Promoting healthy eating through advice-giving: an experimental investigation (with Arthur E. Attema & Werner B.F. Brouwer) [pdf]
Abstract: Construal level theory (CLT) suggests that decision made at greater psychological distance induces rationality or “should-oriented” reasoning. We introduce such distance by having participants advise others before choosing for themselves and test whether this shift of perspective influences their own subsequent food choices. We conducted a laboratory experiment with 182 university students, where the treatment group was asked to give advice to another participant on snack food before making their own choices, while the control group directly made their own choices. We found that the intervention significantly increased the probability of choosing healthy food for themselves by around 19.8 percentage points. This effect was especially marked among women and among individuals who were not hungry. We also measured the time preference for money, healthy food, and unhealthy food, and found the treatment was more effective for individuals who were more patient with unhealthy food than healthy food. The main result reflects that placing oneself in an other-oriented mindset, i.e. thinking about advising others before making one’s own decisions, could benefit their own subsequent decisions. The comparatively large effect indicates that advice-first prompt is a low-cost promising strategy to promote healthy eating.
Time and lexicographic preferences in the valuation of EQ-5D-Y with time trade-off methodology (with Stefan A. Lipman, Koonal K. Shah & Arthur E. Attema) [Link] The European Journal of Health Economics, 24(2), 293‑305.
Abstract: In the valuation of EQ-5D-Y-3L, adult respondents are asked to complete composite time trade-off (cTTO) tasks for a 10-year-old child. Earlier work has shown that cTTO utilities elicited in such a child perspective are generally higher than when adults take their own perspective. We explore how differences in time preference in child and adult perspectives could explain this effect. Furthermore, as cTTO valuation in a child perspective involves explicit consideration of immediate death for a child, we also consider how cTTO utilities could be affected by decision-makers lexicographically avoiding death in children. We report the results of an experiment in which 219 respondents valued 5 health states in both adult and child perspectives with either a standard cTTO or a lead-time TTO only approach, in which immediate death is less focal. Time preferences were measured in both perspectives. Our results suggest that utilities were lower when lead-time TTO, rather than cTTO, was used. We find large heterogeneity in time preference in both perspectives, with predominantly negative time preference. The influence of time preferences on utilities, however, was small, and correcting for time preferences did not reduce differences between utilities elicited in both perspectives. Surprisingly, we found more evidence for differences in utilities between adult and child perspectives when lead-time TTO was used. Overall, these results suggest that time and lexicographic preferences affect time trade-off valuation in child and adult perspectives, but are not the explanation for differences between these perspectives. We discuss the implications of our findings for EQ-5D-Y-3L valuation.
Is episodic future thinking effective in mitigating the influence of time preference in time trade-off? (co-first authored with Zhongyu Lang, and with Stefan A. Lipman, Bradely Sugden, Kim Rand & Arthur E. Attema) [Link] The European Journal of Health Economics (2025)
Abstract
Objectives: The composite time trade-off (cTTO) method has been found to be influenced by time preferences for future life years, which typically results in a downward bias on cTTO utilities without adjustment. Contrary to prior research that adjusted for this distortion ex-post, this study takes an ex-ante approach, using Episodic Future Thinking (EFT), to potentially prevent time preference distortion. We aim to investigate the effect of EFT on time preference and cTTO utilities compared to using alternative methods.
Methods: A total of 150 participants from the UK general public were recruited for interviewer-led online interviews and randomly assigned to either the control or treatment group. In the control group, they were asked to recall recent memories using the Episodic Recent Thinking (ERT) protocol, serving as filler tasks; in the treatment group, they were asked to imagine life in the next 10 to 20 years, i.e. using an EFT protocol. Afterwards, respondents were asked to value seven EQ-5D-5L health states with cTTO tasks, followed by a nonparametric method to measure time preference.
Results: We observed a similar pattern of time preference across the two groups, with the majority discounting positively. EFT did not significantly affect time preference. In addition, the difference between cTTO utilities mitigated by EFT and those adjusted using the ex-post approach for time preference is minimal.
Conclusions: In conclusion, EFT does not seem to mitigate time preference for life years and has negligible effect on cTTO utilities, necessitating alternative strategies for reducing bias in health utilities.
Minimum Legal Drinking Age and Adolescent Alcohol Use: Evidence from the Netherlands (with Elisa de Weerd, Hans van Kippersluis, John Cawley, Ruud Roodbeen, Karin Monshouwer, & Saskia van Dorsselaer) [Slides]
Abstract: A common strategy to limit adolescent alcohol consumption is to impose a minimum legal drinking age (MLDA). This paper examines the effect of the MLDA in the Netherlands, combining bi-annual self-reported survey data from school-aged children with data on vendor compliance and administrative data on alcohol-related hospital visits from Dutch administrative data. The Netherlands provides an interesting context to reassess the effectiveness of MLDA given relatively low demand and supply side compliance, and a raise in the MLDA from 16 to 18 in the middle of our observation period. Although adolescent alcohol consumption has decreased over time, the policy change was not followed by an immediate decline in alcohol consumption among the age group directly affected by the policy change. Using regression discontinuity analyses at age 16, we find evidence that the reform was associated with a shift in parental attitudes and perceived access to alcohol, yet do not find clear evidence of discontinuities in adolescent alcohol consumption, either before or after the reform. Heterogeneity analyses indicate that, pre-reform, the MLDA only reduced drinking among adolescents around the age-threshold in areas with relatively high vendor compliance and limited cross-border access. These findings contrast with prior evidence from other countries on the effectiveness of a MLDA, highlighting the importance of enforcement and broader social norms in influencing policy outcomes.
A head-to-head comparison of methods for measuring time preference (with Arthur E. Attema & Stefan A. Lipman)
Abstract
Objectives: This study aims to compare the performance of various time preference elicitation methods between the monetary and health domains. Specifically, we evaluated their feasibility, internal validity, and concurrent validity within each domain, and predictive validity of these methods on health behaviors. Additionally, we are interested in detecting which methods can potentially useful that were not yet applied to health.
Methods: We identified five methods based on a ranking of methodological and theoretical importance for a head-to-head comparison in measuring time preferences (i) the direct method (Attema et al., 2012), (ii) the non-parametric method (Abdellaoui et al., 2016; Lipman et al., 2019), (iii) the onset of disease method (Attema and Versteegh, 2013), (iv) patience auctions (Olivola and Wang, 2016), and (v) the monetary choice questionnaire (Kirby and Maraković, 1995). Methods (i), (ii), and (iii) are health-framed, while methods (iv) and (v) are non-health methods. A total of 336 subjects participated in an experiment using these methods. In the experiment, participants completed tasks for all methods within a within-subject design, while domain framing (health or monetary) was randomized between subjects. The tasks were identical in structure but differed in framing, except for method (iii), which relies on time trade-off in health state valuation and is hardly adjusted to a monetary version. After completing the time preference tasks, participants answered questions about health behaviors. Measurements were taken at two time points (T1 and T2) with a two-week interval to assess test-retest reliability.
Results: we found that the direct method, non-parametric method and patience auction method yielded similar time preference estimates within each domain, with average AUC values between 0.51 and 0.57. This indicates that most people exhibited positive time preferences, bhough a significant proportion (20%-48%) showed negative time preferences, particularly with the non-parametric method. The predictive validity of time preferences on health behaviors was limited. There was no consensus on the magnitude of time preferences between the health and monetary domains across methods. Test-retest reliability was relatively low for the direct method, non-parametric method, and patience auctions, whereas the monetary choice questionnaire demonstrated good reliability.
Discussion: Evaluating the feasibility of measuring time preference with different methods is challenging. Different measures, in general, yield relatively similar conclusions about discounting. However, the internal validity of the selected methods was poor. Comparing time preferences across health and monetary domains also offered little insight. Extending non-health time preference measurement methods to the health domain did not reveal significant differences in time preferences between the monetary and health domains. Besides, these time preference measures demonstrated limited applicability for predicting health behaviors.
Adolescent alcohol consumption did not additionally decrease after the minimum legal drinking age was raised (The article is in Dutch, "Alcoholconsumptie jongeren niet extra afgenomen na verhoging minimumleeftijd") (with Elisa de Weerd, Hans van Kipperluis & Ruud Roodbeen) [Link]
ESB Economisch Statistische Berichten, 109 (4839), 491-493.