Experienced versus decision utility: large‐scale comparison for income–leisure preferences*

• Published date: 01 October 2023
• Author: Alpaslan Akay,Olivier B. Bargain,H. Xavier Jara
• Date: 01 October 2023
• DOI: http://doi.org/10.1111/sjoe.12538

Scand. J. of Economics 125(4), 823–859, 2023

DOI: 10.1111/sjoe.12538

Experienced versus decision

utility: large-scale comparison

for income–leisure preferences*

Alpaslan Akay†

University of Gothenburg, SE-40530 Gothenburg, Sweden

alpaslan.akay@gu.se

Olivier B. Bargain‡

Bordeaux School of Economics, FR-33608 Pessac, France

olivier.bargain@u-bordeaux.fr

H. Xavier Jara

London School of Economics and Political Science, London, WC2A 2AE, UK

h.x.jara-tamayo@lse.ac.uk

Abstract

Subjective well-being (SWB) data are increasingly used to perform welfare analysis. Interpreted

as “experienced utility”, it has recently been compared to “decision utility” using small-scale

experiments most often based on stated preferences. We transpose this comparison to the

framework of non-experimental and large-scale data commonly used for policy analysis,

focusing on the income–leisure domain where redistributive policies operate. Using the British

Household Panel Survey, we suggest a “deviation” measure, which is simply the difference

between actual working hours and SWB-maximizing hours. We show that about three-quarters

of individuals make decisions that are not inconsistent with maximizing their SWB. We discuss

the potential channels that explain the lack of optimization when deviations are signif‌icantly

large. We f‌ind proxies for a number of individual and external constraints, and show that

constraints alone can explain more than half of the deviations. In our context, deviations partly

ref‌lect the inability of the revealed preference approach to account for labor market rigidities,

so the actual and SWB-maximizing hours should be used in a complementary manner. The

suggested approach based on our deviation metric could help identify labor market frictions.

Keywords: Decision utility; experienced utility; labor supply; subjective well-being

JEL classif‌ication:C90; I31; J22

*A. Akay acknowledges visiting f‌inancial support from the COV-POP project, funded by

the Nouvelle Aquitaine region, which has allowed completing the revision of this paper. We

acknowledge f‌inancial support from the Chaire BEWELL and the GPR HOPE, funded by the

“Investments for the Future” program (PIA) IdEx Universit´

e de Bordeaux. We are grateful to

discussants and participants at the ASSA meeting, the RES Conference, and various seminars

(ISER, DIAL, LISER, AMSE, IZA, BSE) for useful suggestions. Any errors remain our own.

†Also aff‌iliated with Universidad Antonio de Nebrija, Madrid, Spain.

‡Also aff‌iliated with IZA.

c

2023 The Authors. The Scandinavian Journal of Economics published by John Wiley & Sons Ltd on behalf of F¨

oreningen

f¨

or utgivande av the SJE.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution

and reproduction in any medium, provided the original work is properly cited.

824 Experienced versus decision utility: income–leisure preferences

1. Introduction

In economics, the standard approach to measure well-being relies on the

observation of decisions made by supposedly rational (utility-maximizing)

agents. The object derived from the “revealed preference” approach is

sometimes referred to as a “decision utility”. For more than two decades,

some authors have claimed that this decision utility is not always consistent

with the well-being associated with different experiences. They recommend

developing measures that focus more directly on “experienced utility”

(e.g., Dolan and Kahneman, 2008), such as self-reported information

on happiness, life satisfaction, or mental health. A growing amount of

evidence has shown that such subjective well-being (SWB) information

is not pure statistical noise: it ref‌lects some individual heterogeneity that

is closely associated with objective measures of well-being and, to some

extent, with behavior.1Yet, SWB is still seen by many as one argument,

among others, in the grand utility function of an individual (Rayo and

Becker, 2007; Benjamin et al., 2012; Glaeser et al., 2016).2Other studies

postulate that SWB answers, commonly provided in survey questionnaires,

are consistent with people’s revealed preferences (Oswald and Wu, 2010;

Decancq et al., 2015).

Whether there is congruence between individual decisions and the SWB

derived from these choices is still an open question. This is especially

disputable for key economic decisions (such as labor supply), which

imply a trade-off between several important dimensions of a good life

(e.g., consumption versus leisure). On the one hand, observed choices may

ref‌lect heuristics, optimization errors, or the fact that people have imperfect

information about what is good for them. Choices are also potentially limited

by many personal constraints (e.g., family obligations) and external factors

(e.g., market imperfections), the importance of which is diff‌icult to assess in

welfare analyses. On the other hand, SWB may not encompass the totality

of what humans are trying to achieve when they make decisions. Individual

choices may ref‌lect other life goals (e.g., fame) or values (e.g., helping others)

that partly differ from, or sometimes conf‌lict with, the pursuit of well-being

as we measure it in subjective data. Despite these sources of discrepancy, it

seems crucial to test whether there is (at least) minimal consistency between

decision utility and experienced utility.

1See Krueger and Schkade (2008) and Oswald and Wu (2010), as well as critical reviews

in Senik (2005), Clark et al. (2008), Kahneman and Krueger (2006), and Fleurbaey and

Blanchet (2013).

2K¨

oszegi and Rabin (2008) argue that both subjective and choice-based measures of well-being

contain unique information on a person’s true welfare, so that the ideal measure should perhaps

combine both types of data.

c

2023 The Authors. The Scandinavian Journal of Economics published by John Wiley & Sons Ltd on behalf of F¨

oreningen

f¨

or utgivande av the SJE.

A. Akay, O. B. Bargain, and H. X. Jara 825

This paper proposes a tangible approach to address this question

in the context of labor supply decisions. Rather than confronting the

ordinal preferences consistent with a decision-based versus experience-based

welfare metric,3we directly compare actual working hours (consistent

with decision-utility maximization) and optimal working hours (from the

perspective of experienced utility; i.e., hours that maximize income–leisure

satisfaction). The comparison is done on a large scale using nationally

representative data, from the British Household Panel Survey (BHPS). We

necessarily focus on single people, because the joint decision in couples is

diff‌icult to apprehend for individual welfare comparisons. Our single-value

“deviation” metric is a practical and convenient representation of the potential

discrepancies between decision and experienced utilities, which can be used

for inference and for exploring the determinants of these discrepancies – here

in the context of labor supply or, more generally, in analyses that traditionally

rely on the revealed preferences approach.

The suggested procedure goes as follows. We start by calculating the

distribution of deviations in the sample. To this end, we combine income and

leisure satisfaction domains to construct a proxy of experienced utility in the

income–leisure domain. We use this SWB measure to estimate an experienced

utility function, adopting a f‌lexible approach borrowed from the labor supply

literature. Using the estimated parameters and discretized income–leisure

bundles, we numerically search for the amount of working hours that would

maximize experienced utility and compare them with actual choices. We

f‌ind a broad overlap between actual work hours and SWB-maximizing work

duration. The average deviation is close to zero (−2.9 hours). The negative sign

implies that people “overwork” on average according to SWB maximization,

but the deviation is not signif‌icantly different from zero for 72 percent of the

individuals in the sample. In other words, for a majority of people, actual

decisions are not inconsistent with the maximization of their income–leisure

satisfaction. We then attempt to describe the large discrepancies observed for

specif‌ic population subgroups (e.g., by gender, family composition, region

of residence, etc.), either characterized as “overworking” (a negative mean

deviation) or “underworking” (a positive mean deviation) from an SWB

perspective. Results suggest intuitive patterns regarding the direction of

the deviations. For instance, those living in London signif‌icantly overwork

(suggesting social norms or labor market constraints) while those with children

tend to work too little (suggesting childcare constraints or labor contracts that

are not f‌lexible enough). A detailed analysis by levels of worked hours

suggests that signif‌icant deviations are primarily due to those at the two ends

3This alternative approach is used in Akay et al. (2020) where money metrics are derived from

ordinal preferences consistent with either decisions or subjective experience.

c

2023 The Authors. The Scandinavian Journal of Economics published by John Wiley & Sons Ltd on behalf of F¨

oreningen

f¨

or utgivande av the SJE.