Time preference, time discounting, and smoking decisions

Abstract

This study examines the relationship between time discounting, other sources of time preference, and choices about smoking. Using a survey fielded for our analysis, we elicit rates of time discount from choices in financial and health domains. We also examine the relationship between other determinants of time preference and smoking status. We find very high rates of time discount in the financial realm for a horizon of 1 year, irrespective of smoking status. In the health domain, the implied rates of time discount decline with the length of the time delay (hyperbolic discounting) and the sign of the payoff (the sign effect). We use a series of questions about the willingness to undergo a colonoscopy to elicit short- and long-run rates of discount in a quasi-hyperbolic discounting framework, finding no evidence that short-run and long-run rates of discount differ by smoking status. Using more general measures of time preference, i.e., impulsivity and length of financial planning horizon, smokers are more impatient. However, neither of these measures is significantly correlated with the measures of time discounting. Our results indicate that subjective rates of time discount revealed through committed choice scenarios are not related to differences in smoking behavior. Rather, a combination of more general measures of time preference and self-control, i.e., impulsivity and financial planning, are more closely related to the smoking decision.

Introduction

The recent economics literature on anomalies of intertemporal choice and self-control has been focused on alternatives to the standard assumption of exponential time discounting. Building on Strotz (1956) and Phelps and Pollack (1968), research by Laibson (1997) and O’Donoghue and Rabin (1999) sparked a large new literature that explores the consequences of quasi-hyperbolic discounting in many areas, including savings behavior, labor, environmental, and health economics, and corporate finance. Models of hyperbolic discounting have often augmented or replaced the insights derived from standard models, rationalized puzzling behaviors, and generated new testable predictions. Importantly, certain welfare consequences of government policy have been shown to depend critically on whether consumers are quasi-hyperbolic rather than standard, exponential discounters (O’Donoghue and Rabin, 2005). For example, if agents are hyperbolic discounters, a measure of the welfare benefits of an increased tax on cigarettes may be greatly magnified because problems of self-control induce net costs of smoking that are internal to the smoker (Gruber and Köszegi, 2001, Gruber and Köszegi, 2004, Sloan et al., 2004).

The literature on hyperbolic discounting has grown rapidly, in part, because experimental evidence of hyperbolic discounting is voluminous. When choosing now to commit to present or future intertemporal tradeoffs, individuals commonly reveal declining (hyperbolic) rates of time discount. As Strotz (1956) first demonstrated, declining rates of time discount imply time-inconsistency and problems of self-control when choices are uncommitted. There are, however, other potential sources of time preference and problems of self-control not reflected in time discount functions and thus in the committed choices of individual decision makers. In models of dual selves (or cognitive processes) and of temptation costs, for example, present-biased time preference, tastes for commitment, and apparent time-inconsistency may emerge for reasons unrelated to the time discount function.¹ Recent empirical work is also consistent with the idea that substantially improving models of intertemporal choice may require more than the proper calibration of the discount function (Bernheim et al., 2001, Ameriks et al., 2003, Ameriks et al., 2004).

In this study, using data collected for our research, we examine the relationship between time discounting, other sources of time preference, and choices about smoking. The decision to smoke represents an interremporal tradeoff with substantial implications for individual and social welfare, and thus provides a natural context in which to study these issues. We elicit rates of time discount from choices in both financial and health domains. We also examine the relationship between other determinants of time preference and smoking status. Specifically, we employ a proxy for respondents’ degree of self-control using measures of impulsivity in individual behaviors. We investigate whether these and other measures of self-control differ by smoking status and to what extent these measures are correlated with subjective rates of time discount.

Using standard questions regarding committed intertemporal choices in the financial and health domains, we find patterns consistent with previous research on subjective rates of time discount. Responses to now standard hypothetical choice scenarios reveal very high rates of time discount in the financial realm for a horizon of 1 year, irrespective of smoking status. Consistent with previous research, we find evidence that these implied rates of time discount decline with the length of the time delay, (hyperbolic discounting), and depend on the sign of the payoff (the “sign” effect). We also find an effect of the size of stakes on choices for gains but not for losses (the “magnitude” effect). Further, we use a series of questions about the willingness to undergo a colonoscopy, a procedure recommended for all persons in the age group of respondents to our survey, irrespective of smoking status, to elicit short- and long-run rates of discount in a quasi-hyperbolic discounting framework. We find no evidence that, for adults, short-run and long-run rates of discount differ by smoking status.

While several measures of time discounting in our data replicate patterns seen consistently in the literature, there is no correlation between these measures and smoking status. Our findings thus indicate that variation in time discounting is not a driving force behind differences in smoking behavior.²

However, we find that measures of impulsivity and the length of the financial planning horizon are related to smoking decisions. Those who are more impulsive are also significantly more likely to have smoked or to continue to smoke.³ Similarly, current smokers tend to have shorter financial planning horizons, even conditional on their longevity expectations. Neither of these measures of planning or self-control is significantly correlated with the standard measures of time discounting. We find that subjective rates of time discount revealed through committed choice scenarios are unrelated to smoking status. Rather a combination of more general measures of self-control, impulsivity and financial planning, is related to the smoking decision.

This paper proceeds as follows. The next section describes our data source and shows how the smokers in our sample exhibit many of the characteristics and behaviors noted previously in the literature. Section 3 presents our methods and results regarding subjective rates of time discount in the financial and health domains, and their relationship to smoking decisions. This section ends with analysis of alternative sources of time preference – impulsivity and length of planning horizon. Section 4 concludes.