applying different methods of eliciting time preferences for health. European Journal of Health Economics, 3, 10–16.Find this resource:

Hardisty, D. J., & Weber, E. U. (2009). Discounting future green: Money versus the environment. Journal of Experimental Psychology: General, 138, 329–340.Find this resource:

Harris, C. R. (2012). Feelings of dread and intertemporal choice. Journal of Behavioral Decision Making, 25, 13.Find this resource:

Hennig-Schmidt, H., Selten, R., & Wiesen, D. (2011). How payment systems affect physicians’ provision behavior: An experimental investigation. Journal of Health Economics, 30, 637–646.Find this resource:

Hermann, D., & Musshoff, O. (2016). Anchoring effects in experimental discount rate elicitation. Applied Economics Letters, 23, 1022–1025.Find this resource:

Hojgard, S., Enemark, U., Lyttkens, C. H., Lindgren, A., Troeng, T., & Weibull, H. (2002). Discounting and clinical decision-making: Physicians, patients, the general public, and the management of asymptomatic abdominal aortic aneurysms. Health Economics, 11, 355–370.Find this resource:

Horowitz, J. K., & Carson, R. T. (1990). Discounting statistical lives. Journal of Risk and Uncertainty, 3, 403–413.Find this resource:

Jevons, W. S. (1911). The theory of political economy, by W. Stanley Jevons, with notes and an extension of the bibliography of mathematical economic writings, by H. Stanley Jevons (4th ed.). London: Macmillan.Find this resource:

Jit, M., & Mibei, W. (2015). Discounting in the evaluation of the cost-effectiveness of a vaccination programme: A critical review. Vaccine, 33, 3788–3794.Find this resource:

Johannesson, M., & Johansson, P. O. (1996). The discounting of lives saved in future generations: Some empirical results. Health Economics, 5, 329–332.Find this resource:

Johannesson, M., & Johansson, P. (1997a). Quality of life and the WTP for an increased life expectancy at an advanced age. Journal of Public Economics, 65, 219–228.Find this resource:

Johannesson, M., & Johansson, P. (1997b). The value of life extension and the marginal rate of time preference: A pilot study. Applied Economics Letters, 4(1), 53–55.Find this resource:

Khwaja, A., Silverman, D., & Sloan, F. (2007). Time preference, time discounting, and smoking decisions. Journal of Health Economics, 26, 927–949.Find this resource:

Krahn, M., & Gafni, A. (1993). Discounting in the economic evaluation of healthcare interventions. Medical Care, 31, 403–418.Find this resource:

Laibson, D. (1997). Golden eggs and hyperbolic discounting. Quarterly Journal of Economics, 112, 443–477.Find this resource:

Lazaro, A., Barberan, R., & Rubio, E. (2001). Private and social time preferences for health and money: An empirical estimation. Health Economics, 10, 351–356.Find this resource:

Lazaro, A., Barberan, R., & Rubio, E. (2002). The discounted utility model and social preferences: Some alternative formulations to conventional discounting. Journal of Economic Psychology, 23, 317–338.Find this resource:

Lipscomb, J. (1989). Time preference for health in cost-effectiveness analysis. Medical Care, 27(3 Suppl.), S233–253.Find this resource:

Loewenstein G. (1987) Anticipation and the valuation of delayed consumption. The Economic Journal, 97, 666–684.Find this resource:

Loewenstein, G., & Prelac, D. (1991). Negative time preference. American Economic Review, 81, 347–352.Find this resource:

MacKeigan, L. D. (1993). Time preference for health gains versus health losses. PharmacoEconomics, 3, 374–386.Find this resource:

Mazur, J. E. (1987). An adjusting procedure for studying delayed reinforcement. In M. L. Commons, J. E. Mazur, J. A. Nevin, & H. Rachlin (Eds.), Quantitative analyses of behaviour V: The effect of delay and of intervening events on reinforcement value. Hillsdale: Erlbaum.Find this resource:

Meerding, W. J., Bonsel, G. J., Brouwer, W. B. F., Stuifbergen, M. C., & Essink-Bot, M. (2010). Social time preferences for health and money elicited with a choice experiment. Value in Health, 13, 368–374.Find this resource:

Olsen, J. A. (1993). Time preferences for health gains: An empirical investigation. Health Economics, 2, 257–265.Find this resource:

Olsen, J. A. (1994). Persons vs years: Two ways of eliciting implicit weights. Health Economics, 3, 39–46.Find this resource:

Olson, M., & Bailey, M. J. (1981). Positive time preference. Journal of Political Economy, 89, 1–25.Find this resource:

Poulos, C., & Whittington, D. (2000). Time preferences for life-saving programs: Evidence from six less developed countries. Environmental Science and Technology, 34, 1445–1455.Find this resource:

Rae, J. (1905). The sociological theory of capital. London: Macmillan.Find this resource:

Redelmeier, D. A., & Heller, D. N. (1993). Time preference in medical decision-making and cost-effectiveness analysis. Medical Decision Making, 13, 212–217.Find this resource:

Robberstad, B. (2005). Estimation of private and social time preferences for health in Northern Tanzania. Social Science and Medicine, 61, 1597–1607.Find this resource:

Senior, N. W. (1836). An outline of the science of political economy. London: Clowes and Son.Find this resource:

Severens, J. L., & Milne, R. J. (2004). Discounting health outcomes in economic evaluation: The ongoing debate. Value in Health, 7, 397–401.Find this resource:

Stavem, K., Kristiansen, I. S., & Olsen, J. A. (2002). Association of time preference for health with age and disease severity. European Journal of Health Economics, 3, 120–124.Find this resource:

Story, G. W., Vlaev, I., Seymour, B., Darzi, A., & Dolan, R. J. (2014). Does temporal discounting explain unhealthy behavior? A systematic review and reinforcement learning perspective. Frontiers in Behavioral Neuroscience, 8Find this resource:

van der Pol, M., & Cairns, J. (1999). Individual time preferences for own health: An application of a dichotomous choice question with follow-up. Applied Economics Letters, 6, 649–654.Find this resource:

van der Pol, M., & Cairns, J. A. (2000). Negative and zero time preference for health. Health Economics, 9, 171–175.Find this resource:

van der Pol, M., & Cairns, J. (2001). Estimating time preferences for health using discrete choice experiments. Social Science and Medicine, 52, 1459–1470.Find this resource:

van der Pol, M., & Cairns, J. (2011). Descriptive validity of alternative intertemporal models for health outcomes: An axiomatic test. Health Economics, 20, 770–782.Find this resource:

van der Pol, M., Walsh, D., & McCartney, G. (2015). Comparing time and risk preferences across three post-industrial UK cities. Social Science and Medicine, 140, 54–61.Find this resource:

Viscusi, W. K., & Moore, M. J. (1989). Rates of time preference and valuations of the duration of life. Journal of Public Economics, 38, 297–317.Find this resource:

Zweifel, P., & Breyer, F. (1997). Health economics. New York: Oxford University Press.Find this resource:

Notes:

(1.) One of the exceptions is risk-premia on dangerous jobs. Viscusi and Moore (1989) use revealed preference data on the trade-off between wages and risk to estimate individuals’ time-preference rates.

(2.) Even if the experimenter introduced a randomization over the two outcomes, it would be dominant to write a large monetary value (to increase the expected value of the gamble); or, if time is the free variable, to write “now” for the larger option.