A series of three studies explored the effects of uncertainty on time preferences for financial gains and losses. When future outcomes were uncertain, participants preferred immediate, certain gains and losses. Thus, in a departure from Prospect Theory, people appear to be risk averse for both gains and losses in the intertemporal context. This theory was further supported in a second study, where participants avoided immediate uncertainty in favor of future certainty, for both gains and losses. Yet when participants considered only immediate gains and losses, they showed a Prospect Theory-consistent pattern of risk preferences, establishing that general risk aversion is unique to the intertemporal context. The effects of outcome uncertainty on discount rates were quite strong, particularly when compared to the effects of demographic factors. The results have broad implications for policy, as the presence or absence of uncertainty can be used to encourage future-oriented choices.
Temporal preferences for uncertain gains and losses
Many financial choices involve both delay and uncertainty. With delayed gains, people often think of the uncertainty about whether they will actually receive the money or not (Patak & Reynolds, 2007; Takahashi, Ikeda, & Hasegawa, 2007). Therefore, uncertainty leads people to prefer the immediate (and therefore more certain) gain. With delayed losses, however, people may often consider whether or not they will be able to pay in the future, which leads people to prefer the immediate payment (which is a known quantity, and their financial situation is also known). Previous research on intertemporal choice has investigated the first type of uncertainty (e.g., whether the future reward will actually be received; Anderson & Stafford, 2009; Keren & Roelofsma, 1995), but has virtually ignored the second type (time preferences for uncertain losses). The present research fills that gap, by exploring whether uncertainty has different effects on time preferences for gains and losses.
Standard economic models consider risk and time as separate, orthogonal dimensions (Pindyck & Rubinfeld, 2008), where the consideration and analysis of one should not affect the other: outcomes are first multiplied by their probabilities (to calculate the expected value) and then discounted according to their delay to determine their value. Perhaps for this reason, most studies of temporal discounting have ignored the potential effects of uncertainty (Frederick, Loewenstein, & O'Donoghue, 2002).
However, a growing literature has documented that people implicitly associate future gains with uncertainty, even when it is not mentioned in the experimental procedure. There is evidence that delayed rewards were rated as increasingly uncertain at progressively longer delays (Patak & Reynolds, 2007; Reynolds, Patak, & Shroff, 2007; Takahashi et al., 2007). Furthermore, ratings of uncertainty were correlated with discount rates1 (r=.55, r=.37, and rho=.47, respectively). Along the same lines, risk aversion and temporal discounting of rewards are often correlated (Anderhub, Guth, Gneezy, & Sonsino, 2001; Jones & Rachlin, 2009), such that participants preferring certain gains also tend to prefer immediate gains. Taken together, these results suggest that people think future gains are inherently uncertain and therefore prefer immediate, certain gains.
In spite of the fact that risk preferences are different for gains and losses (Kahneman & Tversky, 1979) and temporal preferences are also different for gains and losses (Hardisty & Weber, 2009; Mischel, Grusec, & Masters, 1969; Thaler, 1981), most research on temporal preferences for risky outcomes has only explored gains. This gap in the literature is especially problematic because the existing findings on time and risk seem to contradict each other. In the domain of risk, people are risk averse for gains and risk seeking for losses (Kahneman & Tversky, 1979). If time delay is implicitly associated with uncertainty, this should translate into a preference for immediate gains and future losses (in other words, high discount rates for both gains and losses). However, while people do indeed show a strong desire for immediate gains, the converse is not true for losses: many people choose to get losses over with as soon as possible (Hardisty & Weber, 2009; Mischel et al., 1969; Thaler, 1981). How can this apparent contradiction be reconciled?
To the best of our knowledge, only two papers have empirically compared time preferences for uncertain gains and losses. One found greater discount rates for losses than gains (Shelley, 1994), while the other found no difference in time preferences for gains and losses (Ahlbrecht & Weber, 1997). However, the generalizability of these studies may be limited for two reasons. First, the samples used (MBA students and finance students, respectively) probably had different risk and time preferences than the general population, especially because they had taken courses on the "correct" ways to deal with these issues. Second, both studies included extensive training periods (45 minutes and 15 minutes, respectively) which may have created demand characteristics. Indeed, the training protocol in Ahlbrecht and Weber (1997) explicitly suggested appropriate responses to participants. Thus, the effects of uncertainty on time preferences for gains and losses in the general population remain unexplored.
In the present research, we investigated time preferences for certain and uncertain gains and losses, with a diverse sample of U.S. residents. Furthermore, we explored two different types of uncertainty: probability and variability. The probability condition involved financial outcomes with only a 50% chance of occurring, mirroring previous research on uncertainty. In the variability condition, all outcomes were certain to occur, but the exact amounts were unknown; rather, a possible range of amounts was specified. For example, a choice was offered between losing $100 today or $75 to $225 in one year. In both cases, we predicted that uncertainty would be aversive to participants. In the case of gains, people avoid future uncertainty because they worry about not receiving the full amount, while in the case of losses, people avoid uncertainty because they are worried about not being able to pay the full amount. Therefore, we hypothesized that future uncertainty would lead to greater discounting of gains but lower discounting of losses.
We examined medium ($100) to large ($10,000) magnitude outcomes, because research has found that magnitude affects discount rates for gains and losses in different ways (Hardisty, Appelt, & Weber, 2012; Mitchell & Wilson, 2010). We predicted that magnitude would show the same effects as in previous research (with lower discounting of larger magnitude gains, but greater discounting of larger magnitude losses).
The above predictions concern how time preferences are affected by uncertainty in outcomes. There is also a small literature on how time preferences are related to uncertainty in the life of the decision maker. When people feel disconnected from their future selves, they show higher discount rates, valuing immediate utility at the expense of future utility (Bartels & Rips, 2010; Bartels & Urminsky, 2011; Ersner-Hershfield, Wimmer, & Knutson, 2008). Thus, someone who is uncertain about their future self is also uncertain about the subjective value of future rewards and, as a consequence, devalues those rewards. Another line of literature has documented links between impulsivity and risk seeking behavior (Chabris, Laibson, Morris, Schuldt, & Taubinsky, 2008; Reimers, Maylor, Stewart, & Chater, 2009). Participants with higher discount rates for rewards (as measured in the lab) are more likely to smoke, less likely to exercise, more likely to be overweight, and more prone to sexual infidelity. Discount rates are also related to financial insecurity (Chabris et al., 2008; Meier & Sprenger, 2010; Meier & Sprenger, 2012): participants that select a smaller, sooner financial reward in lab experiments tend to have lower income, more credit card debt, and worse credit scores. Despite these interesting findings, however, the relationship between time preferences for financial losses and demographic uncertainty is unknown.
In the present research, we collected information about financial uncertainty in participants' lives. Specifically, we assessed their employment status, available financial resources, feelings of financial security, and whether they worked for a young company (with the idea that new companies are somewhat insecure and more likely to fail in the near future). We hypothesized that greater financial insecurity in participants' lives would be associated with greater discounting of both gains and losses: when considering gains, economically insecure people want immediate rewards, to help them with their present financial difficulties. Similarly, economically insecure people prefer to postpone losses, in hopes that their financial situation may improve in the future and thus it will be easier to pay their bills. However, based on the previous weak effects of demographic variables predicting discount rates (Chabris et al., 2008; Reimers et al., 2009), we expected that the power of economic insecurity to predict discount rates would be relatively weak, especially when compared with the situational factors of gains versus losses and magnitude effects.
A sample of 150 U.S. residents (mean age = 45, SD=15) was recruited from Amazon Mechanical Turk and Survey Sampling International for a study on decision making. Participants were only eligible to participate if they were at least 18 years old and passed an attention check (Oppenheimer, Meyvis, & Davidenko, 2009) on the first page of the study. Participants completed several different tasks, in counterbalanced order. The overall design was a 2 (presence of uncertainty, within) x 2 (type of uncertainty, between) x 2 (sign, within) x 2 (magnitude, within).
Intertemporal choice scenarios
All participants responded to four typical intertemporal choice scenarios, in counterbalanced order: medium gain, large gain, medium loss, and large loss. In the medium gain scenario, participants read the instruction, "Please imagine you face a set of choices about receiving $100 from investments immediately, or another amount 1 year from now. Please indicate which option you would choose in each case:" This was followed by six intertemporal choices (detailed in Online Supplemental A), such as "Receive $100 immediately OR receive $110 in 1 year." In the medium loss scenario, participants read the instruction, "Please imagine you face a set of choices about paying a $100 bill immediately, or another amount 1 year from now. Please indicate which option you would choose in each case:" This was followed by six intertemporal choices (detailed in Online Supplemental A), such as "Pay $100 immediately OR pay $110 in 1 year." The large gain and large loss scenarios were identical, except that all amounts were 100 times larger.
Intertemporal choice scenarios with future uncertainty
All participants also responded to four intertemporal choice scenarios involving uncertain future outcomes, which were structured the same as the regular intertemporal choice questions (with medium and large gains and losses). Participants were randomly assigned to one of two different uncertainty conditions: probabilistic outcomes or variable outcomes.
In the probabilistic condition, the future options were twice as large, but only had a 50% chance of occurring (for the complete list of choice options, see Online Supplemental A). Thus, the expected value of the options was the same in the certainty and uncertainty conditions. At the beginning of the medium probabilistic gain scenario, participants read the instruction, "Please imagine you face a set of choices about receiving $100 from investments immediately, or another amount 1 year from now that would be uncertain (only a 50% chance of receiving it, which would be determined randomly, one year from now). Please indicate which option you would choose in each case:" The instructions for the large magnitude condition were the same except that the amount was $10,000, and the instructions for the loss conditions were the same except that they concerned paying a bill. There were six intertemporal choice pairs for each scenario, such as "Receive $100 immediately OR 50% chance of receiving $220 in 1 year."
In the variability condition, all outcomes were certain to occur, but the exact amounts were uncertain. Specifically, the uncertain amounts ranged from -50% to +50% of the base amount. The instructions in the medium gain condition were " Please imagine you face a set of choices about receiving $100 from investments immediately, or another amount 1 year from now that would be a variable amount. (The exact amount would be determined randomly, one year from now.) Please indicate which option you would choose in each case:" An example choice pair was "Receive $100 immediately OR receive $55 to $165 in 1 year." The instructions and choice pairs for the loss scenarios were similar.
After all the intertemporal choice questions, participants answered several demographic questions, including age, annual household income, education, employment status (not employed, part time, or full time), and whether they were a smoker. A subset of 30 participants also answered a question about their available financial resources: "Imagine that you had to pay an unexpected bill immediately. For example, suppose that you needed an expensive medical treatment that was not covered by insurance. Considering all possible resources available to you (including savings, borrowing, etc.), what is the maximum amount that you could come up with on short notice?" Subsequently, participants who indicated being employed also answered two questions about the size and age of the organization employing them.
Participants' responses to the intertemporal choice questions were converted into indifference points by taking the average of the values around the point they switched between the immediate and future option. For example, if a participant chose to receive $100 today over $120 in one year, and chose $140 in one year over $100 today, then the participant was judged to be indifferent between receiving $100 today and $130 in one year. If the participant maxed out the scale (for example, by always choosing the immediate option), the indifference point was put one step beyond the maximum option (or minimum option, as appropriate). To easily compare preferences across magnitudes and conditions, indifference points were converted into discount rates using the continuously compounded exponential formula V=Ae-kD (Samuelson, 1937), where V is the immediately available amount (e.g., $10), A is the future amount (e.g., $13), e is the constant (2.718), D is the delay in years, and k is a fitted parameter, the discount rate. We chose this equation (rather than the hyperbolic model or the area-under-the-curve method) because it is easily interpretable. For example, a k of .6 indicates a discount rate of 60%, in the standard economic sense (similar to an interest rate). Higher numbers indicate greater discounting, a k of zero means no discounting, and negative k values indicate negative discounting. As choices in this study all involved the same two time points (immediate outcomes vs. outcomes in one year), exponential and hyperbolic modeling would fit the data equally well, so there was no advantage to using the hyperbolic model (which is known to generally describe data better than the exponential model, Kirby, 1997; Kirby & Marakovic, 1995; Mazur, 1987).
Time preferences for medium and large gains and losses
As seen in Figure 1, large gains were discounted more than medium gains, but this magnitude effect was eliminated or even reversed for losses. A 2x2 repeated measures GLM revealed main effects of magnitude, F(1,327)=52.3, p<.001, η2=.14, and sign (i.e, gains versus losses), F(1, 327)=147.0, p<.001, η2=.31, which were qualified by a sign by magnitude interaction, F(1, 327)=63.8, p<.001, η2=.16. In dollar terms, participants were indifferent between receiving $100 today or $130 next year, receiving $10,000 today or $11,600 next year, paying $100 today or $104 next year, and paying $10,000 today or $10,600 next year. This pattern of time preferences replicates previous research (Baker, Johnson, & Bickel, 2003; Estle, Green, Myerson, & Holt, 2006; Hardisty et al., 2012).
In sharp contrast, when the immediate outcome was certain and the future outcome was uncertain, participants strongly preferred the immediate, certain option. A 2x2 GLM showed a strong sign effect, F(1,149)=263.1, p<.001, η2=.64, but no magnitude effect, F(1,149)=0.1, p=.72, η2=.00, or interaction, F(1,149)=0.3, p=.57, η2=.00. In dollar terms, participants were indifferent between receiving $100 today or a 50% chance of $306 next year, receiving $10,000 today or a 50% chance of $307,200 next year, paying $100 today or a 50% chance of $189 next year, and paying $10,000 today or a 50% chance of $18,700 next year. The results for probability and variability were fairly similar, so we have grouped them together (for separate results, see Online Supplemental B). The only significant difference was that probabilistic losses were discounted less than variable losses, F(1,148)=22.6, p<.001, η2=.13. For both forms of uncertainty, uncertain gains were discounted more than certain gains, F(1,149)=67.5, p<.001, η2=.31, and uncertain losses were discounted less than certain losses, F(1,149)=98.8, p<.001, η2=.40.
Relationship of discount rates to demographic variables
Discount rates for gains and losses were correlated with demographic variables in all three studies. The relationships were quite similar, so the combined data from all three studies is presented in Table 1 (separate correlation tables for each study can be found in Online Supplemental C). Non-parametric correlation (Spearman's rho) was used because the income and financial resource variables were extremely skewed. Overall, discount rates correlated with demographic uncertainty in the expected direction, in that people with greater financial insecurity showed higher discount rates. Specifically, participants who were employed part time or worked for a young organization showed higher discount rates, while participants with more income, financial resources, education, and economic security evidenced lower discount rates. These relationships were relatively weak, however, similar to other demographic variables studied in previous research (Chabris et al., 2008; Reimers et al., 2009).
As predicted, participants avoided future uncertainty, for both gains and losses. This translated into very high discount rates for gains, and very low discount rates for losses. This contrasts with the results of earlier research on discount rates for uncertain gains and losses, which found no difference between gains and losses (Ahlbrecht & Weber, 1997) or greater discounting for losses (Shelley, 1994). One explanation for the observed difference is that the previous studies used business school students as participants who had, moreover, been trained in appropriate ways to deal with delay and probability, while we used a sample from the general population that was not trained in how to make intertemporal choices.
Moreover, this pattern of results -- uncertain gains discounted more than certain gains, and uncertain losses discounted less than certain losses -- was true for two different forms of uncertainty: probability and variability. This lends additional confidence to the findings.
While Study 1 only examined future uncertainty, in some cases the present is more uncertain than the future. For example, the payoff from a project or investment may not be immediately clear, but may become certain in the future (as additional time and effort are put in). Therefore, in Study 2 we examined time preferences when choosing between immediate, uncertain outcomes and future, certain outcomes. We predicted that, like Study 1, participants would prefer to avoid the risky options. This would lead to lower discount rates for gains, but higher discount rates for losses (as compared with certain outcomes).
Another issue concerns the fact that our findings seem to contradict Prospect Theory (Kahneman & Tversky, 1979). We found that participants avoided risk for both future gains and future losses, whereas Prospect Theory predicts risk aversion for gains and risk seeking for losses. One possible explanation is that people have different risk preferences when considering risky prospects in an intertemporal context than when both outcomes are immediate. Therefore, in Study 2 we presented participants with choices between certain and uncertain immediate outcomes, to test the standard Prospect Theory predictions and ensure there was nothing untoward about our samples or the design of the studies.
A sample of 76 participants was recruited in the same manner as Study 1. The intertemporal choice questions were the same as in Study 1, except that the immediate outcomes were uncertain rather than the future outcomes (for the complete text, see Online Supplemental A). Participants also answered a series of risky choice questions in which both outcomes were immediate. The medium magnitude risky choice questions were always presented before the large questions, but the order of the gain and loss questions was counterbalanced. Participants read the instruction "Please imagine you are actually faced with the following choices, and indicate which option you would prefer each case:" Participants then made choices between the following options: "50% chance to receive $200, 50% chance to receive nothing OR receive $100 for sure", "50% chance to receive $20,000, 50% chance to receive nothing OR receive $10,000 for sure", "50% chance to pay $200, 50% chance to pay nothing OR pay $100 for sure", and "50% chance to pay $20,000, 50% chance to pay nothing OR pay $10,000 for sure."
Finally, participants answered demographic questions and questions about financial security, which have already been reported in the Study 1 methods and results.
The discount rates for certain gains and losses replicated the results of Study 1. A 2x2 GLM found main effects of sign, F(1,75)=37.6, p<.001, η2=.33, magnitude, F(1,75)=23.1, p<.001, η2=.24, and the interaction, F(1,75)=18.2, p<.001, η2=.20; large gains were discounted more than medium gains, but this magnitude effect was eliminated for losses.
As seen in Figure 2, when balancing immediate uncertainty versus future certainty, participants avoided the uncertainty (parallel to the results of Study 1). This led to preference for larger, later gains (i.e, lower discount rates for gains) and larger, later losses (i.e, higher discount rates for losses). A 2x2 GLM found no main effect of sign, F(1,75)=2.4, p=.12, η2=.03, but did find a magnitude effect, F(1,75)=11.2, p<.001, η2=.13, and an interaction, F(1,75)=8.1, p=.006, η2=.10. Focused comparisons confirmed lower discount rates for gains in the uncertainty condition than the certainty condition, F(1,75)=45.3, p<.001, η2=.38, and higher discount rates for losses in the uncertainty condition than in the certainty condition, F(1,75)=6.3, p=.01, η2=.08.
However, when all outcomes were immediate, participants showed a pattern of results more typical of Prospect Theory, as seen in Figure 3. Participants were more risk averse for gains than for losses, F(1,75)=49.5, p<.001, η2=.40. Participants were also more risk averse for large outcomes than small outcomes, F(1, 75)=16.6, p<.001, η2=.18. The interaction of magnitude and sign on risk preferences was not significant, F(1, 75)=0.8, p=.36, η2=.01.
As in Study 1, participants avoided uncertainty in the intertemporal context, for both gains and losses. However, because the immediate outcomes were uncertain (rather than the future outcomes), this led to lower discount rates for gains and higher discount rates for losses, wiping out the intertemporal "sign effect" altogether.
When there was no temporal component, however, participants evidenced a pattern of risk seeking consistent with Prospect Theory, showing strong risk aversion for gains and something like risk-neutrality for losses. This demonstrates that our findings on uncertainty are unique to the intertemporal context.
It is also the case that in the real world people often cannot choose between certainty and uncertainty and, instead, confront choices all of which entail some degree of uncertainty. Therefore, in Study 3, we examined the case where both immediate and future outcomes were uncertain. In this case, there would be no opportunity for participants to avoid the uncertainty, so we predicted that their discount rates would be similar to the condition where both outcomes are certain.
A sample of 102 participants was recruited in the same manner as Study 1 and Study 2. The intertemporal choice scenarios were the same as before, except that in this case, both the immediate and future outcomes were uncertain (for the full text, see Online Supplemental A).
Discount rates for the certain outcomes replicated the results from the previous studies, with main effects of sign, F(1,101)=42.7, p<.001, η2=.30, magnitude, F(1,101)=13.2, p<.001, η2=.12, and the interaction, F(1,101)=15.3, p<.001, η2=.13. When both immediate and future outcomes were uncertain, discount rates were extremely similar to those in the certainty condition, as seen in Figure 4. Participants showed the sign effect, F(1,101)=48.9, p<.001, η2=.33, magnitude effect, F(1,101)=4.3, p=.04, η2=.04, and the interaction, F(1,101)=9.2, p=.003, η2=.08. Differences between the certainty and uncertainty conditions were small and not significant; a GLM found no main effect of uncertainty (versus certainty), F(1,101)=1.0, p=.33, η2=.01, no interaction of uncertainty and sign, F(1,101)=3.7, p=.06, η2=.04, no interaction of uncertainty and magnitude, F(1,101)=1.1, p=.31, η2=.01, and no three-way interaction with uncertainty, F(1,101)=0.4, p=.51, η2=.00. A focused comparison of discount rates for uncertain versus certain gains was not significant, F(1,101)=3.7, p=.06, η2=.04, although there was a trend for uncertain gains to be discounted more than certain gains. For losses, there was no difference between uncertain versus certain losses, F(1,101)=0.4, p=.51, η2=.00.
These null effects have at least two possible explanations. One is that because uncertainty was present in all outcomes, and therefore unavoidable, it had no effect on participants' thoughts and choices: they may have mentally edited out the uncertainty and focused on the features that were different between each choice option. A second interpretation is that participants were strongly affected by the uncertainty (as in Study 1 and Study 2), but the two sources of uncertainty pushed them in opposite directions, cancelling each other out and resulting in a null effect. This interpretation is supported by the fact that the discount rates for uncertain outcomes in Study 3 are not significantly different from the average of the discount rates for uncertain outcomes in Studies 1 and 2. This is shown in Figure 5.
Overall, uncertainty had very different effects on people's time preferences for gains and losses, depending on whether the future outcome was uncertain, the immediate outcome was, or both. To the best of our knowledge, these are the first studies with a national sample to compare intertemporal choice for uncertain gains and losses.
When future outcomes were uncertain participants strongly preferred certain, immediate gains and certain, immediate losses, in line with our predictions. We theorize that participants dislike uncertain future gains because they are unsure whether they will ever receive them, and dislike uncertain future losses because they are unsure whether they will be able to pay them. Our results thus diverge from Prospect Theory (Kahneman & Tversky, 1979), which would predict that uncertain future losses should be quite attractive because people are risk seeking for losses. Prospect Theory was developed with immediate gains and losses, and our results show that when all outcomes are immediate, participants are indeed risk averse for gains and risk seeking (or risk neutral) for losses. Thus, it appears that people's risk preferences are different when making intertemporal tradeoffs (or, equivalently, that their time preferences are different in the presence of risk). However, when both immediate and future outcomes are equally uncertain, the uncertainty has no effect on people's time preferences. Because there is no way to avoid the uncertainty, people may edit out the common feature and answer as if all outcomes were certain.
Across all these conditions of uncertainty, we found similar effects for probabilistic outcomes and variable outcomes. Thus, it appears that the observed effects are due to uncertainty in the broad sense, rather than probability in particular.
We also found that demographic financial uncertainty predicted time preferences; people in more precarious financial situations tended to prefer immediate gains and future losses. This relationship, although significant, was generally weak. The best demographic predictor of time preferences was the amount of financial resources currently available to the participant.
Our findings on the interaction of risk and time depart from standard economic analysis, in which one should not affect the other. The question remains, however, whether this interaction should be considered a bias (to be avoided), or simply a matter of taste. Therefore, while this paper has documented the behavioral consequences of different types of uncertainty, future research should empirically explore the process mechanisms underlying these differences. For example, do people think about different types of uncertainty when considering future gains and future losses? Thought listings might be useful to establish whether people do indeed think of outcome uncertainty when considering future gains, but personal financial uncertainty when considering future losses. If so, it may be reasonable for people to express different discount rates for gains and losses, and the sign effect would not be an irrational bias, but rather a reasonable way to avoid the pain of uncertainty.
Another shortcoming of the present research is that it has focused on pure gain scenarios and pure loss scenarios, whereas many real-world financial decisions involve tradeoffs between gains and losses at different times. For example, when buying a car, the benefits begin immediately and are relatively certain, while the costs are often delayed and somewhat uncertain (gas prices fluctuate, repair costs are unknown, etc). Therefore, future research should explore the effects of uncertainty for mixed gain-loss intertemporal scenarios.
As delays and uncertainty are so common in everyday financial decisions, the present results have potentially broad policy implications. People underinvest and allocate too high a percentage of their investments to relatively riskless bonds and money market funds. People's aversion to uncertain future gains may partly explain this "equity premium puzzle" (Hardin & Looney, 2012; Mehra & Prescott, 1985) concerning why people prefer bonds over stocks, even though stocks typically offer higher rates of return. People are risk averse and have a very high discount rate for uncertain future rewards, as our results demonstrate. Therefore, attempts to encourage people to value the future more and to take more risk might well try to introduce some sense of immediate uncertainty. Conversely, effort to encourage individuals to promptly pay their bills should try to eliminate immediate uncertainty and attempt to make future uncertainty more salient. For example, future credit card payments could be made more uncertain by tying the payment amounts(not just the interest rates) to future interest rates. Our results suggest that the future uncertainty and possibility of larger payments could drive people to prefer immediate and more certain payments. Our general point is that having people make decisions using lower discount rates and thereby encouraging long-term financial prudence depends importantly on how uncertain the future seems compared to the present.
Ahlbrecht, M., & Weber, M. (1997). An empirical study on intertemporal decision making under risk. Management Science, 43(6), 813-826. doi: 10.1287/mnsc.43.6.813
Anderhub, V., Guth, W., Gneezy, U., & Sonsino, D. (2001). On the interaction of risk and time preferences: An experimental study. German Economic Review, 2(3), 239-253.
Anderson, L. R., & Stafford, S. L. (2009). Individual decision-making experiments with risk and intertemporal choice. Journal of Risk & Uncertainty, 38(1), 51-72. doi: 10.1007/s11166-008-9059-4
Baker, F., Johnson, M. W., & Bickel, W. K. (2003). Delay discount in current and never-before cigarette smokers: Similarities and differences across commodity, sign, and magnitude. Journal of Abnormal Psychology, 112, 382-392. doi: 10.1037/0021-843X.112.3.382
Bartels, D. M., & Rips, L. J. (2010). Psychological connectedness and intertemporal choice. Journal of Experimental Psychology: General, 139, 49-69.
Bartels, D. M., & Urminsky, O. (2011). On intertemporal selfishness: How the perceived instability of identity underlies impatient consumption. Journal of Consumer Research, 38, 182-198.
Chabris, C. F., Laibson, D., Morris, C. L., Schuldt, J. P., & Taubinsky, D. (2008). Individual laboratory-measured discount rates predict field behavior. Journal of Risk and Uncertainty, 37. doi: 10.1007/s11166-008-9053-x
Ersner-Hershfield, H., Wimmer, G. E., & Knutson, B. (2008). Saving for the future self: Neural measures of future self-continuity predict temporal discounting. Scan 1-8. doi: 10.1093/scan/nsn042
Estle, S. J., Green, L., Myerson, J., & Holt, D. D. (2006). Differential effects of amount on temporal and probability discounting of gains and losses. Memory & Cognition, 34, 914-928. doi: 10.3758/BF03193437
Frederick, S., Loewenstein, G., & O'Donoghue, T. (2002). Time discounting and time preference: A critical review. Journal of Economic Literature, 40, 351–401. doi: 10.1257/002205102320161311
Hardin, A. M., & Looney, C. A. (2012). Myopic loss aversion: Demystifying the key factors influencing decision problem framing. Organizational Behavior & Human Decision Processes, 117, 311-331.
Hardisty, D. J., Appelt, K. C., & Weber, E. U. (2012). Good or bad, we want it now: Fixed-cost present bias for gains and losses explains magnitude asymmetries in intertemporal choice. Journal of Behavioral Decision Making, in press.
Hardisty, D. J., & Weber, E. U. (2009). Discounting future green: Money versus the environment. Journal of Experimental Psychology: General, 138(3), 329-340. doi: 10.1037/a0016433
Jones, B. A., & Rachlin, H. (2009). Delay, probability, and social discounting in a public goods game. Journal of the Experimental Analysis of Behavior, 91(1), 61-73. doi: 10.1901/jeab.2009.91-61
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263-291. doi: 10.2307/1914185
Keren, G., & Roelofsma, P. (1995). Immediacy and certainty in intertemporal choice. Organizational Behavior and Human Decision Processes, 63(3), 287-297. doi: 10.1006/obhd.1995.1080
Kirby, K. N. (1997). Bidding on the future: Evidence against normative discounting of delayed rewards. Journal of Experimental Psychology: General, 126, 54-70. doi: 10.1037//0096-34184.108.40.206
Kirby, K. N., & Marakovic, N. N. (1995). Modeling myopic decisions: Evidence for hyperbolic delay-discounting with subjects and amounts. Organizational Behavior and Human Decision Processes, 64, 22-30. doi: 10.1006/obhd.1995.1086
Mazur, J. E. (1987). An adjusting procedure for studying delayed reinforcement. In M. L. Commons, J. E. Mazure, J. A. Nevin & H. Rachlin (Eds.), Quantitative analyses of behavior: Vol. 5. The effect of delay and intervening events on reinforcement value (pp. 55-73). Hillsdale, NJ: Erlbaum.
Mehra, R., & Prescott, E. C. (1985). The equity premium: A puzzle. Journal of Monetary Economics, 15(2), 145-161.
Meier, S., & Sprenger, C. (2010). Present-biased preferences and credit card borrowing. American Economic Journal: Applied Economics, 2(1), 193-210. doi: 10.1257/app.2.1.193
Meier, S., & Sprenger, C. (2012). Time discounting predicts creditworthiness. Psychological Science, 23(1), 56-58.
Mischel, W., Grusec, J., & Masters, J. C. (1969). Effects of expected delay time on the subjective value of rewards and punishments. Journal of Personality and Social Psychology, 11(4), 363-373.
Mitchell, S. H., & Wilson, V. B. (2010). The subjective value of delayed and probabilistic outcomes: Outcome size matters for gains but not for losses. Behavioural Processes, 83(1), 36-40. doi: 10.1016/j.beproc.2009.09.003
Oppenheimer, D. M., Meyvis, T., & Davidenko, N. (2009). Instructional manipulation checks: Detecting satisficing to increase statistical power. Journal of Experimental Social Psychology, 45(4), 867-872. doi: 10.1016/j.jesp.2009.03.009
Patak, M., & Reynolds, B. (2007). Question-based assessments of delay discounting: Do respondents spontaneously incorporate uncertainty into their valuations for delayed rewards? Addictive Behaviors, 32, 351-357. doi: 10.1016/j.addbeh.2006.03.034
Pindyck, R., & Rubinfeld, D. (2008). Microeconomics. Upper Saddle River, New Jersey: Pearson Education, Inc.
Read, D. (2004). Intertemporal choice. In D. Koehler & N. Harvey (Eds.), Blackwell handbook of judgment and decision making. Oxford: Blackwell.
Reimers, S., Maylor, E. A., Stewart, N., & Chater, N. (2009). Associations between a one-shot delay discounting measure and age, income education and real-world impulsive behavior. Personality and Individual Differences, 47, 973-978. doi: 10.1016/j.paid.2009.07.026
Reynolds, B., Patak, M., & Shroff, P. (2007). Adolescent smokers rate delayed rewards as less certain than adolescent nonsmokers. Drug and Alcohol Dependence, 90(2-3), 301-303. doi: 10.1016/j.drugalcdep.2007.04.008
Samuelson, P. (1937). A note on measurement of utility. Review of Economic Studies, 4, 155-161. doi: 10.2307/2967612
Shelley, M. K. (1994). Gain/loss asymmetry in risky intertemporal choice. Organizational Behavior and Human Decision Processes, 59(1), 124-159. doi: 10.1006/obhd.1994.1053
Takahashi, T., Ikeda, K., & Hasegawa, T. (2007). A hyperbolic decay of subjective probability of obtaining delayed rewards. Behavioral and Brain Functions, 3, 52. doi: 10.1186/1744-9081-3-52
Thaler, R. (1981). Some empirical evidence on dynamic inconsistency. Economics Letters, 8, 201-207. doi: 10.1016/0165-1765(81)90067-7
Non-parametric correlations (Spearman's rho) between discount rates and various demographic factors.
discounting of gains
discounting of losses
employed part time
discounting of gains
discounting of losses
employed part time
† indicates p < .10, * indicates p < .05, ** indicates p < .01
Mean discount rates when future outcomes are certain versus uncertain, in Study 1. Error bars indicate +/- one standard error.
Mean discount rates when immediate outcomes are certain versus uncertain, in Study 2. Error bars indicate +/- one standard error.
Risk preference for immediate gains and losses, in Study 2. Error bars indicate +/- one standard error.
Mean discount rates when immediate and future outcomes are certain versus uncertain, in Study 3. Error bars indicate +/- one standard error.
Comparison of discount rates for uncertain outcomes. The left hand side shows the average of discount rates when uncertainty is in the future (Study 1) or uncertainty is immediate (Study 2). The right shows discount rates when both immediate and future outcomes are uncertain (Study 3). Error bars indicate +/- one standard error.
1 Throughout the manuscript, we use the terms "time preference" and "discounting" to refer to the overall preference for when something may occur or how much something is worth at a given delay, respectively. Therefore, observed time preferences and discount rates are the end result of multiple factors. We will not address the highly specific "pure time preference" for utility that is sometimes indicated in economic papers by the use of the terms "time preference" and "discounting." As noted by Read (2004) and others, the measurement of pure time preference relies on a number of key assumptions that are routinely violated in experimental investigations.