Abstract
By using a panel survey of Japanese adults, we show that smoking behavior is associated with personal time discounting and its biases, such as hyperbolic discounting and the sign effect, in the way that theory predicts: smoking depends positively on the discount rate and the degree of hyperbolic discounting and negatively on the presence of the sign effect. Positive effects of hyperbolic discounting on smoking are salient for naïve people, who are not aware of their self-control problem. By estimating smoking participation and smokers’ cigarette consumption in Cragg’s two-part model, we find that the two smoking decisions depend on different sets of time-discounting variables. Particularly, smoking participation is affected by being a naïve hyperbolic discounter, whereas the discount rate, the presence of the sign effect, and a hyperbolic discounting proxy constructed from sign effect behavior vis-à-vis doing homework assignments affect both types of decision making. The panel data enable us to analyze the over-time instability of elicited discount rates. The instability is shown to come from measurement errors, rather than preference shocks on time preference. Several evidences indicate that the detected associations between time preferences and smoking behavior are inter-personal one, rather than within-personal one.
The original article first appeared in the Health Economics 23(12):1443–1464, 2014. A newly written addendum has been added to this book chapter.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Ikeda et al. (2010) construct a similar proxy of hyperbolic discounting and detect a positive association between it and the degree of obesity. In Ikeda and Kang (2015), we adopt the same idea to identify whether a respondent is naïve or sophisticated, and thereby show that a naïve, hyperbolic discounter is more likely to be a debtor than an exponential discounter, whereas a sophisticated hyperbolic respondent is as likely to be a debtor as an exponential respondent.
- 2.
With regard to the assertion that there are higher discount rates among addicts, see, for the issue of smoking, Mitchell (1999), Odum et al. (2002), Bickel et al. (1999), Baker et al. (2003), Ohmura et al. (2005), Reynolds et al. (2004) and Ida and Goto (2009); for the issue of drug abuse, Madden et al. (1997) and Kilby et al. (1999). With regard to the prevalence of hyperbolic discounting among addicts, see Madden et al. (1999) for heroin users and Bickel et al. (1999), Odum et al. (2002), and Ida (2010) for smokers. See also Blondel et al. (2007), a study whose findings show that there is no difference between the discount rates of drug users and drug nonusers if risk attitudes are controlled for.
- 3.
- 4.
- 5.
Theoretically, it is difficult to show analytically the precise effect of hyperbolic discounting on smoking behavior by obtaining a closed-form solution in a dynamic optimization framework. However, Gruber and Kőszegi (2004) verify, using a quadratic utility model, that naïve hyperbolic discounters have a higher marginal propensity to smoke than sophisticated hyperbolic discounters, and that sophisticated hyperbolic discounters in turn display a higher smoking propensity than exponential discounters. Laibson (1997, 1998) shows that under certain conditions, the effective discount rate obtained by transforming the hyperbolic discounting function into the exponential one is greater than the pure exponential discount rate. This implies that persons with hyperbolic discounting tend to smoke more than those with exponential discounting.
- 6.
In the economics literature, there have been two attempts to incorporate the sign effect into economic models. First, Wakai (2008, 2011) provide a utility-smoothing model in which future felicity is discounted at a lower rate when it is smaller than the current utility level (i.e., when felicity is going to decrease in the future) than when it is larger than the current utility level (i.e., when felicity is going to increase in the future). Second, Loewenstein and Prelec (1992) propose a property of loss amplification, by which the rate of change in the values of gains is perceived as smaller than that in the values of equivalent losses.
- 7.
The JHPS was initiated in 2004 as a project of the Osaka University COE program and continues as a project of the Osaka University Global COE program, both of which are supported by the Ministry of Education, Culture, Sports, Science and Technology.
- 8.
We exclude from our sample the data from 2004, for two reasons. Unlike in 2005–2008, the queries to elicit discount rates in the 2004 survey were asked in a matching form in which respondents were asked to write down equivalent amounts of present-day money to a given amount of future money. The resulting discount rate data are considered to contain large measurement errors. The descriptive statistics of elicited discount rates in the 2004 survey indeed differ from those in other years. In addition to the choice conditions of the discounting, the queries also differ from those in other years.
- 9.
These trends are consistent with the reported data of the National Survey of Health and Nutrition (NSHN) conducted by the Ministry of Health, Labour and Welfare of Japan. According to NSHN data from 2004 to 2005, the number of cigarettes consumed by male smokers decreased from 21.5 to 21.0 per day, and the rates of regular smoking declined from 43.3 to 39.3 %. For females, the number of cigarettes consumed increased from 14.6 to 15.6 per day, while their rates of regular smoking decreased from 12.0 to 11.3 %.
- 10.
In contrast to some experimental studies (e.g., Harrison et al. 2002), the current study is based on the non-incentivized hypothetical question survey. We have a great deal of evidence that there is no systematic difference between discount rates estimated from real and hypothetical monetary reward choices (e.g., Baker et al. 2003; Johnson and Bickel 2002; Simpson and Vuchinich 2000). See also Khwaja et al. (2007), Grignon (2009), and Ikeda et al. (2010) for studies that use hypothetical choice surveys to elicit discount rates. However, we could be still skeptical of the usefulness of responses to our limited hypothetical questions in explaining smoking behavior. We shall discuss on within-person variations in our response data in Sect. 3.3, and on measurement errors in Sect. 5.
- 11.
As in the literature (e.g., Harrison et al. 2002), respondents who displayed multi-switching points are omitted from the sample. Respondents who left an option unselected are also omitted.
- 12.
It is well known that the discount rates differ wildly depending on various choice conditions. See, e.g., Frederick et al. (2002), which summarizes the elicited discount rates reported in the vast literature. It is also known that discount rates elicited from experiments and questionnaire surveys are much higher than market interest rates. See Frederick et al. (2002).
- 13.
It is not unusual for the discount rate for future payments (or losses) to be negative. Indeed, many subjects in experiments are reported to prefer to incur a loss immediately rather than delay it (e.g., Benzion et al. 1989; Chapman 1996). Average discount rates for losses are sometimes reported as negative (Loewenstein 1987; Chapman 1996; Ganiats et al. 2000). Wakai (2011) shows theoretically that a negative discount rate for deteriorating future felicity causes people to strongly dislike future volatility, so that intertemporal preferences become non-monotonic. Our result that average respondents prefer earlier repayment with a negative interest rate is consistent with this kind of non-monotonicity of intertemporal preferences.
- 14.
Although the standardized average DISCRATE of the elicited discount rates should theoretically satisfy E (DISCRATE) = 0 and σ (DISCRATE) = 1, neither equality is actually met, as seen in Table 9.4. This comes from the fact that the numbers of effective responses differ among the five discount rate questions.
- 15.
In addition, although we have not included the results of the t-test in the table, DR3, the discount rate for JPY 10,000 (around USD 93.32) is significantly higher than DR4, which applies for JPY 1 million (around USD 9,331.84). This implies that people are more patient for larger amounts than for smaller amounts. This tendency, called the magnitude effect, is also commonly observed in the literature (e.g., Benzion et al. 1989; Frederick et al. 2002).
- 16.
In Japanese elementary and high schools, students are usually assigned a great deal of homework during summer vacation.
- 17.
Options for QA and QB do not have any categories to capture behavior to do homework late or not to do it at all. As another problem, responses may capture other respondent attributes than an inclination toward procrastination (e.g., their upbringing by parents, availability of other options, or any social connections). To partially take this possibility into account, we conducted another main regression analysis by controlling for the education levels of parents, and we found our main results to be robust, even against this consideration. Furthermore, in spite of the above-mentioned limitations, HYPERBOLPROXY and the interaction with NAÏVE (HYPERBOLPROXY*NAÏVE) obtained from QA and QB have strong correlations with respondents’ degrees of obesity, inclination toward over-borrowing, and addictive behavior other than smoking (e.g., drinking and gambling), all of which are predicted by economic theory to be affected by peoples’ present-biased tendency. See Appendix, which shows the correlations of HYPERBOLPROXY and HYPERBOLPROXY*NAÏVE with the actual current behavior.
- 18.
Note that throughout the sample period, JHPS asked the same set of questions regarding time discounting.
- 19.
We also construct the variables that are time-averaged responses to QA instead of HYPERBOLPROXY, and that which are constructed by taking time-averaged differences between responses to QA and QB instead of NAÏVE. With these alternative specifications, the main results in the present study do not change substantially. The estimation results are available upon request.
- 20.
This model is also called the ‘double-hurdle model’ or ‘two-tier model.’
- 21.
For robustness checks, we estimate five alternative specifications by using both the pooled regression and random-effects models: (i) the binary probit model, (ii) the linear regression model, (iii) the Tobit model, (iv) the ordered probit model, and (v) the interval data regression model. The results are fairly consistent with those in Table 9.7: the estimated impacts of time-preference variables on smoking behavior are confirmed in all specifications in the same manner as in the Cragg model. These results are available upon request.
- 22.
The results of one-wave estimations are available upon request.
- 23.
The cases of taking 0.75 and 1 as the threshold values are also estimated, wherein our results are robust against the modifications. The results are available upon request.
- 24.
However, in the consistent sample, the negative impacts of SIGN become weaker, possibly due to the sample reduction.
- 25.
There could be several reasons for the measurement errors. First, our data are not based on incentivized experiments. For this point, see footnote 10. Second, our questions to elicit discount rates might be somewhat demanding for respondents. However, we do not think that this possibility is so serious for the following reasons. First, our average respondents replied to the same time-discounting questions in 2.5 waves. It would not be so difficult for them to reply to the questions. Second, in the 2011 wave, more than 80 % of respondents made consistent choices even though options in payoff tables (e.g., Table 9.2) were arranged in more complex manners, i.e., in random orders without listing imputed interest rates, instead of the order in accordance with the listed value of imputed interest rates as in Table 9.2. Third, our time discounting data display associations with various behavioral attributes, such as the degree of obesity, debt holdings, and habits of gamble and drinking, in theoretically predicted ways. For the same reasons, we do not guess that the measurement errors in our data are larger than in computer-based studies in the literature where series of simple binary choices are posed stepwise.
- 26.
For example, the proportion of hyperbolic discounters in our dataset is 67.3 % (see Sect. 3.2.1), whereas hyperbolic discounters occupy 25 % of 3,200 Italians and 1,400 Dutch respondents (see Eisenhauer and Ventura 2006), 36 % of 606 Americans (see Meier and Sprenger 2010), and 44.9 % of 2,236 Japanese (see Ikeda and Kang 2015).
- 27.
Instead of using Kimball’s procedure, we could impute discount rates by simply assigning median interest rates of categories to their estimates. In that case, however, we cannot elicit discount rates when respondents do not switch at any given choices. Partly due to the resulting reduction of the sample size, estimation results become weaker when using the median interest rates. The Kimball method enables us to estimate discount rates even when responses do not switch.
- 28.
Since the 2010 JHPS wave does not make inquiries regarding behavioral inclinations for the homework assignment, we assume here that variables HYPERBOLPROXY and NAÏVE take the same values as in the 2009 wave.
- 29.
The relationships are robust even if the cut-off points of HYPERBOLPROXY and NAÏVE are arbitrarily changed.
- 30.
Time-invariant indicators make associations in Table 9.11 stronger: in the case of hyperbolic discounting, the t-values are changed to −5.480 for the mean difference between smokers and nonsmokers, and to −1.210 for the difference between heavy smokers and light smokers; and in the case of the sign effect, the former is changed to −3.627, and the latter to −1.843.
References
Ainslie G (1992) Picoeconomics. Cambridge University Press, Cambridge
Baker F, Johnson MW, Bickel WK (2003) Delay discounting in current and never-before cigarette smokers: similarities and differences across commodity, sign, and magnitude. J Abnorm Psychol 112(3):382–392
Bardsley P, Olekalns N (1999) Cigarette and tobacco consumption: have anti-smoking policies made a difference? Econ Rec 75(3):225–240
Becker GS, Murphy KM (1988) A theory of rational addiction. J Polit Econ 96(4):675–700
Becker GS, Grossman M, Murphy KM (1994) An empirical analysis of cigarette addiction. Am Econ Rev 84(3):396–418
Benzion U, Rapoport A, Yagil J (1989) Discount rates inferred from decisions: an experimental study. Manag Sci 35(3):270–284
Bickel WK, Odum AL, Madden GJ (1999) Impulsivity and cigarette smoking: delay discounting in current never, and ex-smokers. Psychopharmacology 146(4):447–454
Blondel S, Lohéac Y, Rinaudo S (2007) Rationality and drug use: an experimental approach. J Health Econ 26(3):643–658
Chaloupka F (1990) Men, women, and addiction: the case of cigarette smoking, NBER working paper 3267. National Bureau of Economic Research, Cambridge, MA
Chaloupka F (1991) Rational addictive behavior and cigarette smoking. J Polit Econ 99(4):675–700
Chapman G (1996) Temporal discounting and utility for health and money. J Exp Psychol Learn Mem Cogn 22(3):771–791
Cragg JG (1971) Some statistical models for limited dependent variables with application to the demand for durable goods. Econometrica 39(5):829–844
Eisenhauer J, Ventura L (2006) The prevalence of present bias: some european evidence. Appl Econ 38(11):1223–1234
Frederick S, Loewenstein G, O’Donoghue T (2002) Time discounting and time preference: a critical review. J Econ Lit 40(2):351–401
Ganiats G, Teodore RT, Carson RM, Hamm SB, Cantor W, Summer S, Spann J, Hagen M, Miller C (2000) Health status and preferences: population-based time preferences for future health outcome. Med Decis Mak 20(3):263–270
Grignon M (2009) An empirical investigation of heterogeneity in time preferences and smoking behaviors. J Socio-Econ 38(5):739–751
Gruber J, Kőszegi B (2001) Is addiction “rational”? Theory and evidence. Q J Econ 116(4):1261–1303
Gruber J, Kőszegi B (2004) Tax incidence when individuals are time-inconsistent: the case of cigarette excise taxes. J Public Econ 88(9–10):1959–1987
Gruber J, Mullainathan S (2005) Do cigarette taxes make smokers happier. Adv Econ Anal Policy 5(1):1–43
Harrison GW, Lau MI, Williams MB (2002) Estimating individual discount rates in Denmark: a field experiment. Am Econ Rev 92(5):1606–1617
Heidhues P, Kőszegi B (2010) Exploiting naïvete about self-control in the credit market. Am Econ Rev 100(5):2279–2303
Ida T (2010) Anomaly, impulsivity, and addiction. J Socio-Econ 39(2):194–203
Ida T, Goto R (2009) Simultaneous measurement of time and risk preferences: stated preference discrete choice modeling analysis depending on smoking behavior. Int Econ Rev 50(4):1169–1182
Ikeda S, Kang M-I (2015) Hyperbolic discounting, borrowing aversion, and debt holding. Japanese Economic Review. doi:10.1111/jere.12072
Ikeda S, Kang M-I, Ohtake F (2010) Hyperbolic discounting, the sign effect, and the body mass index. J Health Econ 29(2):268–284
Johnson MW, Bickel WK (2002) Within-subject comparison of real and hypothetical money rewards in delay discounting. J Exp Anal Behav 77(2):129–146
Kang M-I, Ikeda S (2013) Time discounting, present biases, and health-related behavior, ISER discussion paper 885. ISER, Osaka
Kang M-I, Ikeda S (2014) Time discounting and smoking behavior: evidence from a panel survey. Health Econ 23(12):1443–1464
Keeler TE, Hu T-W, Barnett BG, Manning WG (1993) Taxation, regulation, and addiction: a demand function for cigarettes based on time-series evidence. J Health Econ 12(1):1–18
Khwaja A, Silverman D, Sloan F (2007) Time preference, time discounting, and smoking decisions. J Health Econ 26(5):927–979
Kilby KN, Petry NM, Warren WK (1999) Heroin addicts have higher discount rates for delayed rewards than non-drug-using controls. J Exp Psychol 128(1):78–87
Kimball MS, Sahm CR, Shapiro MD (2008) Imputing risk tolerance from survey responses. J Am Stat Assoc 103(483):1028–1038
Laibson D (1997) Golden eggs and hyperbolic discounting. Q J Econ 112(2):443–477
Laibson D (1998) Life-cycle consumption and hyperbolic discount functions. Eur Econ Rev 42(3–5):861–871
Loewenstein G (1987) Anticipation and the valuation of delayed consumption. Econ J 97(387):666–684
Loewenstein G, Prelec D (1992) Anomalies in intertemporal choice: evidence and an interpretation. Q J Econ 107(2):573–597
Luo F, Abdel-Ghany M, Ogawa I (2003) Cigarette smoking in Japan: examination of myopic and rational models of addictive behavior. J Fam Econ Iss 24(3):308–317
Madden GJ, Petry NM, Barger GJ, Bickel WK (1997) Impulsive and self-control choices in opioid-dependent patients and non-drug-using control patients: drug and monetary rewards. Exp Clin Psychopharmacol 5(3):256–262
Madden GJ, Bickel WK, Eric EA (1999) Discounting of delayed rewards in opioid-dependent outpatients: exponential or hyperbolic discounting functions? Exp Clin Psychopharmacol 7(3):284–293
Meier S, Sprenger C (2010) Present-biased preferences and credit card borrowing. Am Econ J Appl Econ 2(1):193–210
Mitchell SH (1999) Measures of impulsivity in cigarette smokers and non-smokers. Psychopharmacology 146(4):455–464
O’Donoghue T, Rabin M (1999) Doing it now or later. Am Econ Rev 89(1):103–124
Odum AL, Madden GJ, Bickel WK (2002) Discounting of delayed health gains and losses by current, never- and ex-smokers of cigarettes. Nicotin Tob Res 4(3):295–303
Ohmura Y, Takahashi T, Kitamura N (2005) Discounting delayed and probabilistic monetary gains and losses by smokers of cigarettes. Psychopharmacology 182(4):508–515
Reynolds B, Karraker K, Horn K, Richards JB (2004) Delay and probability discounting as related to different stages of adolescent smoking and non-smoking. Behav Process 64(3):333–344
Sato M, Ohkusa Y (2003) The relationship between smoking initiation and time discounting factor, risk aversion and information. Appl Econ Lett 10(5):287–289
Shoda Y, Mischel W, Peake PK (1990) Predicting adolescent cognitive and self-regulatory competencies from preschool delay of gratification: identifying diagnostic conditions. Dev Psychol 26(6):978–986
Simpson CA, Vuchinich RE (2000) Reliability of a measure of temporal discounting. Psychol Rec 50(1):3–16
Thaler R (1981) Some empirical evidence on dynamic inconsistency. Econ Lett 8(3):201–207
Wakai K (2008) A model of utility smoothing. Econometrica 76(1):137–153
Wakai K (2011) Modeling nonmonotone preferences: the case of utility smoothing. J Math Econ 47(2):213–226
Wan J (2006) Cigarette tax revenues and tobacco control in Japan. Appl Econ 38(14):1663–1675
Acknowledgment
Our special thanks go to Y. Fukuta, K. Hirata, M. Nakagawa, F. Ohtake, J. Wan, and participants at the International Workshop on the Economics of Obesity and Health 2009 for helpful comments. We acknowledge financial supports from: the COE and Global COE Programs of Osaka University; a Grant-in-Aid for Scientific Research (B 21330046) from the Japan Society for the Promotion of Science; and the Joint Usage/ Research Center Projects of ISER from the Ministry of Education, Culture, Sports, Science and Technology.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendices
Appendix
Table 9.10 shows the correlation coefficients of HYPERBOLPROXY and HYPERBOLPROXY*NAÏVE with actual behavioral traits.
Addendum: Validation Using Recent Wave Data in JHPS
This addendum has been newly written for this book chapter.
This addendum checks robustness of the results in the Kang and Ikeda (2014) text article. To do so, we use the most recent waves of JHPS in 2009 and 2010. These waves of data enable us to reconsider whether time-variations of preference data are attributed to individual preference shocks or possible measurement errors, and whether detected relationships between time preferences and smoking capture behavioral differences due to within-personal preference shocks or interpersonal differences of time preferences.Footnote 28
Table 9.11 shows that associations between time-discounting and smoking in the 2009 and 2010 data are consistent with those in the text article. All of the individual discount rates (DR1 to DR4), with the exception of the discount rate for paying money (DR5), and the impatience measure (DISCRATE) show significant and positive associations with smoking.
In the 2009 and 2010 waves, associations regarding behavioral biases of time discounting also show consistency with our previous findings, in that heavier smokers are less likely to exhibit the sign effect (SIGN), and more likely to exhibit tendencies toward procrastination (HYPERBOLPROXY) and time-inconsistency (NAÏVE).Footnote 29 In contrast, we find association between smoking and hyperbolic discounting regarding monetary choices (HYPERBOL) to be different from our hypothesis again.
As discussed in the text, time-varying indicators regarding monetary discounting biases, such as HYPERBOL and SIGN, might contain non-negligible measurement errors. Table 9.12 shows autocorrelations of DISCRATE, HYPERBOL, and SIGN in the 2009 and 2010 waves. In spite of the significant autocorrelations observed in case of the three variables, the reported magnitudes for HYPERBOL and SIGN are not large enough. The time-variations in the indicators are attributable to measurement errors. Indeed, associations between time-discounting biases and smoking become stronger when we use time-invariant indicators that take one only if a respondent reports the incident of the corresponding time-discounting biases in both waves.Footnote 30 This fact implies that, in time-variations of time preferences in JHPS, measurement errors dominate preference shocks, and therefore, in accordance with our assertion in the text article, the results in Table 9.11 reflect interpersonal associations between time discounting and smoking, rather than within-personal ones.
By using a titration-type questionnaire, which asks sequentially three queries of binary choices on immediate future and distant future trade-offs, we successfully detect expected associations between monetary hyperbolic discounting and several actual behaviors, such as health-related behavior including smoking in Kang and Ikeda (2013) and borrowing behavior in Ikeda and Kang (2015). Therefore, we now believe that unstable associations of monetary hyperbolic discounting in the text article might not reflect any true relations between hyperbolic discounting and smoking, but would rather reflect some shortcomings of the methodology in detecting hyperbolic discounting.
Rights and permissions
Copyright information
© 2016 Springer Japan
About this chapter
Cite this chapter
Kang, MI., Ikeda, S. (2016). Time Discounting and Smoking Behavior: Evidence from a Panel Survey. In: Ikeda, S., Kato, H., Ohtake, F., Tsutsui, Y. (eds) Behavioral Economics of Preferences, Choices, and Happiness. Springer, Tokyo. https://doi.org/10.1007/978-4-431-55402-8_9
Download citation
DOI: https://doi.org/10.1007/978-4-431-55402-8_9
Publisher Name: Springer, Tokyo
Print ISBN: 978-4-431-55401-1
Online ISBN: 978-4-431-55402-8
eBook Packages: Economics and FinanceEconomics and Finance (R0)