Session 4 l1

Relative Income, Happiness, and Utility: An Explanation for the Easterlin Paradox and
Other Puzzles
Author(s): Andrew E. Clark, Paul Frijters and Michael A. Shields
Source: Journal of Economic Literature , Mar., 2008, Vol. 46, No. 1 (Mar., 2008), pp. 95-
144
Published by: American Economic Association
Stable URL: http://www.jstor.com/stable/27646948
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic
Literature
This content downloaded from

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
All use subject to https://about.jstor.org/terms
Journal of Economie Literature 2008, 46:1, 95-144
http:www.aeaweb.org/articles.php?doi=10.1257/jel.46.1.95
Relative Income, Happiness, and Utility:

An Explanation for the
Easterlin Paradox and Other Puzzles
Andrew E. Clark, Paul Frijters, and Michael A. Shields
The well-known Easterlin paradox points out that average happiness has remained
constant over time despite sharp rises in GNP per head. At the same time, a micro
literature has typically found positive correlations between individual income and
individual measures of subjective well-being. This paper suggests that these two find
ings are consistent with the presence of relative income terms in the utility function.
Income may be evaluated relative to others (social comparison) or to oneself in the
past (habituation). We review the evidence on relative income from the subjective
well-being literature. We also discuss the relation (or not) between happiness and
utility, and discuss some nonhappiness research (behavioral, experimental, neu
rological) related to income comparisons. We last consider how relative income in
the utility function can affect economic models of behavior in the domains of con
sumption, investment, economic growth, savings, taxation, labor supply, wages, and
migration.
Every pitifulest whipster that walks within a skin has had his head filled with the
notion that he is, shall be, or by all human and divine laws ought to be, "happy."
Thomas Carlyle
1. Income, Happiness, and the catalysts in the literature on income and hap
Easterlin Paradox piness has been Richard A. Easterlin's semi
nal article (1974, updated in 1995), setting
out the "paradox" of substantial real income
Studying thebecome
happiness has causes andhot correlates
one of the topics of human growth in Western countries over the last
in economics over the last decade, with both fifty years but without any corresponding rise
the size and depth of the literature increas in reported happiness levels. Similar studies
ing at an exponential rate (Daniel Kahneman
and Alan B. Krueger 2006). One of the main John Helliwell, Felicia Huppert, Danny Kahneman,
Brian Knutson, George Loewenstein, Sonja Lyubomirsky,
Dan Mroczek, Yew-Kwang Ng, Matthew Rabien, Larry
* Clark: Paris School of Economics and IZA. Frijters: Samuelson, David Schkade, Wolfram Schultz, Dylan
Queensland University of Technology. Shields: Univer Smith, Oded Stark, Arthur Stone, Stephen Wheatley
sity of Melbourne. We are grateful to the editor and two Price, Peter Ubel and Dan Wilson for invaluable advice.
anonymous referees for very constructive comments. We Andrew Oswald and Bernard van Praag provided espe
also thank Colin Camerer, Richie Davidson, Ed Diener, cially detailed suggestions. Frijters and Shields would like
Dick Easterlin, Ada Ferrer-i-Carbonell, Carol Graham, to thank the Australian Research Council for funding.
95

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
96 Journal of Economie Literature, Vol. XLVI (March 2008)
T 40,000
2.5
-h 30,000 o
o
o
C\|
CU
1.5 + 20,000
10,000
0.5
Happiness - - - Real income per capita
_l_I_I_I_I_I_L_ _l_I_I_I_l_ PH
1973 1977 1981 1985 1989 1993 1998 2003
Year
Figure 1. Happiness and Real Income Per Capita in the United States, 1973-2004
Source: World Database of Happiness and Penn World Tables. Happiness is the average reply to the following
question: "Taken all together, how would you say things are these days? Would you say that you are...?" The
responses are coded as (3) Very Happy, (2) Pretty Happy, and (1) Not too Happy. Happiness data are drawn
from the General Social Survey.
have also since been conducted by psy of little importance and should, therefore,
chologists (Ed Diener, Marissa Diener, and not be the primary goal of economic policy
Carol Diener 1995) and political scientists (Andrew J. Oswald 1997). Richard Layard
(Ronald Inglehart 1990). Figure 1 shows an (2005) goes as far as arguing that we need a
Easterlin graph for the United States over "revolution" in academia, where every social
the period 1973-2004. While real income scientist should be attempting to understand
per capita almost doubles, happiness (from the determinants of happiness, and it should
the General Social Survey) shows essentially be happiness which is the explicit aim of gov
no trend. From this figure, to borrow a term ernment intervention.1
from health economics, it looks as if indi This "paradox" is not specifically a U. S. phe
viduals in the United States are "flat of the nomenon. The same picture can be drawn for
curve," with additional income buying little Japan (Easterlin 1995), which has seen one of
if any extra happiness. It has been argued the largest increases in real per capita income
that once an individual rises above a poverty of any country since World War II, and also
line or "subsistence level," the main source for Europe. Figure 2 shows trends in average
of increased well-being is not income but life satisfaction for five European countries
rather friends and a good family life (see, for since 1973. As in the United States, there has
example, Robert E. Lane 2000). This "sub
sistence level" could be as low as US$10,000
1 It is interesting to note that this "modern" viewpoint
per annum (as reported in Bruno S. Frey and of the role of government in promoting happiness con
Alois Stutzer 2002b and Darrin M. McMahon trasts sharply with that of the ancient Greeks and much of
the world of antiquity (see McMahon 2006 for a history of
2006). Following on with this argument, the the philosophy of happiness). Erik Angner (2005) provides
radical implication for developed countries a fascinating account of the modern history of subjective
at least is that economic growth per se is well-being.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Clark, Frijters, and Shields: Relative Income, Happiness, and Utility 97
4 -,
3.5 -
3-
2.5
2 -
g 1.5 -
1 -
UK France ? ? - Germany < Italy Netherlands
0.5 -
.i i i i i i i ' ' '' I I I I I I
1973 1977 1980 1983 1986 1988 1990 1993 1996 2000 2004
Year
Figure 2. Life Satisfaction in Five European Countries, 1973-2004

Source: World Database of Happiness. Happiness is the average reply to the following question: "On the
whole how satisfied are you with the life you lead?" The responses are coded as (4) Very Satisfied, (3)
Fairly Satisfied, (2) Not Very Satisfied, and (1) Not at all Satisfied. Life satisfaction data are drawn from the
Eurobarometer Survey.
been no obvious increase in life satisfaction a large amount of evidence suggesting that
over a thirty-year period, even though real money does matter. There are three stylized
incomes per capita have increased sharply in facts in this second literature.
all five countries. The only trend found is in
Italy, the poorest country of the five, where 1) A regression of happiness on income using
average life satisfaction increased from 2.67 cross-section survey data from one country
in 1973 to 2.88 in 2004, a rise of 9.3 percent. (with or without standard demographic
Easterlin (2005b) provides a useful summary controls) generally produces a significant
of this macro empirical literature. positive estimated coefficient on income.
The same time-series data in transitional This holds for both developed (see, for
countries, however, suggest a larger role example, David G. Blanchflower and
for income. Consider figure 3, which shows Oswald 2004; Michael A. Shields and
average life satisfaction and real income in Stephen Wheatley Price 2005) and devel
East Germany during the decade following oping (Carol Graham and Stefano Pettinato
reunification. East Germans experienced a 2002; Orsolya Lelkes 2006) countries.
substantial increase in real income between However, the income-happiness slope is
1991 and 2002, and reported a considerable larger in developing or transition than in
rise in their life satisfaction over the same developed economies.
period.
However, we should be cautious in con 2) Recent work has used panel data to control
cluding from these graphs, which illustrate for unobserved individual fixed effects,
bivariate correlations, that income does not such as personality traits, and concludes
buy happiness in the developed world. A par that changes in real incomes are corre
allel body of work has produced what is now lated with changes in happiness (see, for

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
T 5,000
6.5
4,000 ?
i
0) 5.5
--3,000 13 I
2
%-. life satisfaction Real income
2,000
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
Figure 3. Life Satisfaction and Income in East Germany, 1991-2002

Source: Frijters et al. (2004a). Data are drawn from the German Socio-Economic Panel Study. Respondents
are asked: "How satisfied are you at present with your life, all things considered?" The responses run from
0 (completely dissatisfied) to 10 (completely satisfied).
example, Liliana Winkelmann and Rainer that control for country fixed-effects, have
Winkelmann 1998, Martin Ravallion and shown that happiness co-moves with mac
Michael Lokshin 2002, Ada Ferrer-i roeconomic variables including GDP, GDP
Carbonell and Paul Frijters 2004, Claudia growth, and inflation (see, for example,
Senik 2004, Ferrer-i-Carbonell 2005, Rafael Di Telia, Robert J. MacCulloch, and
and Andrew E. Clark et al. 2005). Further, Oswald 2003, John F. Helliwell 2003, and
a number of these studies have been able Alberto Alesina, Di Telia, and MacCulloch
to utilise exogenous variations in income 2004). A useful set of recent figures is to be
to establish more firmly the causal effect found in Andrew Leigh and Justin Wolfers
of income on happiness (e.g., Jonathan (2006).
Gardner and Oswald 2007; Frijters, John
P. Haisken-DeNew, and Shields 2004a; The bulk of the evidence in 1) - 3) thus
Frijters, Haisken-DeNew, and Shields suggests that income does raise happiness.
2004b; Frijters et al. 2006). It is again One of the key challenges for the nascent
the case that income has a larger effect economics of happiness literature is therefore
in transition than in developed countries. to render the significant income coefficient
In addition, the slope of the income found in much of empirical literature consis
happiness relationship is not necessar tent with the time profiles shown in figures 1,
ily the same between groups (Clark et 2, and 3 and to identify the ensuing implica
al. 2005, Frijters, Haisken-DeNew, and tions of the fundamental income-happiness
Shields 2004b, Lelkes 2006). relationship for both economic theory and
policy design.
3) Recent detailed studies of the "macreco This paper attempts to respond to that
nomics of happiness" using very large sam challenge. Our answer is based on the con
ples and cross-time cross-country models cept of income comparisons?both to others

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
in the relevant reference group (social com 2. Explaining the Easterlin Paradox by
parisons) and to oneself in the past (adapta Relative Income
tion or habituation). In section 2, we provide
a unified account of the observed income The explanation of the Easterlin para
happiness gradients in both the micro and dox detailed in this paper rests on the ways
macro literature by presenting them as in which income translates into utility. It is
straightforward extensions of the textbook important to be clear about the logical step
utility function. that we are taking here. While the paradox is
We then turn to the question of microeco couched in terms of income and happiness,
nomic evidence that is consistent with the we are going to appeal to a specific type of
presence of income comparisons in the utility utility function to account for it. In other
function. The recent growth of the empiri words, we imagine that happiness scores
cal literature on income and happiness has provide information about utility. We will
produced much information in this respect. maintain this hypothesis over both this sec
We summarize these new findings in sec tion and the micro-level income and happi
tion 3, especially focusing on studies that ness results described in section 3. Section 4
have used panel data from surveys such as will then explicitly set out the evidence link
the British Household Panel Survey (BHPS), ing happiness and utility.
the German Socio-Economic Panel Study In this section, we consider the implica
(GSOEP), and the Russian Longitudinal tions of relative or comparison income terms
Monitoring Survey (RLMS). These panel in the individual utility function. These com
studies allow researchers to track individu parisons may concern others, or oneself in
als' income and happiness over long periods the past, evoking the possibility that indi
(now over twenty years in the case of the viduals adapt or "get used to" their changing
GSOEP) and to control for individual fixed income (Easterlin 2001). Both of these types
traits, the latter having been shown to be of comparison can be presented as simple
crucial for the empirical modeling of sub extensions to the standard economics text
jective well-being (Ferrer-i-Carbonell and book utility function. Consider a utility func
Frijters 2004). tion of the form:
In section 4, we directly address the ques
tion, "Is happiness related to utility?" In (i) ut = u(Ul(Yt),u2(x\y:)MT - it,zu)i
particular, we consider a number of findings
from the analysis of objective data, experi where U(.) is a common function over all
mental economics, and neuroscience which individuals indicating how the subutilities ux,
are consistent with relative income play u2, and u3 are combined into final utility U;
ing a role in the individual utility function. the subscript t refers to time.
However, we also underline a number of In this specification, Yt is the vector of
possible objections to the use of happiness incomes yt from t = 0 to t and uY{.) can be
data to reveal such income comparisons. thought of as the classic function showing
Section 5 then highlights some of the main utility from consumption, which is increas
implications of income comparisons for a ing at a decreasing rate in its argument. As
range of issues relating to economic theory we are thinking of a vector of incomes in
and policy design. The economic issues we general, past incomes may affect current
focus on include many of the central con consumption, for example via wealth. In a
cerns of economics: consumption, invest one-period model, or in a model without
ment, economic growth, savings, taxation, savings, income will equal consumption
labor supply, wages and migration. Finally, ct, so that ux(Yt) = ux{yt) = u^c^. The
section 6 concludes. subutility function u3(T ? Zi?Zlf) picks up

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
the influence of leisure, (T ? lt), with lt rate in Yt, but decrease at an increasing rate
denoting hours at work, and a vector of other in y*. The status function is homogeneous of
socio-economic and demographic variables, degree zero, so thatu2(aYt\aY*) ? u2(Yt\Y*):
z1(. status is unaffected by proportional increases
The empirical application of (1) typically in income and reference income. In many
appeals only to current values of Yt and a par cases, u2(Yt\Y*) = ct/ct, where ct is average
tial log-linear specification: reference group consumption, but the for
mulation is sufficiently general to encompass
(2) Ut = A ln(yt) + ?2ln(yt?t) + z'ty, the bulk of the specifications used in the
literature.
where yt is usually a measure of real indi In the following subsections, we show how
vidual or household income, y* is some spe this basic model can easily explain the Easterlin
cific reference income, and Z includes both paradox, first considering comparisons to
demographics and hours of work. others, and then comparisons to one s past.
While the first and third parts of the utility
function in (1) are standard, the second is less
2.1 Social Comparisons
so, and shows the influence of status or habit To illustrate the main forces at work when
u?t ion. The economic analysis of such relative individuals compare to others, consider the
income effects (or, more generally, interde following stylised implication of the relation
pendent preferences) can be dated back to at ship between income and happiness across
least Thorstein Veblen (1899), and then James countries when: i) income is the only system
S. Duesenberry (1949). More recent contribu atic difference between countries (so that we
tions include Robert A. Pollak (1976), Robert can relegate u3 in equation (1) to a constant
H. Frank (1985), and Jon Elster and John E. and ignore it); and ii) reference income is
Roemer (1991). average income within the country. This case
The variable Y?* is often called "reference is depicted in figure 4, for the function:
group" or "comparison" income, and the
ratio Yt/Y* is called "relative" income. Any
empirical test of such a utility function will yi -r A
require us to specify individual reference
groups. In this respect, we can distinguish with y i being average income in the country
between internal reference points, such as where individual i lives, and A being a
own past income or expected future income, positive constant. The functional form here
and external reference points, where com is deliberately chosen to ensure that the
parisons refer to distinct demographic groups benefit of an across-the-board proportional
such as own family, other workers at the rise in income tends to zero as income goes
individual's place of employment, people in to infinity: a general rise in income leaves the
the same neighborhood, region, country, or second term unchanged, and has an effect on
even people across a whole set of countries. the first term which tends to zero as income
With external reference points, u2(Yt | Y*) can increases. Note that this is not true of the
be interpreted as the "status return" from formulation in (2) where a growth in log
income, or the positional or conspicuous con income by x will increase utility by x?x for
sumption aspect of income.2 This status func any level of income.
tion is assumed to increase at a decreasing The main prediction of this model is that
the gradient between income and happiness
will be steeper within a country at a point in
2 Others' income might also matter for noncomparison
reasons: for example if a general rise in income leads to time than over time by country. This is due
higher prices. We only consider social comparisons here. to the status benefit of high income within a

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Relationship within a
country at some time
Country t2
Relation across
time in a country
0 Income
Figure 4. The Relationship between Income and Happ
happiness
country. Crucially, however, this (the dotted
status ben lin
than the
efit has no aggregate impact relationship b
on country-level
happiness (in this model,income
the moreand status
aggregate
one h
person has, the less others
by thehave:
thickstatus is the
line). In a
zero-sum game). Over individuals
time in a given coun
are relatively
try, the only effect of no aggregate
income benefit at th
on aggregate
higher
happiness will be via the income, but
consumption com there
individual
ponent of the utility function status return t
(%).
Figure 4 is easiest to interpret if we
This stylized imag
illustration
sive opinion
ine that, over time, individuals over the last
in a particular
country move from the leftthe ellipses
relationship betwee
to the right
piness
ellipses. At t0, the population inat the
this individual
country is
marginal
poor and the slope between utility
individual from
income
and individual happiness approaches
(which is zero
shown as by
count
(in equation
the dotted line) is relatively (2),f1?
steep. At this mar
the
population has become?i/y; in the richer,
somewhat specification
and
the relationship between it isindividual income
/3iA/(?/f+A)2). On the
and individual happinessginalis less steep
utility than
of extra stat
zero,
at t0. In the third period, because
t2, inincome
average general y
is high and the slope between
income) individual
rises in line with o
model thus
income and individual happiness explains
is fairly flat. the E
It is clear that in all adox
three periods
and concursthe sta m
with
tus return from income logical and
yields some of the e
a relationship
between individual income andin
At a point individual
time, those

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
enjoy higher consumption and higher status this case, the discussion applied to figure 4
(and are thus happier); over time, as every is valid for each country, although individual
one becomes richer, as the amount of status countries at a point in time may be on different
is fixed, the only benefit to the country is portions of the unbroken line, depending on
from higher consumption, the value of which their income level. However, if one country's
drops toward zero. GDP grows relative to that of its neighbors,
This simple model can be embellished by then y?/y* will change, and the high-growth
considering the relationship between income country will enjoy greater happiness. The
and happiness across several countries simul best outcome for each country is to have
taneously, as in Di Telia, MacCulloch, and high income while its neighbors have low
Oswald (2003). Here the authors estimate incomes. However, unless one country can
individual happiness equations over twelve increasingly outstrip its neighbors, the addi
countries and eighteen years, controlling for tional benefit of more income is subject to
not only individual demographic variables, decreasing returns.3 This type of happiness
but also country fixed effects, time dummies, function can help explain why countries are
and macroeconomic variables such as lagged locked in an arms race over growth, even
GDP. They find, as Arie Kapteyn, Bernard M. though, on aggregate, that growth will only
S. Van Praag, and Floor G. Van Herwaarden bring utility via the consumption function.
(1976) had previously argued, that "social ref In each country, the component ?3ln(y*t/y*)
erence spaces" (reference groups) can include produces a strong relationship between GDP
whole countries, and that happiness within a and happiness. However, analogously to the
country is strongly positively correlated with individual argument within a country, the
GDP growth over the last year. This can be happiness return from being richer than
squared with the general observation that, other countries is, from the world perspec
over long periods of time, GDP and happi tive, a zero-sum game.
ness are uncorrelated in richer countries by At the individual level, these kinds of sta
an expanded happiness function with two tus-races can lead to suboptimal outcomes
different kinds of comparison: if they crowd out nonstatus activities. This
can be illustrated using the general (one
(3) UiJt = ?Mytft) + ?2\n{yijt/yft) country) utility function (1), where a higher
income for individual i reduces the utility of
everyone whose reference group includes i.
+ ?3ln(y?/y;) + Zl?y. In the specification proposed, higher income
comes about at the expense of leisure time.
Here Ui? is the happiness of an individual i Consider the parameterization:
in country j at time t, yt is average income in
country j at time t; and y * is average income (4) Uit = ?Mya) + ?2^(yitlyt),
over the whole set of countries (say Europe)
at time t. This happiness function is of the + yln(T - yitlwt),
same nature as that appealed to in (2) to
describe happiness within a country, but with where the expression y In (T ? yitlwt) reflects
an added component (/33ln(^*/j/*)) reflecting the utility from leisure (which is written as
the income of one country relative to that in
another set of countries. This added compo 3 Note that if the "true" happiness function does indeed
nent shows individuals' utility gain from liv depend (negatively) on some measure of reference group
income, but we instead estimate a happiness equation that
ing in a relatively successful country.
does not include y*, then the negative effect of higher lev
If income in all countries grows at the same els ofy* over time will show up as a negative time-trend (as
pace, then yt/y* will remain unchanged. In in Di Telia, MacCulloch, and Oswald 2003).

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Marginal happiness of
additional income to an
individual
Marginal happiness of
additional income to a
country
Figure 5. The Marginal Happiness of Additional Income for an Individual versus a Country
T minus the number of hours spent earning same time (i.e., when all individuals make th
income yit at wage wt). Figure 5 illustrates same choice). This effectively removes th
the individual's utility-maximizing choice status benefit of higher income. The secon
of income relative to that pertaining in the curve lies below the first due to the negativ
social optimum (where status externalities externality of yt in the term ?2\n(yit/yt)
are internalized). Individuals choose income of Y2, where thei
In this figure, the top curve shows dUit/ marginal utility of income is zero, whereas
dyit, the marginal utility to the individual of the societal optimum, taking externalitie
additional labor income. This marginal utility into account, is at the lower income of YJ. I
is positive up to income Y2, at which point the is tempting to relate figure 5 to the literature
detrimental effect to the individual of less lei on excessive work hours (see Juliet B. Scho
sure is exactly balanced by the increased con 1991).
sumption and higher status that come with The above illustrations considered, for
simplicity, reference groups defined at the
more income. The lower curve in this figure
dUit. = dUit. , _ country or supracountry level, but the same
represents-\yt yit H-\y* yit, which is thegeneric argument holds when reference
tyit tyt groups are defined at a finer level. The empir
effect of additional income in the country
ical literature on relative utility has typically
when everyone's income increases at the to more disaggregated reference
appealed

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
that earned over the past three years, we

groups. This is partly for intuitive reasons of
social distance (people living in London are would have:
more likely to compare themselves to other
Londoners than to people living in Glasgow
(5) Uit = ?1 In (yit) + ?2 In (y?/tf) + Ttty
or Cardiff; people compare more within
their age cohort than outside of it), and partly
to obtain sufficient variation in comparison
income, j/*, to allow for a tight estimate of its
coefficient.
The reference groups appealed to in the
discussion above can be thought of as exter uu = ?Myu)
nal. The next subsection discusses a utility
function with internal reference points, spe
+ &[>(?/?) - aln(j/i(_1)
cifically the individuals own past income or
income aspirations for the future.
- yln(i/i(_2)
2.2 Adaptation
- (l-a-y)ln((/j(_3)]
The second main explanation of the Easter
lin paradox relies on adaptation to the argu
ments of the utility function. Here we are + z;ty.
principally concerned with adaptation to
income, although recent work by economists
and psychologists has covered other life In the final utility function, we have the
domains, such as unemployment, marriage, logs of current income and income over
divorce, and health. With income adaptation, the last three periods.4 This equation can
individuals get used to their circumstances, in principle be extended to include further
so that changes in income have only tran lags of current income; if aspirations are
sient effects. Shane Frederick and George F. important (another internal reference point,
Loewenstein (1999) define adaptation as "a but this time forward-looking) it may also
reduction in the affective intensity of favor include expected future incomes. One of
able and unfavorable circumstances" and the the main implications of this specification
concept of reversion back to some baseline is that the short-run effect of an increase in
hedonic level following temporary highs log income equals ?x + ?2 whilst the long
and lows in happiness has been termed the run effect is only ?Y. This is obviously analo
"hedonic treadmill" (Philip Brickman and gous to the social comparison case, where
Donald T Campbell 1971). Miles Kimball the marginal utility of higher income was
and Robert Willis (2006) provide a fuller greater when others' income remained con
review of work on the psychology of adapta stant than when others' income rose in line.
tion and reference points.
From an economist's point of view, a sim
ple way of thinking of adaptation to income 4 We do not specify here whether incomes are nomi
is in terms of an internal backward-look nal or real. Practically, models using lagged income terms
express them in real terms or include time dummies.
ing reference point. We thus remain in the However, in the case of money illusion individuals may
general framework of equation (2), but now compare nominal rather than real amounts. A recent arti
cle (Stefan Boes, Markus Lipp, and Winkelmann 2007)
consider that y* is formed from own past
uses long-run panel data to test for the presence of money
incomes. If the individual compares her own illusion in subjective well-being judgements, concluding
income at time t to (a geometric average of) that it is largely absent.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Income
happiness
Happiness jump at the end of second year
followed by gradual dissipation
Jump in income at the

end of second year
2 3
Time in years
Figure 6. Change in Happiness Following an Income Shock
In terms of figure 4, the short-run benefit of have continually rising income.5 Adaptation
higher income is illustrated by the dotted therefore potentially explains the Easterlin
lines, whereas the (flatter) thick line shows paradox of a flattish long-run relationship
the long-run benefit. between income and happiness, but a steeper
Figure 6 illustrates a simple case where ?i short-run slope.
= 0 (so that there is no consumption benefit This section has proposed two flavors of
from income), and a = y = 1/3, which cor a model of income comparisons in order to
responds to the situation where the short-run explain the Easterlin paradox. This paradox
benefit of higher income dissipates linearly is expressed in terms of income and happi
over the following three years. For illustrative ness; in this section we have worked under
purposes we have smoothed this dissipation. the assumption that happiness and utility
The top line denotes happiness and the are synonyms, and have proposed explana
lower line income. The latter is constant for tions based on modifications of the utility
the first two years, jumps at the beginning function. The following section summarizes
of year two, and remains constant thereaf developments and issues in the recent lit
ter. At the time of the income shock, hap erature that has used individual-level happi
piness also jumps, but due to the gradual ness information to try to find evidence that
adaptation of reference income, happiness
returns to its initial level by the beginning 5 Not only do rising wage profiles discourage turnover
of period five. In this set-up, the only way to for incentive reasons, in this model they also provide util
achieve permanently greater happiness is to ity to the worker.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
relative income really does matter. Section 4 "someone like me", where the regression
then lists a number of ways of testing for the controls for individual characteristics such
presence of relative income terms in the util as age, sex, education and region, as in Clark
ity function that do not rely on subjective and Oswald (1996). Second, perhaps more
well-being data. simply, we can compute cell averages (for
example, average wage by region, sex, and
education). This latter calculation can either
3. Evidence of Comparisons using be carried out within the dataset, or matched
Happiness Data in from an external source.
The growing economics of happiness A crucial issue in the econometric lit
literature is testament to the fact that an erature is that of identification: y* is typi
increasing number of economists believe cally estimated as a linear function of some
that self-reported well-being data contain explanatory variables X1 in the wage equa
valuable information that can complement tion approach. To then identify the effect
our understanding of individual behavior.6 In of y* on happiness, we need either exclusion
terms of the specific subject of this review, restrictions (some variables which appear
happiness data are the cornerstone of the in Xj but which do not enter the happiness
Easterlin paradox; this section asks whether equation), or identification directly from the
the same data can be used to resolve this para functional form (such as when the prediction
dox, by empirically demonstrating the impor of y* enters in a different functional form
tance of social comparisons and adaptation. in the happiness regression to the variables
A rapidly growing number of econometric in Xj). The cell average approach relies on a
studies have used survey data on happiness more subtle exclusion restriction that indi
or life satisfaction to evaluate the importance viduals compare themselves only to the aver
of "absolute" versus "relative" income. Under age income within each cell.
the maintained hypothesis that happiness is The empirical literature mostly started
a good proxy measure of utility, this corre by considering job satisfaction, reflecting
sponds to estimating the relative size of the economists' interest in wages and the labor
coefficients ?1 and ?2 in equation (2). market, and perhaps also the original
research carried out in industrial psychology,
3.1 Happiness and Social Comparisons
before moving on to global measures of well
All empirical tests of social comparisons being such as happiness and life satisfaction.
over income, whether using happiness data Probably the first economist to estimate
or any other approach, require candidate subjective well-being equations using both
measures of y*. One such candidate is the y and y* was Daniel S. Hamermesh (1977).
income of "people like me" (e.g., those with Although Hamermesh s focus is upon occu
the same age, education, etc., who are doing pational choice and the effects of training
the same kind of job). This reference group in American data, and he does not discuss
income can be calculated in two different relative income in detail, his job satisfac
ways. We can first estimate wage equations tion regressions include the residual from
and then compute the predicted income of a wage equation as an explanatory variable.
This residual, y - y* in our terminology,
6 A search of EconLit for journal articles with either
has a positive and significant effect on job
satisfaction.
"happiness," "life satisfaction," or "well-being" in the title
identifies 465 published articles between 1960 and 2006. The regression approach of calculating the
Of these, 363 (78 percent) have been published since
1995, 285 (61 percent) have been published since 2000,
income of "people like me" was also used by
and one-third of the literature (37 percent, or 173 articles) Clark and Oswald (1996) on the first wave
has appeared in print in just the last three years. of BHPS data. The estimated coefficients

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
on income and comparison income in a job their own income. Graham and Andrew
satisfaction equation are statistically equal Felton (2006) replicate this finding across
and opposite, which is consistent with a eighteen Latin American countries. Helliwell
fully relative utility function: to paraphrase and Haifang Huang (2005) is in the same
Easterlin (1995), in these results increasing vein, calculating average household income
the income of all increases the happiness of by census tract in Canadian GSS data. The
no one. Louis L?vy-Garboua and Claude estimated coefficient on this variable in life
Montmarquette (2004), and P. J. Sloane and satisfaction equations is negative, and equal
H. Williams (2000), using Canadian and in size to the positive coefficient on house
British data respectively, have also found hold income, suggesting that life satisfaction
evidence that econometrically predicted is totally relative in income. As the estimated
comparison income is negatively correlated coefficient on income refers to ?1 + ?2 in
with job satisfaction. equation (2), and that on relative income
Articles which calculate comparison in to -?2, the finding that the coefficients are
come as a cell average, rather than an econo equal and opposite is tantamount to say
metric prediction from individual data, ing that the consumption benefit of higher
include Peter Cappelli and Peter D. Sherer income (ft) is essentially zero, which is con
(1988), who find that pay satisfaction is nega sistent with figures 1 to 3.
tively correlated with an outside "market A novel paper dealing with social compari
wage," calculated by averaging pay for spe sons is John Knight and Lina Song (2006).
cific occupations in other firms (airlines, in This paper appeals to cross-sectional infor
this case). Clark and Oswald (1996) find a mation on 9,200 households in China, and
negative relationship between job satisfac thus refers to an economy which is very
tion in BHPS data and average earnings by different from the Europe-North America
hours of work matched in from the U.K. nexus which has so far dominated the litera
Labour Force Survey. ture. The authors are not only able to iden
Stepping outside of the realm of work, tify which villages their respondents came
a number of recent papers have found from, but also confirm that 70 percent of
comparison income effects using cell means. individuals indeed see their village as their
Ferrer-i-Carbonell (2005) calculates com reference group (by simply asking them to
parison income as an average within fifty whom they compare themselves), making
cells defined by sex, age, and education in their rural sample well-suited to the question
six years of German GSOEP data; Michael of how important reference groups really are.
McBride (2001) uses 1994 data from the Controlling for own income, and for village
General Social Survey and defines compari income, those respondents who say that their
son income as average earnings of the indi income was much above the village aver
vidual's cohort, defined as those who are age report far higher happiness than those
between five years younger and five years who say that their income was much below
older than her. Blanchflower and Oswald the village average. The difference between
(2004) use GSS data over the period 1972 the two estimated coefficients implies a hap
98, with y* defined as average income by piness boost of one point, on a zero to four
state. Erzo F. P. Luttmer (2005) also takes scale, making relative income the most
a geographic approach to reference groups, important right-hand side variable.
and calculates average income by local area The above work considers y* as the income
identified in a number of waves of the U.S. of "people like me" or those living in the same
National Survey of Families and Households; neighborhood. Another potential peer group
this is shown to be negatively correlated with is those with whom the individual comes into
respondents' life satisfaction, conditional on close daily contact: her family, friends, and

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
work colleagues. With respect to the latter, the first published article being Van Praag
and despite the current abundance of micro (1971). This project involved asking indi
economic data, very few papers have related viduals to assign income levels (per period)
individual well-being to coworkers' wages. to six different verbal labels (such as "excel
One direct test is Gordon D. A. Brown et al. lent," "good," "sufficient," and "bad") and
(forthcoming), who use matched employer then, based on the values given, estimating
employee data from the British Workplace for each individual a lognormal "Welfare
Employee Relations Survey. Individuals Function of Income." The resulting indi
were asked to report their satisfaction with vidual estimated means (jul) and variances
the amount of influence they have, their (a) were then used as dependent variables
pay, their achievement, and the respect they in regressions which sought to explain which
receive. Controlling for own wage, the (nor types of individuals require a higher level of
malized) rank of the individual in the firm income to be satisfied, and which individu
wage distribution is correlated positively and als have valuations that are more sensitive to
significantly with all four measures of satis changes in income.
faction (see their table 6b). The results using cross-country data pro
The situation is equally sparse with respect duced a number of important findings. In
to family and friends. Clark (1996b) uses terms of this paper's subject, we would like
BHPS data to relate individual job satisfac to know who has a higher value of ?x (i.e.,
tion, conditional on own wage, to the wages who needs more money to be satisfied?).
of their partners and the average wage of Comparisons to others were analyzed via the
other household members. The results show inclusion in the regressions of reference group
that individuals do indeed report lower job income (usually cell average income over age,
satisfaction scores the higher are the wages education and certain other individual or job
of other workers in the household. McBride characteristics) as a right-hand side variable.
(2001) also introduces a family benchmark, The empirical results (for example, Aldi J.
appealing to the question in the GSS: "com M. Hagenaars 1986 and Huib Van de Stadt,
pared to your parents when they were the Kapteyn, and Sara Van de Geer 1985) show
age you are now, do you think your own that, ceteris paribus, the higher is the refer
standard of living now is: much better, ence group's income, the higher are the lev
somewhat better, about the same, somewhat els of income assigned by individuals to the
worse, or much worse?" While this is a valid six verbal labels, as social comparisons over
approach, it is worth noting that it is perhaps income would imply.
a poor candidate to explain the flat income One of the very few papers ever to
happiness relationship, as it remains fixed appeal to respondent-defined (rather than
over time. In other words, for the same indi researcher-defined) reference groups is
vidual, y* does not change with y, although Bertrand Melenberg (1992). He uses 1985
new cohorts will presumably have higher val and 1986 Dutch Socio-Economic Panel data
ues of yl than will older cohorts. in which individuals are asked about their
Modeling the utility function via proxy social environment?the "people whom you
variables, such as life or job satisfaction, meet frequently, like friends, neighbors,
is not the only way to demonstrate social acquaintances or possibly people you meet
comparisons. One method that essentially at work." Respondents are asked to indicate
inverts the question is that of the Welfare the average age, household size, income, edu
Function of Income, associated with the cation and labor force status in this group.
Leyden school in the Netherlands and, par Melenberg shows that the average income of
ticularly, with Van Praag. This predates the this (respondent-defined) reference group is
work on satisfaction by some years, with positively and significantly correlated with

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
the estimate of ?jl from the WFI: those who circumstances, such that, in the long run, sta
associate with higher-earners need more ble characteristics do not affect well-being.
money in order to describe their income as In the same tradition, Clark (1999) uses
good or adequate. two waves of BHPS data to look at the rela
tionship between workers' job satisfaction
3.2 Happiness and Adaptation
and their current and past labor income.
There is a large literature in psychology The panel nature of the BHPS makes it pos
that deals with the general issue of adapta sible to concentrate on individuals who stay
tion in many life domains (see Frederick in the same firm, and in the same position
and Loewenstein 1999), but only a very few (i.e., have not been promoted or moved job in
studies have focused on income adaptation any other way). Both current and past labor
(see the work reviewed on their page 313). income and hours are used as explanatory
Perhaps the most famous example is that of variables. Past income attracts a negative
Brickman, Dan Coates, and Ronnie Janoff coefficient in the job satisfaction equation,
Bulman (1978), who show using a very small and past hours a positive coefficient, consis
sample of lottery winners (n = 22) that this tent with a utility function that depends on
group with their positive income shock do not changes in these variables. The data suggest
have significantly higher life satisfaction than a completely relative function, with job satis
a control group.7 They propose an explana faction depending only on the annual change
tion based on the twin concepts of contrast in the hourly wage. Christian Grund and Dirk
(i.e., winning money opens up new pleasures Sliwka (2007) find similar results in German
but also makes existing pleasures less enjoy GSOEP panel data. Matthew Weinzierl
able) and habituation (winners get used to a (2005) introduces both past income and ref
new standard of living). More recent exam erence group income (calculated as a cell
ples of adaptation in nonmonetary spheres mean by gender, age, and education) in life
are Richard E. Lucas et al. (2003) and satisfaction equations using the GSOEP data,
Lucas (2005) with respect to marriage and and finds negative and significant coefficients
divorce, Stephen Wu (2001) and Oswald and for both. Last, Tania Burchardt (2005) finds
Nattavudh Powdthavee (2005) for adaptation evidence of adaptation in income satisfaction
to illness or disability, and Lucas et al. (2004) in ten years of BHPS data, with a suggestion
regarding unemployment. of greater adaptation to rises in income than
Here we are especially interested in adap to falls in income.
tation to income changes. One early article is A recent detailed study of life satisfaction
Inglehart and Jacques-Ren? Rabier (1986), and income adaptation was carried out by
who use pooled Eurobarometer data from ten Di Telia, Haisken-DeNew, and MacCulloch
Western European countries between 1973 (2007), who analyze longitudinal data for
and 1983 to show that life satisfaction and around 8,000 individuals drawn from the
happiness scores are essentially unrelated to West German sample of the GSOEP over
the level of current income, but are positively the period 1984 to 2000. They find that
correlated with a measure of change in finan the effect of an income increase after four
cial position over the past twelve months. years is only about 42 percent of the effect
Their conclusion is that aspirations adapt to after one year: the majority of the short-term
effect of income vanishes over time.
An alternative to using individual income,
7 Important though this paper is, it is worth noting that and its lags, is to concentrate on aggregate
the paper is cross-section ex post: no shock is observed. income. Di Telia, MacCulloch, and Oswald
Further, winners were actually more satisfied than non
winners, but, given the small sample size, the difference (2003) examine individual happiness in
was not significant. data covering eighteen years across twelve

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
110 journal of Economic Literature, Vol. XLVI (March 2008)
European countries, and argue that some of A second individual-level reference point
their results on current and lagged GDP per is aspirations. The concept is the same as that
capita show that "bursts of GDP produce of adaptation: if aspirations rise with own
temporarily higher happiness" (p. 817). actual income, then the effect of income on
The Leyden Group (e.g., Hagenaars 1986, happiness will be muted.
Van de Stadt, Kapteyn, and Van De Geer As might be imagined, there is only rela
1985, Erik J. S. Plug 1997, and Van Praag 1971; tively little work here, as it is difficult to
for a review, see Van Praag and Frijters 1999) know how to accurately measure income
explicitly attempted to measure the degree aspirations.9 Easterlin (2005a) uses direct
of adaptation to income. The cornerstone of measures to show that material aspirations
this empirical work is the Welfare Function (the big-ticket consumer items that make up
of Income, as described in section 3.1 above. the good life) seem to increase in line with
Questions permitting a direct estimate of ownership of such consumer items. However,
the income needed to achieve a fixed level this is not true with respect to marriage,
of welfare were posed in the GSOEP, in the where over forty percent of those who have
EUROSTAT surveys of the 1980s, in Russian been single their entire lives, and are aged
panels, in the Dutch Socio-Economic Panel, 45 and over, cite a happy marriage as part of
and in various other surveys. The relation the good life. Two recent papers have taken
ship between this required income level and different approaches to measuring income
the individual's past income can then be seen aspirations, and relating them to subjective
as a direct measure of adaptation, or as Van well-being. Stutzer (2004) combines the
Praag (1971) calls it, "preference drift." The analysis of subjective data with the income
stylized finding for about twenty European evaluation approach of the Leyden school,
countries is that a $1 increase in the income by using the answer to the Minimum Income
of a household leads to a 60 cents increase Question10 as a measure of individual income
(within about two years8) in what people aspirations (and thus one measure of y*) in a
consider to be a "excellent," good," "suffi life satisfaction equation.
cient," and "bad" income. Income adaptation McBride (2006) introduces a novel way of
is therefore high, but not complete in this calculating aspirations directly in a matching
methodology. pennies game, where individuals play against
The individual-level reports match up with computers. The computer chooses heads or
what is found at the aggregate level concerning tails according to (known) probability dis
subjective poverty (having an income lower tributions (for example 80 percent heads, 20
than that was deemed minimal). European percent tails). After each round of playing,
countries which are on average poorer (such individuals report their satisfaction with the
as Greece and Portugal) are found to have outcome. McBride's first contribution is to
many more respondents whose own income introduce social comparisons in some of the
was below an insufficient level than richer treatments (by telling the individual the out
European countries such as Germany or comes of the other players). Second, he is able
Switzerland. For instance, subjective poverty to identify an aspiration effect by varying the
was about 3 percent in West Germany in the
1990s, but up to 90 percent in Russia in 1993
9 Suggestive indirect evidence is easier to find. Clark
(Van Praag and Frijters 1999).
(1997), for example, suggests that the stubbornly higher
job satisfaction reported by British women in BHPS data
might partly reflect their lower expectations.
8 The 60 percent finding was initially based on cross 10 Where individuals are asked to indicate the sum
sectional within-country data, but has since also been per period they think is the absolute minimum net family
found to hold over time. See Van Praag and Frijters (1999) income their family requires to make ends meet. This was
for specific longitudinal results. introduced in Theo Goedhart et al. (1977).

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
heads and tails probabilities played by the function is shown in figure 4, which rep
computer. Each subject has five pennies to resents the basic aspects of the Easterlin
play. When paired with a 80 percent heads, "Paradox" shown in figures 1 and 2. It is
20 percent tails computer, the best strategy is possible that even this small positive long
to always play heads, which gives an expected run effect may be an overestimate, as new
payoff of four pennies. When paired with a generations or cohorts may start with higher
65 percent heads, 35 percent tails computer, aspiration levels than older generations. Any
the best strategy is still to always play heads, such intergenerational adaptation of aspira
but now the expected payoff is only 3.25 tions would further diminish the long-run
pennies. By manipulating the probabilities, aggregate effect of higher income, but
McBride creates variations in aspirations. is at present still ill-accounted for in the
The empirical analysis shows that satisfac literature.
tion is (a) higher the more one wins, (b) lower
3.4 Key Challenges for Empirical Work
the more others win, and (c) lower the higher
was the aspiration level. Akin to many areas of applied econom
ics, establishing the nature of the empirical
3.3 Do Social Comparisons and Adaptation
relationship between income and happiness
Explain the Easterlin "Paradox"?
faces a number of challenges, even if we pre
Some of the research that we have cited sume that happiness is perfectly measured
above allows us to undertake tentative back and conforms to experienced utility. Here we
of-the-envelope calculations of the relation highlight a number of the main difficulties.
ship between income and happiness. For Firstly, economic theory often dictates that
example, we can take the key finding in the the relevant measure of welfare is consump
Leyden literature that adaptation over time tion, not income, and that income in happiness
accounts for around 60 percent of the effect regressions is only a noisy proxy for consump
of income (i.e., income's long-run effect tion (Weinzierl 2005). As such, researchers
is only 40 percent of its short-run effect), will tend to underestimate the importance of
which corresponds closely to the results in material circumstances on happiness. Bruce
Di Telia, Haisken-DeNew, and MacCulloch Headey and Mark Wooden (2004) go some
(2007). We can further appeal to one of the way toward addressing this issue. They use
best sources of information on the extent Australian panel data (HILDA) and find that
of social comparisons, Knight and Song's "net worth," which is arguably a better proxy
(2006) finding that relative income is at least for current consumption than a transitory
twice as important for individual happiness measure of income, matters broadly at least
as actual income, even in poor regions (in as much as does income in determining hap
their case rural China). Together, this sug piness. As they conclude, "the unimportance
gests a utility function in which two-thirds of material circumstances has been exagger
of aggregate income has no effect because ated." The main reasons why consumption and
it is status-related, and thus disappears in income may differ are the consumption that
a zero-sum game, and where 60 percent of individuals obtain directly from others, and
the effect at the individual level evaporates
within two years due to adaptation. Hence
formation in Enrichetta Ravina (2005), using panel data
only around 13 percent of the initial individ on U.S. credit card holders' consumption expenditure.
ual effect will survive in the long run at the Weinzierl (2005) includes both cell-average reference
aggregate level.11 Precisely such a happiness group income (by age, sex, and education) and lagged
income in a life satisfaction equation. The estimated coef
ficients imply that satisfaction is completely relative with
11 These percentage figures are remarkably close to respect to income. We do not know, however, whether this
the estimates of interdependent preferences and habit definition of the reference group is apt.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
deferred consumption via savings. Regarding areas. Only very few studies ask individuals
the first of these, individuals in developed about their reference groups, rather than
economies are provided with a great deal simply imposing one. As noted in section 3.1,
of consumption goods via the State, such as Melenberg (1992) asks respondents directly
education, health care, and transfers-in-kind, about the income of the people with whom
which are only rarely taken into account in they interact often. We are only aware of one
empirical estimations. If public-goods con study where respondents were given a list of
sumption is not directly measured, then options and asked to explicitly state to whom
proxy variables, such as local area or country they compare themselves. As mentioned
income, which are related to public goods above, in Knight and Song (2006), 68 percent
via taxation, will attract positive coefficients. of survey respondents in China reported that
This will pollute the status effect of aggregate their main comparison group consisted of
incomes on happiness, so that the coefficient individuals in their own village, whereas only
on aggregate income in happiness regressions 11 percent stated that their main comparison
will suffer from upward bias if public good group consisted of individuals from outside
consumption is not taken into account. of the village.12
Even measuring personal consumption Almost all of the rest of the literature has
is difficult. Not only do individuals likely resorted to assuming a particular reference
have trouble remembering how much of income, and therefore inserts variables into
their income they have saved in financial the empirical model such as the individual's
assets, but more fundamentally it is difficult predicted income according to her charac
to establish empirically a clean borderline teristics or the income in some geographical
between purchases that have only current area, which is less convincing. The generic
consumption benefits and purchases with problem with using constructed reference
some future consumption benefit. How groups is that they might pick up effects other
much of a car or a house purchased today than social comparison: average income by
should be counted as current consumption geographical area will likely also measure
and how much as future consumption? How local public good consumption; coworkers'
much of education is current status con income may pick up measurement error in
sumption, and how much investment? Issues own reported income; and income predicted
such as these, which relate to the majority from a regression may reflect own expected
of major purchase decisions, are very tricky future income. Therefore, in the absence
and create a significant rift between theoret of accurate information about reference
ical models and empirical estimates of con groups, we should be cautious in claiming
sumption. If we do use individual income to have evaluated the importance of social
instead of consumption in happiness regres comparisons over income from happiness
sions, we should remember that income is data.
an overestimate of what is consumed when A third point is that the group of individu
young (when we save) and an underestimate als (or countries), to whom individuals com
when old (when we dissave). Forcing income pare is assumed to be exogenous, and not a
to have a single coefficient over all ages then matter of choice. Armin Falk and Markus
implies an upward bias in the effect of age
on happiness. 12 Wave 3 (2006) of the European Social Survey will go
The second major empirical difficulty, as some way to filling this lacuna. Individuals are first asked
already briefly mentioned above, is to cor "How important is it to you to compare your income with
other people's incomes?" They are then asked "Whose
rectly identify reference groups, especially income would you be most likely to compare your own
when individuals move a great deal in their with?," with responses on a showcard of work colleagues,
lifetimes and reside in high population-density family members, friends, and others.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Knell (2004) ask what happens if individuals A fourth challenge concerns the timing
can partly choose their reference groups.13 of income changes: the empirical prediction
To obtain interior solutions for this choice, from the loss-aversion hypothesis of Amos
the psychological literature has distinguished Tversky and Kahneman (1991) is that the
between "self-enhancement" and "self im absolute effect of a loss of one dollar, from an
provement" motives. A concern for status initial reference position, on individual hap
implies that individuals prefer low-income piness is greater than the effect of a gain of
reference groups: this is "self-enhancement." one dollar. Any test of this prediction, which
In the extreme, everyone would compare is highly relevant for many economic phe
themselves to the poorest individual(s), which nomena (see section 5), will require precise
clearly does not fit reality. The "self improve observation of the timing of both income
ment" motive then posits some indirect movements and reference income move
benefit to having a higher-income reference ments. Panel data, in which individuals are
group. One such benefit works through the typically interviewed only once per year, is
cost of effort: "people perform better and are consequently severely limited in its ability to
more successful if they set themselves high distinguish asymmetric happiness responses
goals or compare with high reference stan to incomes that are above and below the ref
dards" (p. 421). The main result of Falk and erence position. At present, only experiments
Knell's model is that the endogenously chosen can address this asymmetry, but even these
reference level increases with individual abil face well-known limitations: experimental
ity (as measured by the rate of transformation subjects are very often nonrepresentative;
of effort into output), so that higher-ability the laboratory situation itself may lack real
individuals will choose higher-income refer ism; and laboratory experiments on social
ence groups. The choice of reference group phenomena are inherently unsuitable for the
will then be based on the trade-off between measure of meaningful adaptation (such as
status and the higher output that comes from the adaptation of reference groups) as sub
lower effort cost. Matthew Rabien (2006) jects cannot be kept in the laboratory for suf
considers an explicit dynamic model where ficiently long periods of time. Until we can
agents face self-control problems (there are better track movements in both income and
future benefits from current effort). He reference income, the loss-aversion hypoth
shows that the "planner," who maximizes the esis will remain difficult to verify in this
individual's intertemporal utility, may find it literature.
optimal to introduce a reference level into the A fifth challenge is to deal with the issue
utility function. The optimum reference level of missing variables. No data set has all the
comes from the trade-off between the direct variables one might wish and their absence
utility cost of evaluating outcomes relatively often leads to problems. Missing variables
and the future benefits from higher current lead on to the issues of the endogeneity of
effort levels. Oded Stark (2005, 2006a) has key variables and spurious relations between
written a number of papers which appeal to income and happiness, and the problem of
reference-group choice to better explain the slope heterogeneity.
migration decisions of heterogenous indi The first effect of missing variables is to
viduals. It is important to note, however, that render income potentially endogenous. It
the empirical happiness literature is still in seems plausible that happy people, or, equiv
its infancy on this issue. alently, individuals with "happy" personality
traits, are more likely to obtain better jobs
(see Graham, Andrew Eggers, and Sandip
13 A related question is treated in Robert J. Oxoby
(2004): What if individuals can choose the domains over Sukhtankar 2004 and Sonja Lyubomirsky,
which status comparisons take place? Laura King, and Diener 2005). David J. P.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Barker (2005) similarly concludes that many DeNew, and Shields (2004b), and Frijters et
later life outcomes depend on adverse influ al. (2006) consider the large changes in real
ences during early development, and spe incomes observed in East Germany (follow
cifically links both income and depression to ing reunification) and Russia (following tran
birth size. The lack of personality traits and sition) as exogenous, and find a greater effect
early life influence variables in the data then of income on happiness than in much of
implies that income is endogenous. Drawing the existing literature. Gardner and Oswald
on these arguments, Ferrer-i-Carbonell and (2007) use information on lottery winnings
Frijters (2004) find in GSOEP data that in the BHPS as reflecting exogenous income
the partial correlation coefficient between movements. In both level and panel equa
changes in income and changes in happi tions, lottery winnings are found to sig
ness is smaller than that between levels nificantly reduce mental stress scores. It is
of income and levels of happiness. They worth underlining that natural data will only
advocate panel data techniques to account very rarely produce truly exogenous income
for unobserved fixed individual traits that movements, although this is an issue for all
produce endogeneity problems. However, work in applied microeconomics for which
even fixed-effect estimation will not identify income is important.
time-varying factors that lead to both greater Missing variables at the aggregate level are
happiness and higher income, producing important since any variable that correlates
spurious correlation. Good health, which positively with income and negatively with
allows individuals to obtain better jobs and happiness may, if excluded from the data,
increases well-being, is a good candidate for give the false impression that income does
a missing factor that may lead to such a spu not lead to greater happiness and would thus
rious correlation; marital stability and good be able to explain the Easterlin Paradox.
relations with co-workers are other possi Some candidates which might spring to
bilities. While the omission of these types of mind in this context are pollution, (lower)
variable in happiness regressions leads to an social capital, and hours of work. Can any
upward bias on the income coefficient, the of these indeed explain why growth is not
reverse holds with respect to variables that making us happier? Probably the most
are themselves influenced by income and detailed attempt at tackling this research
which are included as separate regressors in question is Di Telia and MacCulloch (2005)
a happiness regression. Health again fits the using twenty-three years of Eurobarometer
bill, as does housing and even marital sta data and twenty-eight years of American
tus: these outcomes are improved by higher GSS data. They examine a series of poten
incomes but are included in the regression as tial omitted variables which could explain
exogenous factors, producing a smaller esti why increasing income has not led to more
mate on the income coefficient. The balance happiness. These are life expectancy, pol
of such conflicting effects is hard to predict. lution (measured as kilograms of Sulphur
Recent years have seen a number of papers Oxide emissions per capita), unemployment
appealing to natural experiments to skirt the and inflation, hours worked, the divorce
issue of endogeneity by providing some exog rate, crime, and income inequality. Their
enous variation in income.14 Frijters, Haisken empirical results show that most of these are
DeNew, and Shields (2004a), Frijters, Haisken indeed correlated with life satisfaction in the
expected manner. However, their inclusion
14 An alternative is to instrument income, although the
task of finding instruments which affect income but not characteristics in a sample of U.K. university graduates,
subjective well-being is a hard one. Reamonn Lydon and which leads to a doubling of the size of the income coef
Arnaud Chevalier (2001) instrument income via spousal ficient in a job satisfaction equation.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
as right-hand side variables does not explain departures from log-linearity. Specifically, it
why rising income has not produced rising may miss the presence of kinks, not only over
well-being because, like income, these addi time (as in loss-aversion), but also regarding
tional variables have mostly also improved comparisons to others: is the return to having
over time without increasing happiness: in one dollar more than the neighbor the exact
their own words "introducing omitted vari opposite of having one dollar less? Better
ables worsens the income-without-happiness data and more flexible estimation techniques
paradox." are needed to address this challenge.
Missing variables may also lead to different
individuals having a different marginal ben
4. Is Happiness Related to Utility?
efit from income, i.e., "slope heterogeneity."
Presuming the same coefficient on income In this section we ask what basis there is
over the whole sample may not be appropri for believing that happiness is a reasonable
ate if there are important interacting variables measure of the economic notion of (deci
omitted from the data. Has the literature sion) utility, i.e., the thing whose maximiza
found any such interacting variables? The tion leads to choice behavior. It is, of course,
answer appears to be yes: a recent example surprisingly difficult to say whether any
is Lelkes (2006), who shows that the religious given series of numbers conforms to utility
were less affected, in life satisfaction terms, or not. The full scale of the identification
by income movements during economic tran problem can be gauged by reflecting on the
sition in Hungary. If religiosity were a miss two requirements that (decision) utility must
ing variable in this example, there would have fulfil in textbook treatments:
been slope heterogeneity on unobservables in
Hungary. Dylan M. Smith et al. (2005) pro 1. Utility guides individual choice in the
pose the same type of mediating relationship sense that choices serve to maximize the
for health. Clark et al. (2005) argue that such expected stream of utility.
slope heterogeneity is likely to be present in 2. Utility itself is the outcome of both choices
many more settings and propose to identify it and chance factors that were outside
on functional form assumptions on the error the control of the individual but whose
term and the allowed types of slope heteroge possibility was taken into account when
neity. They use latent class techniques applied decisions were made.
to three waves of European Community
Household Panel data to identify four differ The first identification problem is that
ent classes, in terms of both intercept and the in practice we are not able to say with any
estimated coefficient on income in financial precision what choices individuals really
satisfaction equations. have available to them at a moment in time.
A sixth and final challenge is the issue of Having children, getting a job, getting mar
the estimation method. Frey and Stutzers ried, health, etc. are only partially outcomes
(2002b) plea for greater use of panel tech of our own choices as they also depend on
niques to overcome some of the missing choices made by others and other factors
variables problems signaled above has largely outside of our control. This is not only the
been heeded. However, little attention has case for events in the past but also (and even
been paid to the exact specification of the more so) for possible events in the future,
independent variables and one can think of of which there are many more than actually
many nonlinearities that may be important in eventuate. Which jobs, marriage partners,
actual work but that are usually ignored. In and schools could an individual choose from
particular, the consensus use of log income and at which prices one may ask? We usually
in well-being equations may hide important do not know. This makes it in practice

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
extremely difficult to check that an observed 3. To formulate a theory for how the brain
outcome indicator of utility (say, happiness) comes up with a happiness number and
does indeed represent the best outcome then see whether choice behavior is con
attainable by that individual. A second and sistent with the happiness predictions of
related problem is that observed happiness that theory.
may not be the same construct as expected 4. To verify in laboratory and natural experi
happiness: behavior is driven by expecta ments that the found best-guess causal
tions and not necessarily by realizations. In determinants of happiness, such as rela
order to prove that a series Sit is the same as tive concerns, are also a determinant of
utility, we would therefore need to observe choice behavior in settings where all other
what the individual expected Sit to be in factors are kept constant.
all future periods under all possible future
We next proceed to discuss what each of
states, together with all the probabilities of
all future states of the world. This informa these four approaches has yielded so far, fol
lowed by a number of reasons why happiness
tion is necessary to show that the choices
might not correspond to utility.
undertaken do lead to the highest expected
future stream of Su. We would also need to 4.1 Is Happiness Related to Hard-Wired
be able to check that the realized Su cor Reward-Response Stimuli and Is It
responds to the ex ante expected Sit for the Predictable?
state of the world that came about ex post.
We would then be able to see whether the Well-being scores can be examined in
relation to various physiological and neuro
realized Sit does relate to the same concept
logical phenomena. It is known (see Peter
as the expected Su.
Shizgal 1999; Jos?-Miguel Fern?ndez-Dols
This type of information does not to our
and Maria-Angeles Ruiz-Belda 1995, and Ed
knowledge exist and seems likely to remain elu
Sandvik, Diener, and Larry Seidlitz 1993)
sive for the foreseeable future regarding happi
that there is a strong positive correlation
ness or any other candidate empirical measure
between emotional expressions like smil
of utility. What circumstantial evidence can
ing, and frowning, and answers to well-being
we then turn to support the hypothesis that
questions. Tiffany A. Ito and John T. Cacioppo
happiness is a good measure of utility?
(1999) showed that positive and negative
emotions are associated with the extent of
There have been four main approaches:
the startle response, and various measures of
facial expressions (facial electromyography).
1. Presuming that choice behavior is some A recent literature has looked at the rela
how evolutionarily hard-wired, we can
tionships between positive and negative
look for evidence that happiness or any
states, on the one hand, and neurological
other measure of utility relates to observ
measures, on the other. Obtaining physical
able hard-wired reward-response mech
anisms in the brain. If individuals are measures of brain activity is an important
step in showing that individuals' self-reports
also presumed to interact strategically,
it further needs to be shown that we are reflect real phenomena.15 Particular interest
evolutionarily geared to be able to predict
15 Davidson (2004) notes that "The identification of the
other people's happiness. brain circuitry responsible for different aspects of affec
2. To compare the trade-offs implicit in the tive processing has helped to parse the domain of emotion
best-guess estimates of the causal deter into more elementary constituents in a manner similar to
that found in cognitive neuroscience, where an appeal to
minants of happiness and to see whether the brain has facilitated the rapid development of theory
these match up reasonably well to observed and data on the subcomponents of various cognitive pro
choice behavior in those spheres. cesses" (p. 1395).

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
has been shown in prefrontal brain asym in other words, left-right asymmetry is not
metry16 In right-handed people, positive just about pleasurable feelings. Davidson
feelings are generally associated with more (2004) describes further work in which left
alpha power in the left prefrontal cortex right asymmetry is associated with quicker
(the dominant brain wave activity of awake recovery from negative affect challenge (i.e.,
adults are called alpha waves), and nega "shocks" to happiness), over and above its
tive feelings with more alpha power in the effect on baseline well-being.
right prefrontal cortex.17 This relationship Brain asymmetry is also associated with
was initially suggested by the observations physiological measures, such as cortisol and
of patients with unilateral cortical damage corticotropin releasing hormone (CRH),
(see Richard J. Davidson 2004), but more which are involved in response to stress,
recently has been explored using tech and with antibody production in response to
niques to measure localised brain activity, influenza vaccine (Davidson 2004). In gen
such as electrodes on the scalp in Electro eral, it seems that brain asymmetry is not
encephalography (EEG) or scanners in Mag only associated with measures of subjective
netic Resonance Imaging (MRI). well-being, but general measures of Wellness
A recent example is Urry et al. (2004). In of the organism's functioning.19
this study, eighty-four right-handed individu Since there is a distinct advantage in stra
als (drawn from the Wisconsin Longitudinal tegic games in knowing what the other per
Study) provide answers to questions on posi son's utility function looks like, it would seem
tive and negative affect, and measures of reasonable to ask whether a proposed mea
both hedonic well-being (using global life sure of utility is predictable by others. Many
satisfaction scores) and eudaimonic well studies have shown that individuals are able
being.18 Brain activity is measured via EEG. to a large extent to recognise and predict the
Left-right brain asymmetry is shown to satisfaction level of others. In interviews in
be associated with higher levels of positive which respondents are shown pictures or
affect, and with both hedonic and eudai videos of others, they accurately identify
monic well-being. Interestingly, the correla whether the individual shown to them was
tion between brain asymmetry and positive happy, sad, jealous, and so on (see Sandvik,
affect explains all of the correlation with Diener, and Seidlitz 1993; and Diener and
hedonic well-being, but only some of the Lucas 1999). This is also the case when
correlation with eudaimonic well-being; respondents were shown individuals from
16 Other approaches have also been explored. Brian 19 A recent review article by Sarah D. Pressman
Knutson et al. (2001) explore the relationship between and Sheldon Cohen (2005) describes the relationships
positive emotions and activity in subcortical circuits between affective "style" and physical health; see also
including the nucleus accumbens. Andrew Steptoe, Jane Wardle, and Michael Marmot
17 This is an oversimplification, and recent work has (2005). The medical literature has also found high cor
cast the left-right opposition in terms of approach versus relations in the expected sense between low well-being
withdrawal (anger, a negative approach-related emotion, scores and coronary heart disease (Stephen M. Sales
is associated with more alpha power in the left prefrontal and James S. House 1971), strokes (Felicia A. Huppert
cortex); see Heather L. Urry et al. (2004). 2006), suicide (Heli Koivumaa-Honkanen et al. 2001),
18 Eudaimonia refers to the idea of flourishing or and length of life (Erdman Palmore 1969 and Daniel K.
developing human potential, as opposed to pleasure, and Mroczek and Avron Spiro 2006). Individuals with higher
is designed to capture elements such as mastery, relations life satisfaction scores were less likely to catch a cold
with others, self-acceptance and purpose. Practically, when exposed to a cold virus, and recovered faster if they
eudaimonic well-being is measured by questions on did (Cohen et al. 2003). Blanchflower and Oswald (forth
autonomy, determination, interest and engagement, aspi coming) show that happiness and high blood pressure are
rations and motivation, and a sense of meaning, direction negatively correlated, both at the individual and at the
or purpose in life. country level.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
118 Journal of Economic Literature, Vol. XLVI (March 2008)
other cultures.20 It might then be argued of the respondent. Lastly, respondents are
that there is a common human language of sometimes given open-ended interviews in
satisfaction or happiness, so that subjective conjunction with standard questions about
well-being is at least to an extent observ their well-being. When third parties, who
able and comparable between individuals. It do not know the respondent, are played tape
has also been found that individuals in the recordings of these open-ended interviews,
same language community have a common their evaluation of the respondent's well
understanding of how to translate internal being matches well with the respondent's
feelings into a number scale, simply in order own reply.
to be able to communicate with each other.
4.2 Do the Implicit Trade-offs Look Like
Respondents translate verbal labels, such as
They Correspond to Choice Behavior?
"very good" and "very bad," into roughly the
same numerical values (see Van Praag 1991).21 There are by now many hundreds of iden
A tempting conclusion is that an evolutionary tified "correlates" of happiness and for each
advantage accrues to the accurate evaluation one of them there are difficulties in identify
of how well others are doing.22,23 ing the correct coefficients due to the usual
The general idea of having a third party problems of causality and measurement. If
evaluate respondents' happiness has been we just focus on the variables that show up
used to validate the replies that individuals in most regressions, however, what can we
themselves provide (see Sandvik, Diener, say about how reasonable the signs of the
and Seidlitz 1993 and Diener and Lucas coefficients look, and the plausibility of the
1999). When friends and family are asked implicit trade-offs?
about how happy they believe the respondent Studies looking at happiness or life satis
is, the scores they provide tend to correlate faction have identified clear positive relations
with the respondent's own report.24 Another with income, marriage, job status, health, and
obvious choice is the interviewer: again, the religion (see Kahneman, Diener, and Norbert
answer the interviewer gives tallies with that Schwarz 1999 or more recent surveys, such as
Layard 2005). Improved health, income, and
job status can be seen as extensions of the
20 This is reminiscent of work in the area of "emotion"
budget constraint. Marriage can be viewed
undertaken in the 1960s (see the description in chapter 1
of Daniel Nettle 2005). American respondents were very
as an opportunity for taking advantage of
good at identifying the emotions depicted by American specialization and access to home produc
actors in a series of photographs; but importantly so were tion. Being religious similarly can be seen as
the Dani tribespeople of Papua New Guinea.
21 More precisely, it looks as if individuals convert the
having access to spiritual goods and to psy
verbal labels into cardinal numbers that equally divide chological coping mechanisms. Hence these
up the response space. Practically, this is one reason why findings concur with what we would expect
ordinal and cardinal estimation techniques applied to
subjective measures of well-being or health most often
produce similar results (Ferrer-i-Carbonell and Frijters (1993). Third parties and respondents may share unob
2004). served characteristics which lead them to supply similar
22 Paul Seabright (2004, chapter 3) invokes the pos answers, even though the correlation between the two
sibility that smiling and laughter may have evolved as underlying constructs (how happy the individual thinks
(accurate) signals of trustworthiness. It is not easy to fake she is, and how happy her friend thinks she is) may be
smiles, and extremely difficult to fake laughter. only small. This applies particularly to third-party reports
23 A point worth making is that when asked to report from both friends and family who likely share with the
their level of happiness, life satisfaction or well-being in respondent idiosyncratic uses of language. This is less of a
surveys, only a small minority of respondents do not pro problem with third-party raters who are unknown to the
vide an answer (less than 1 percent of respondents in the respondent. The correlation is not affected if both A and B
BHPS or GSOEP). The concept of happiness is intuitively use the same linear transformation of their real evaluation
understood by almost everyone. into a report, but it will be exaggerated if part of both A's
24 This test is not as clean as it might appear at first and B's answer reflects a norm belief that people "should"
sight, for the reasons underlined in Charles F. Manski score 8 out of 10 on subjective well-being scales.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
from a mainstream view of utility. It is, how see whether the theory resulting from that
ever, worth mentioning that having children outside information correctly predicts the
and additional education only slightly affect series Sit and the associated revealed prefer
utility. Since these have a strong choice ele ence behavior.
ment to them, so that at the margin we would One implicit theory that has been followed
expect the utility effect of an additional child here is that low satisfaction in a domain of
or year of education to be zero, this general life is often (though not always) avoidable.
finding can be rationalized.25 Rational maximizing individuals are then
When we look at trade-offs in terms of predicted to be more likely to walk away
which variable explains most of the variation from jobs or marriages with low job or mari
(and is therefore worth the most happiness), tal satisfaction. If people do indeed display
health usually yields the highest number. this behavior, then this may be taken as evi
Even to the mean income earner, the differ dence that individuals maximize satisfaction.
ence between the best possible health and This theory sounds intuitively plausible but
the worst possible health is worth millions is only valid under the restrictive assumption
per year (i.e., more income than is available). that low satisfaction in a job or a marriage is,
What is also striking is how much a job and on average, predictive of the expectation that
marriage are worth. A recent estimate of individuals have about the alternative i.e.
the implicit value of marriage and a job in their satisfaction is lower when they expect
Australia is that both are worth about twice to be able to do better. This theory is usually
mean yearly income (N. Carroll, Frijters, and only implicit (though not always; see Frijters
Shields forthcoming). While these figures are 2000).
high, they are not that strange if we reflect Many panel data studies have found that
on the time and trouble that people are pre subjective well-being at time t predicts
pared to go through to find partners and jobs. future behavior, in that individuals clearly
Such trade-offs can also be used to calculate choose to discontinue activities associated
the shadow wage, as in Clark (1996a), where with low levels of well-being (see Kahneman
the negative effect of one more hour of work et al. 1993; Frijters 2000; and Baba Shiv and
per week on job satisfaction is cancelled Joel Huber 2000). Measures of life satisfac
out by a pay rise of ?8.60 per week (in 1991 tion have been shown to predict future mar
prices). Similarly, Van Praag and Barbara E. ital break-up (Gardner and Oswald 2006). A
Baarsma (2005) calculate that the negative number of labor market studies have shown
externalities from noise at Schiphol airport that job satisfaction is a strong predictor of
in Amsterdam (in 1998) could be compen job quits, even when controlling for wages,
sated by a tax of around three dollars per hours of work and other standard individual
passenger per flight. and job variables (see, for example, Richard
B. Freeman 1978, Clark, Yannis Georgellis,
4.3 Does it Correspond to Theory?
and Peter Sanfey 1998, Clark 2001, and
One way to check whether happiness cor Nicolai Kristensen and Niels Westerg?rd
responds to utility is to use outside infor Nielsen 2006). A recent example using data
mation about the function u(Xit) and to on the self-employed is found in Georgellis,
John Sessions, and Nikolaos Tsitsianis
(2007). Clark (2003) shows that mental
25 Plug (1997) considered the shadow value of children
in more depth, finding that, while the last child had no stress scores on entering unemployment pre
effect on parents' happiness, the first child did have a sig dict unemployment duration: those who suf
nificant positive shadow value in the order of tens of thou
sands of dollars. This concurs with what we might expect
fered the sharpest drop in well-being upon
from maximizing behavior, and with the trouble many entering unemployment were the quickest
parents are prepared to go to in order to conceive. to leave it.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
4.4 Do the Empirical Correlates of Social comparisons can also be demon

Happiness also Show up in Experiments? strated by asking individuals to express
preferences over hypothetical outcomes.
In controlled experiments, researchers are Francisco Alpizar, Fredrik Carlsson, and
able to change the variable of interest while Olof Johansson-Stenman (2005), Johansson
holding the rest constant. This makes an Stenman, Carlsson, and Dinky Daruvala
indirect test possible of the validity of happi (2002), and Sara J. Solnick and David
ness as utility: if some factor is important for Hemenway (1998) present respondents with
happiness, then it should also be important states of the world which differ in both the
for choice behavior when all other factors are absolute and relative domains. For example,
held constant. In the context of this paper, in Solnick and Hemenway (1998), individu
the key question is whether relative concerns als are asked to choose between states A and
show up in experiments. B, as follows:
One source of evidence on the importance
of comparisons to others in actual choice A: Your current yearly income is
behavior comes from the burgeoning experi $50,000; others earn $25,000.
mental economics literature on fairness.
Survey evidence, such as Kahneman, Jack L. B: Your current yearly income is
Knetsch, and Richard H. Thaler (1986), finds $100,000; others earn $200,000.
that people have strong views about fair
ness in economic exchange. Laboratory evi It is specified that "others" refers to the
dence on ultimatum games (Werner Guth, average of other people in the society, and
Rolf Schmittberger, and Bernd Schwarze emphasised that "prices are what they are
1982; and Vernon L. Smith, 1994) suggests currently and prices (the purchasing power
that individuals will throw away real income of money) are the same in States A and B."
to obtain a fairer division of a smaller pie. All three papers find evidence of strong
Perhaps even more explicitly, Daniel John positional concerns over income, in that
Zizzo and Oswald (2001) report the results individuals say they are willing to give up
of an experiment whereby subjects can pay absolute income in order to gain status
to burn each other's money. A majority of (choosing A over B above). Further, two of
subjects chose to do so, even though it costs the papers ask analogous questions with
them real earnings. The average subject had respect to other life domains, and compare
half of her earnings burnt, and richer sub the resulting taste for status. Concerns about
jects were burnt more often. M. Keith Chen, relative standing in Solnick and Hemenway
Venkat Lakshminarayanan, and Laurie R. (1998) are found to be strongest for attrac
Santos (2006) describe a fascinating set of tiveness and supervisor's praise, and weakest
experiments involving Capuchin monkeys, for vacation time; in Alpizar, Carlsson, and
and find evidence that their preferences Johansson-Stenman (2005) they are stronger
are reference-dependent (see also Sarah F for cars and housing, and weaker for vaca
Brosnan and Frans B. M. De Waal 2003). It tions and insurance.
is tempting to view these experimental out A natural experiment on how reference
comes in the light of some sort of compara positions affects behavior was recently
tive process.26 described by Alexandre Mas (2006). In New
Jersey, police unions bargain over wages
with their municipal employer and in cases
26 Comparisons and fairness are not synonyms how
ever: while the former implies that an individual is happy
of dispute, an outside arbitrator has the
to receive more than others, fairness considerations sug final say. Mas (2006) found a 12 percent
gest that they would prefer not to. increase in the per capita number of crimes

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
solved (cleared) when unions win their case absolute income levels once relative income
compared to when they lose, which he inter is introduced.
prets as evidence that workers care about
4.5 Why Might Happiness Not be Utility?
whether their pay conforms to a reference
position. Despite the work described above, it
Finally, we can appeal to physiological is wise to remain cautious about the link
and neurological evidence regarding status between happiness and utility. One reason
and relative income. A series of well-known why we might think that happiness is not
studies27 relates the level of serotonin in the same as utility is that happiness is an
monkeys to status within the primate group, evaluation of what has occurred, and such
and show experimentally that it is status that an evaluation may not be the same as what
produces serotonin, rather than the inverse. people expected to happen. In other words,
We are not aware of experiments that have individuals may make systematic mistakes in
shown that relative income is associated predicting their happiness. This would drive
with physiological outcomes in human sub a wedge between choice behavior and hap
jects. Animal studies have examined neuro piness maximization and thus between hap
nal activity when faced with pairs of rewards piness and decision utility (though happiness
(here different flavours or quantities of fruit could then still be experienced utility, i.e.,
juice). Previous tests establish the preference the thing that people would want to maxi
ranking over fruit juices for each monkey. mize). This issue is discussed in Kahneman,
The experimental results show that "stria Peter P. Wakker, and Rakesh Sarin (1997).
tal neurons do not process reward informa Loewenstein, Ted O'Donoghue, and Mat
tion in a fixed manner but relative to other thew Rabin (2003) specifically provide a
available rewards" (Howard C. Cromwell, model of misprediction of future prefer
Oum K. Hassani, and Wolfram Schultz 2005, ences (and therefore misprediction of future
p. 522; see also L?on Tremblay and Schultz experienced utility), and apply their model
1999). Equally, there is some evidence that to lifetime consumption and saving, and the
neuron firing is determined by the amount purchase of durable goods.
of relative reward within a gamble (i.e., rela Rabin (1998, pp. 33-34) summarizes the
tive to the amount that could have been experimental findings in this active area of
won). A recent paper (Klaus Fliessbach et psychology: "How do people misperceive
al. 2007) uses MRI techniques to measure their utilities? One pattern is that we tend
the brain activity of pairs of individuals to underestimate how quickly and how fully
engaged in identical tasks. Each individual's we will adjust to changes, not foreseeing
ensuing monetary reward is announced to that our reference points will change ....
both subjects, and both absolute and rela People do not anticipate the degree of such
tive payments were varied. The results with adaptation, and hence exaggerate expected
respect to the ventral striatum show that changes in utility caused by changes in their
relative income is significantly correlated lives."
with blood oxyg?nation in the brain. In fact, If it is indeed the case that people do
brain activity is completely relative in this not fully anticipate changes in reference
respect, as there is no significant role for points, then a wedge will be driven between
27 Nicely summarized in Frank (1999), pp. 140-42.

There is an entire separate literature on health outcomes Daniel J. Wilson 2006), and that suicide and parasuicide
and status; see Marmot (2004) and Lynn Cherkas et al. by the unemployed is actually higher in low unemploy
(2006) for example. Research has also shown that sui ment regions (Stephen Platt, Rocco Micciolo, and Mich?le
cide is a function of relative income (Mary C. Daly and Tansella 1992).

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
happiness and utility.28 The ramifications Mullainathan 2001). Happiness can eas
of systematic errors in anticipating refer ily be manipulated in surveys by reminding
ence group changes are substantial: in eco respondents about something positive or
nomic models with relative utility functions, negative before the question. It is clearly not
it is typically assumed that people perfectly a number that people have on the tip of their
anticipate changes in their reference groups tongue, just waiting to be reported.
(this point reappears in the next section).
Systematic forecast errors of the type Rabin
5. Some Implications for Economic
claims directly and predictably violate the
Theory and Policy Design
rational expectations hypothesis. Frijters,
Haisken-DeNew, and Shields (2002), for The previous sections have discussed evi
example, found that East German respon dence that people value relative outcomes,
dents failed to anticipate in 1991 that their using happiness data (section 3) and nonhap
initial euphoria after German reunification piness approaches (section 4). We now turn
would wear off and therefore structurally to the implications for economic theory and
overestimated their future life satisfaction, policy design of social comparisons and adap
which is consistent with the idea that they tation. Some of these have previously been
failed to realize that their reference position presented in general terms (e.g., Frey and
would adapt to the new situation. Stutzer 2002b, Layard 2005, and Di Telia
Another reason to be cautious about using and MacCulloch 2006), but we here provide
happiness data as a measurement of utility is a more formal and wide-ranging discussion
the argument that there is more to life than of these economic issues. In particular we
happiness. The psychology literature has, for focus on the core areas of economic growth,
example, specifically argued that eudaimonia, labor supply, wage profiles, optimal taxation
which captures functional aspects of well and consumption, savings and investment,
being, plays a separate role to the hedonic and migration.
part of well-being (happiness or life satisfac We will point out in several instances that
tion). These functionings include autonomy, the implications of social comparisons and
competence, personal growth, positive rela adaptation may also result from utility func
tionships, self-acceptance, engagement, and tions without comparisons and adaptation but
meaning (see Edward L. Deci and Richard where there are constraints that are functions
M. Ryan 2000; Carol D. Ryff 1989; Ryff of past and aggregate circumstances. This
and Burton H. Singer 1998; and Martin E. occurs when the effects from reference groups
P. Seligman 2002). In this case, we may well run via an aggregate group outcome (compari
trade off happiness against other constitu sons) or via a function of the past (adaptation),
ent parts of utility, as argued by Kimball and where these reference groups are not explic
Willis (2006). itly identified but rather assumed to underlie
A last reason to mistrust happiness as the observed correlations. In such cases, it is
a measure of utility is the known mal hard to dismiss alternative readings that link
leability of happiness answers (see, for constraints (prices and quantities) to the past
instance, Marianne Bertrand and Sendhil or to aggregate outcomes. Since many prices
and quantities in economics are unobserved
28 It is perhaps worth pointing out here that the gap
(such as the price of home production or indi
between utility and happiness relies on exactly the kind vidual ability or the "fundamentals" of econ
of comparison mechanism that we have appealed to as omies), many models can be proposed with
an explanation for the Easterlin paradox. Choice behav
an unobserved price or quantity generating
ior (based on predicted utility) does not take into account
changes in y*?but these latter are indeed subsequently a relationship between individual behavior
revealed in happiness data. and past actions or group aggregates. Only

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
in some cases can we reasonably argue that of the second subutility function U2(yt \ y*)
reference groups are "needed" to explain as concave, with everyone in the country
unambiguously an empirical regularity; we sharing the same reference income (some
will point out in the applications below when national "norm"), then it is immediate that,
this is the case. at a given level of aggregate income, personal
5.1 Economic Growth consumption, and work choices, a country
with a more unequal income distribution
We start with the ongoing debate about will be less happy on average: the additional
whether economic growth leads to greater status benefit of the individual with one dol
happiness. Easterlin (1974, 1995) and oth lar more than the norm does not compen
ers since (e.g., Lane 2000) have argued that sate the additional status loss of the person
economic growth in Western countries does with one dollar less than the norm. This is
not lead to greater happiness, backed up by an additional rationale for pursuing income
the fact that happiness levels are essentially equality as a policy goal over and above the
flat in Western countries over time (figures 1 usual argument that consumption equality
and 2). Yet, in countries that started out from has welfare benefits due to concavity in the
much lower levels, income growth has been subutility Ui(ct). The effect of economic
associated with modest increases in happi growth on happiness then hinges on the rela
ness (Frijters, Haisken-DeNew, and Shields tionship between growth and inequality.
2004a, Frijters, Haisken-DeNew, and Shields There is also a flip side to the argument
2004b, and Frijters et al. 2006). In terms of that greater economic prosperity at some
the models described in section 3, it can be point ceases to buy more happiness. It can
argued that most developed countries appear be argued that it is actually the concern for
to be at a point of personal consumption ct relative income embodied in the second
where the marginal utility from Ui(ct) is subutility function U2(yt \ y*?) that keeps eco
minimal, while for poorer countries there nomic growth going beyond some wealth
are still gains to be had in Ui(ct) from higher level. The argument here is that relative con
personal consumption. Further economic cerns are more important in rich countries,
growth in developed countries then has little as personal consumption plays an increas
aggregate effect because reference incomes ingly marginal role: status is a luxury good.
increase in line with income, producing no The driving force behind hard work in rich
change in U2(yt\y*t) with higher income. countries, despite high aggregate consump
This explanation for the Easterlin Paradox tion levels, is the concern for status. This is
has been widely adopted (see, for example, indeed one possible evolutionary reason for
Easterlin 1995, 2001; Ruut Veenhoven 1999; having a term U2(yt\y*) in the utility func
McBride 2001; and Layard 2005). Equivalent tion in the first place (Luis Rayo and Gary
formulations are the assertion that at certain S. Becker 2007). This argument has a long
levels of development only conspicuous ancestry in economic debates. Bernard
consumption is important, or that "keeping Mandeville's 1705 "Fable of the Bees" puts
up with the Joneses" is the main economic the argument allegorically. Mandeville jux
motive in rich countries. taposes two hypothetical beehives: one in
However, one possible weak point in this which the bees only care about sustaining
explanation is that it presumes that economic themselves and have no interest in status (i.e.,
growth only affects consumption levels and there is only UY(ct) and U3(T ? lt,Zlt)) and
has no effect on the distribution of income. another where status is what mainly matters
If we relax this assumption, the effect of (i.e., there is mainly U2(yt\y*)). Mandeville
inequality on aggregate happiness also postulates that the first beehive would be
becomes relevant. For example, if we think happy but not very rich, and is ultimately

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
doomed to be taken over by the second bee withstand foreign aggression and economic
hive where the bees are mainly motivated by activity; the ability to attract migrants when
status (by U2(yt | y*)). In that second beehive, income levels are relatively high; and some
the bees would keep working and looking for status utility benefit to a country as a whole
opportunities both within and outside their from having high income compared to other
beehive to further their relative standing, countries. Each of these elements relates to
leading to continual expansion and growth of other literatures which we will only touch
the second beehive. upon in the remainder of the paper.
Mandeville's observations have since been
echoed by many others. Adam Smith, for
5.2 Labor Supply
example, in his Theory of Moral Sentiments, Mandeville and his successors predicted
noted, "To what purpose is all the toil and that labor supply would remain high dur
bustle of this world? ... It is our vanity which ing economic expansions, as a result of sta
urges us on." The modern-day equivalent tus motives. Along the same lines, David
of the argument by Mandeville and Smith Neumark and Andrew Postlewaite (1998)
and many other early economists is made in note that in models where only personal con
theoretical models by Gerhard Glomm and sumption matters, with decreasing marginal
B. Ravikumar (1994), Giacomo Corneo and utility of consumption, we should see falling
Olivier Jeanne (2001), and Thi Kim Cuong aggregate labor supply as aggregate consump
Pham (2005).29 The key aspect of these tion rises, just as the bees in Mandeville's
models is that they specify UY{ct) as ln(ct), first beehive cease to work hard. In the util
and U2(yt\y*t) as ln(kt) - 0 In (/:*), where kt ity function (1) above, however, there is a
denotes wealth instead of income. These limit to the long-run reduction in labor sup
models abstract from the possibility of lei ply with increasing consumption, because
sure, but it is the U2(yt \ y*) part of the utility the relative term U2(yt\y*t) is independent
function in these models that drives contin of consumption. Neumark and Postlewaite
ued economic growth. A related argument argue that status concerns in the income
in Chaim Fershtman, Kevin M. Murphy, happiness relation are the main reason why
and Yoram Weiss (1996) is that social status labor supply has not declined dramatically
is determined in part by relative education, in the twentieth century, despite the very
therefore linking economic growth via edu significant rise in consumption levels. The
cation to status considerations. same conclusions arise if we consider the job,
Normatively speaking, the dominance of rather than the income associated with it, as
the status motive in the income-happiness the carrier of status: here too, labor supply
relationship means that the benefits of eco will be relatively unresponsive to overall con
nomic growth are not to be found in greater sumption levels.
happiness. There are other reasons indi There are of course utility functions with
rectly related to utility that would still pro out relative considerations that are also con
vide a rationale for economic growth, much sistent with labor supply not responding to
in the vein of Mandeville's arguments: the the long-run growth in wages. Examples are
link between the length of life and (aggre utility functions that are log-linear in leisure
gate) income; the link between the ability to and consumption (i.e., Cobb-Douglas util
ity functions in leisure and consumption).
29 Stark (2006b) presents a model in which greater We can object to this alternative by pointing
inequality decreases average social status in a popula out that these do not exhibit any responsive
tion but increases the marginal personal status benefit of
additional income, thereby leading to higher aggregate
ness to wage changes, i.e., labor supply is
incentives to earn more, so that inequality is positively fixed. This is only true for one-period models
correlated with growth. though: we can build in a short-term response

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
to wage changes in such standard models by which signal an abundance of time not used
allowing for borrowing such that individuals to increase Ui(ct). The tax implication is now
would shift labor supply from low-wage to far less clear, as we would ideally want to tax
high-wage periods. Hence it is not necessary all investments into status races equally, so
to resort to relative motives to explain why as to promote nonconspicuous consumption
labor supply has not declined much over the and leisure. Layard (1980) even went so far
past 100 years, even though relative motives as to recognize that one may want to sus
do naturally lead to that observation. tain several status races rather than fewer,
Isolde Woittiez and Kapteyn (1998) and because more races have more winners. The
Maarten CM. Vendrik (1998) point out that question then arises how multiple status
there may be intermediate factors between races can be sustained, all the while counter
relative income concerns and labor supply ing crowding-out effects on nonconspicuous
decisions, such as social norms with respect consumption and leisure. Frijters and Leigh
to appropriate labor supply decisions that (2005) hypothesize that conspicuous leisure
themselves in the long-run are determined is lower in mobile environments because
by the payoffs to a more basic utility func mobility reduces the visibility of leisure
tion. They also argue that female labor sup more than that of consumption. Empirically,
ply has increased in many countries as a they find that U.S. states with higher mobil
result of changing wages, but more slowly ity rates also have higher aggregate levels of
than expected because the associated social investment in conspicuous consumption (i.e.,
norms took time to adjust. Both Neumark and higher labor supply), both at any moment in
Postlewaite (1998) and Yongjin Park (2006) time and through time. The average num
provide empirical tests of female labor sup ber of hours worked per week per working
ply as a function of relative income.30 age person over 1981-2003 in U.S. states
Layard (2005) explicitly argues that the with the highest level of internal mobility is
labor supply implications of income compari about twenty-eight, as compared to twenty
sons provide a rationale for growth-reducing six hours per week for states with the low
taxation designed to bring about greater est level of internal mobility. The authors
leisure. In this context, note that the model advocate mobility taxes to help restore the
in (1) can be extended by supposing that balance between conspicuous consumption
status games may involve not only income, and conspicuous leisure.
but also time investments. Veblen (1899)
5.3 Wage Profiles
recognized this by talking about conspicu
ous consumption and "conspicuous leisure." Kahneman, Knetsch, and Thaler (1991)
This latter consists of all time investments conclude from choice experiments that indi
whose main payoff is demonstrating to other viduals are, at the margin, about twice as
people that one can afford to spend time sensitive to losses as they are to gains; this
on leisure: examples might be showing off is labelled loss-aversion or status quo bias.
(productively) useless skills (such as speak To reflect loss-aversion, we can appeal to a
ing Latin, or playing a musical instrument), specification of the status component of util
ity, U2(yt\y*t), as follows:
30 A related issue is how hard individuals work once

employed: their effective labor supply. A recent paper (6) U2(yt\y*t) = 2f(yt - y*t)
(Clark, David Masclet, and Marie-Claire Villeval 2006)
appeals to both survey and experimental data to show
that effort at work is a function of income relative to that -f(yt - y*t)*i[V,>*}
of one's reference group. Stark and Lukasz Tanajewski t
(2006) appeal to the notion of relative deprivation in the
context of overtime work.
y*t =s=0^ syts,

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Vt Vt + i Ut+ 2
Incomes rising over time
y*t y? + i y*t +2 yi + 3 -^- Time
Incomes decreasing over time
.IT
5
Time -< y* + 3 y*t + 2 y? +1 ?/r
Figure 7. Importance of Kinks in the Utility Function
where the second term reflects the lowerThe opposite occurs in the lower panel,
where the income profile is decreasing. To
marginal utility of income higher than y* rel
make the point that the same number of
ative to income lower than \ft, so that there
is a kink at y\. This reference income itself
dollars are being disbursed in the two profiles,
can be considered as some weighted aver the income figures exactly match vertically.
age of previous incomes. Figure 7 illustratesWith the decreasing profile, the individ
this kind of utility function in the context ofual always receives less than her reference
increasing or decreasing income profiles. For income, which is heavily penalized by this util
ease of illustration, reference income is set to
ity function. Consequently, utility is far lower
equal income in the previous period. under the decreasing income profile than
This figure is read as follows. In the top
under the increasing income profile, despite
panel, where income rises over time, income the actuarial value of the decreasing income
profile being higher: any positive discount
at t exceeds reference income, j/*, so that the
rate will produce higher present-discounted
individual is on the relatively flat part of the
utility function. At time t + 1, the referencevalue from the profile with the higher income
income is now equal to yt, so that the whole first. Note that we do not actually require
utility function shifts to the right. As incomeloss-aversion for this conclusion, which is
driven by the evolution of reference income
at t H- 1 is higher than income at t, the indi
vidual is again on the flat part of the curve.
over time, but that the kinked utility function
As time goes on, the utility function shiftsreinforces the preference for growth. In fact,
further and further to the right. individuals will show similar preferences over

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
two positively-sloped income profiles, where A potentially fruitful avenue for future
loss-aversion plays no role. The steeper profile research along these lines is to test the
will be preferred, ceteris paribus, as income hypothesis that retirement partly results
at each period will be evaluated relative to a from individuals being unwilling to take
lower reference point (a lower past income), a step back within their organization, and
producing higher utility. thus choosing retirement over wage cuts or
Given this preference for income growth, demotion. Retirement would then generi
independent of the income level, employers cally follow the moment at which individual
can save money by offering an increasing productivity peaks, even though workers may
profile with lower actuarial value rather than still have many highly productive years left.
a decreasing profile with higher actuarial This comes about simply because individu
value; individuals prefer the former even if als are loath to accept jobs and wages that
they are perfectly rational. Frank and Robert are below their current reference position.
M. Hutchens (1993) and Loewenstein and In this situation, there is a case for deferring
Nachum Sicherman (1991) use evidence from rewards for production until later in life, i.e.,
small-sample surveys to show that individuals to smooth wages such that they will increase
do indeed express a preference for wage pro up to some fixed age, which in turn raises
files which rise over time, even though these the issue of credible long-term contracts and
have lower present discounted values than mandatory retirement.
alternative profiles with constant or decreasing
5.4 Poverty
wages. Such an observation is very hard, if not
impossible, to square with a fixed utility func The relative importance of the three com
tion that does not depend on past incomes. ponents of the utility function in (1) is cru
Kinks in the utility function around mov cial for the measurement of poverty. One
ing reference points, also termed loss-aver common representation is that individuals
sion, have more implications than simply are in poverty if their material consumption
helping to explain upward-sloping wage levels falls below some subsistence level; this
profiles within firms. We would also expect appeals to a critical level of Ux(ct) rather
employers to be likely to offer contracts than to overall happiness. This approach is
guaranteeing no income reductions over the behind concepts like the absolute poverty
working life, i.e., an endogenous absence of line, the cost of minimum calorie intake
demotion in job titles and institutionalized line, and minimum living standards, such
downward wage rigidity. Ian M. McDonald as the minimum consumption basket defin
(2002) motivates an asymmetric utility func ing the poverty line in the United States, or
tion including loss-aversion, and then argues the 1$ a day poverty line used by the World
in a simple micro-macro model that it creates Bank. However, Amartya Sen (1983) and
downward wage rigidity at the level of firms many others in the poverty literature have
which in turn generates Keynesian business explicitly argued that relative concerns also
cycles. This also fits well with the empiri matter for individuals, and that we should
cal observation of Coen Teulings and Joop base the poverty line on relative rather than
Hartog (1998) that wage decreases are virtu absolute consumption.31 In this vein, the
ally never observed within organizations in OECD publishes statistics on the number
Europe because individuals are sacked rather
than demoted. Andrea Patacconi and Florian
Ederer (2005) also invoke sensitivity to rela 31 One can argue that the concern for relative poverty
results from self-interested insurance against negative
tive decreases in job status to rationalize the shocks, and that the lower tail of the income distribution
lack of observed empirical reductions in rank is informative about the size of the negative shocks cur
and nominal pay within organizations. rently prevalent in the economy.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
of individuals whose income is below half 5.5 Optimal Taxation and Consumption
of median income in member countries,
and the European Union currently employs A burgeoning literature in recent years
a poverty line set at 60 percent of median has addressed the optimal tax implications
income. In terms of equation (1), these rela of utility functions which depend on rela
tive representations of poverty assume that tive income; this literature both relies on
the U2(yt | y*) component of utility is the most and produces predictions about the pre
important for individuals. The measurement cise empirical properties of the income
of poverty via U2 does however pose prob happiness relationship. To our knowledge,
lems once we realise that the norm level of this theoretical literature has not in the past
income, y*, is liable to evolve over time. For been explicitly connected to the empirical
instance, if y* depends on own past income, happiness literature, despite there being
then, at a given level of own current income clear potential gains from such integration
yt, an individual whose income has just (Weinzierl 2005).
increased has higher utility than someone One of the most influential papers on opti
whose income has just decreased, so that mal taxation is Frank (1985), who adopts the
poverty depends on both income levels and following utility function:
income profiles. In practice, taking income
adaptation into account for relative poverty (7) U = U(c0,R(c0),Cl),
measures would seem to be very difficult.
Neither absolute nor relative poverty lines where c0 is the consumption of some posi
introduce any explicit role for the nonmate tional good, potentially including both
rial aspects of utility, and are therefore not positional material goods (conspicuous con
yet based on happiness. To make the distinc sumption) and positional immaterial goods
tion clear between happiness and whatever (conspicuous leisure). This first term of
we currently mean by poverty, think of a (7) corresponds to U^Cf) in equation (1).
factor like sunshine. No known definition of The second term in the utility function,
poverty considers it to be relevant whether R(c0), denotes the individual's consumption
a materially poor person enjoys more hours rank with respect to the positional good: this
of sunshine than a rich person who suf term corresponds broadly to U2(yt\y*). The
fers in a cold climate, even if the materi third term cx denotes a nonpositional good
ally poorer person is happier. Implicitly, and corresponds loosely to U3(T ? lt,Zu).
sunshine and all of the other nonincome The basic point made by Frank (1985) is that
factors influencing happiness are consid utility maximization means that individuals
ered as orthogonal to poverty, even though consume c0 up to the point where total mar
they are highly relevant for both happiness ginal utility is zero:
and individual decision making. Poverty
as currently operationalized concerns the
(8) dU_lR{co)
dc0 dc0 + dR(c0) dU ,.,
dR{c0)
subutilities UY(ct) and U2(yt\y*), instead
of happiness proper (U). A more happi
ness-based poverty measure would take into _ dU dcx
account nonmaterial elements to provide a dci dc<f
broader picture of well-being (the lives that
individuals live), and would also jar less with where ??pCo) is the marginal utility of the
the commonplace observation that individ dc0
uals freely make a number of choices that consumption of the positional good keeping
leave them materially worse off (such as hav dR(c0) dU .
ing children). rank constant, ?:-:?7?H ? is the mar
aco dR{c0)

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
ginal utility of the consumption of the to over-consumption of the observable

positional good via its effect on rank, and good c0 for a wide class of possible informa
tion regimes. He also illustrates the Pareto
?-? is the indirect effect of the increased improvements that can be attained via
dci dc0
income taxation coupled with direct trans
consumption of the positional good via the fers of cl to the poor.
(reduced) consumption of the nonpositional Lars Ljungqvist and Harald Uhlig (2000)
good. The precise form of dcYldc? is given by use a similar utility function, but concentrate
the budget constraint that fixes total income, on changes in optimal tax policy over the
allowing nonpositional goods cY to include business cycle. Their main utility function is:
both leisure and consumption activities.
x. i i i dR(co) dU . (ct - acty-y - 1
Frank emphasizes that ?:-:?7?H ?(10)
is a U 1=- y?-?l,
dc0 dR{c0)
pure externality: changes in rank have no with ct being the population average of indi
social benefit even though they yield private vidual consumption ct, and I denoting labor
benefits. This additional benefit of positional supply. This utility function is analogous
goods to an individual leads to societal over to that in equation (1), albeit with ct and ct
consumption of positional goods, to the det entered as separable functions. The exter
riment of nonpositional goods. Frank then nality embedded in the presence of ct leads
points out that this externality produces a to labor supply that is too high, very much
rationale for the taxation of the positional in the same mould as Michael J. Boskin and
good, in order to promote the nonpositional Eytan Sheshinski (1978) and Frank (1985).
good. If we equate the positional good to Ljungqvist and Uhlig show that this exter
relative income and the nonpositional good nality can be perfectly countered by a con
to leisure, we obtain a rationale for income stant marginal tax on ct, independent of the
taxation in order to promote leisure. Layard business cycle. The analysis is then extended
(2005) adopts this argument.32 by considering the reference position not
A number of other authors have adopted as ct, but as ?Vi, i.e., average consumption
different specifications of the utility func last period. In the presence of aggregate
tion, and of the reference position in par productivity shocks, they find that optimal
ticular, which affect the tax implications. For tax rates co-move with current productiv
example, if we take the utility function in ity shocks, creating countercyclical effects
Norman J. Ireland (1994): of taxation on the economy. This is exactly
in line with the usual Keynesian optimal tax
(9) 17 = UifiwJAco)), policy which is also countercyclical.
5.6 Savings and Investment
where cx is a good whose consumption is
unobservable, and s(c0) is status, specified A related recent theoretical literature has
as the belief spectators have about f(c0,c?) emerged on the dynamic effects of relative
based on observing c0. Exactly as in Frank consumption (or status concerns). The papers
(1985), Ireland derives a general tendency we mention here abstract from the possibil
ity of a term like U3{T ? lt,Zlt) and focus in
32 An older literature argues that "social preferences" the main on the timing of consumption. The
(including altruism) can only be identified from observed key question addressed is the optimality of
transfers under restrictive assumptions (direct utility
measures do not suffer from this drawback); see Oswald savings.
(1983) with respect to taxation and Yew-Kwang Ng (1987) We might naively think that status does not
for the related problem of public good provision. affect savings, because the trade-off between

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
i
Low reference consumption

High reference consumption
0<a< 1
Low reference consumption

High reference consumption
Figure 8. Utility of Consumption with High/Low Reference Consumption, and
current and future status would seem toisbe

Here ct own consumption and ct is the
identical to the trade-off betweengeometric
currentmean of the consumption of a
and future consumption. However,reference
this linegroup, which can be construed
of thinking breaks down when weasconsider
the rest of the population or some slowly
that individuals accumulate wealth over
adjusting social norm containing past gen
their lifetime, and that productivityerations'
generallyor the individual's own previous
increases, implying that in a stylized
consumption.
sense The parameter r/, 0 < r? < 1,
reference income when "old" is always higher
denotes a kind of "weight" for relative con
than that when "young." In this case,
cernsrelative
in individual utility and a > 0 reflects
concerns come into play through the mar (a = 0 implies risk-neutrality).
risk-aversion
ginal utility of consumption over theThe
lifecycle
key characteristics of this utility function
are revealed when we consider that the rela
and, therefore, affect saving and investment
decisions. Most of the generic arguments
tionship between the individual's marginal
that arise here can be illustrated via thefrom
utility util own consumption and reference
ity function introduced by Andrew B. Abel
income ct depends crucially on a. The main
(1990), and subsequently adopted by a num are depicted in figure 8, where
possibilities
ber of other authors: the x axis shows personal consumption and
the y axis utility.
(c)(l-i7)(l-a) fc \r,(l-a)
(11) ut
1 - a= ^-* ^)
\ctJ
The two lines in the top panel of figure 8,
where 0 < a < 1 (so that U is positive), show

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
that a higher value of reference consump lives, when the consumption of others will
tion (the thin line) reduces not only utility, also be higher due to productivity increases.
but also the marginal utility from own con In the terminology of Bill Dupor and
sumption, i.e., at every consumption level Wen-Fang Liu (2003), the case with a > 1
the curve is flatter with higher reference can be called "Keeping up with the Joneses"
consumption. As such, when 0 < a < 1, indi and that with a < 1 "Running away from the
viduals will want to consume more when ref Joneses."33 Crucially, the issue of whether
erence consumption is lower: individuals will the parameter a is in fact greater than one or
plan consumption in the periods when other not can in principle be empirically evaluated
individuals are not consuming, as their mar using happiness data, although this test has
ginal utility of another dollar of consumption not to our knowledge been carried out to
will be greater. This creates a coordination date. A simple method of obtaining the sign
problem, as reference consumption results of (a - 1) results from the cross-derivative of
from simultaneous choices by everyone in equation (11):
the economy (for example if the reference
point refers to average consumption by oth
ers). The difficulty in solving such coor (12) act
^- act
= V(a - l)(ctya5r(1'ayi
dination problems in endogenous growth
models has to date appeared unsurpassable
(Ljungqvist and Uhlig 2000, and Abel 2005, It follows from (12) that the sign of the inter
simply assume that a > 1). Even so, it seems action between own consumption and refer
intuitively plausible to imagine that the mar ence group consumption in a reduced-form
ginal (status) utility of consumption is higher regression reveals the sign of (a - 1).
when the consumption of others is lower. Abel (2005) focuses on this issue in an over
Intuitively also, we might think that when 0 lapping two-generation model with a utility
< a < 1, savings will be too low: individu function for the new-born of the form:
als will not postpone consumption to the
future as future general consumption levels
then will kkely be higher due to productivity (13) /cUt = ?-^?^
) (1-^(1-*) *(c/c?jKi-?)
I ? a
growth.
We obtain exactly the opposite results when
a > 1, corresponding to the lower panel of fig | (ct+?*-r')<-1-aHct+l/et+?vll-a)
ure 8, which is the dominant assumption made
in the literature. Here again, individual utility
falls as reference consumption rises, but now
the marginal utility of consumption increases 33 There is an interesting analogy here with models of
habit formation explaining unemployment persistence in
(roughly speaking, the curve shifts to the macro and labor economics (for a review, see William A.
right, so that that its slope is steeper at any Darity and Arthur H. Goldsmith 1996). The generic argu
given level of consumption). Individuals will ment in this literature is that the unemployed get used
now want to consume more when others con to being unemployed, either via becoming discouraged
(which is a form of adaptation to own circumstances) or via
sume more, producing a kind of herding phe social norms (such as when they conform to the neighbor
nomenon: status is then a bandwagon good in hood; see Clark 2003 for an empirical test). In these mod
els, individuals become less keen on formal work when
the terminology of Duesenberry (1949). There they or their reference group have been out of work for a
is no difficult coordination issue to solve as all
long time and therefore become "locked" into unemploy
consumers will want to consume at the same ment. This corresponds closely to the notion that people
"keep up with the Joneses" rather than "run away from the
time. The corollary is that individuals all save Joneses," for in the latter case they would be more keen
too much at the same time, because they all to have a job when they have been unemployed for a long
want to consume more toward the end of their time or when their reference group is unemployed.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
where ct is own current consumption and ct that there will be upward pressure from tech
is the geometric mean of the current con nological growth on savings, as individuals
sumption of everyone else alive. The term anticipate that the marginal return to future
(Ci)(1-w-a) refers to the part of own con consumption will be higher. However, in
sumption that is independent of others' con the specification of the reference position ct,
sumption, and the term following ? refers there is another effect, related to the speed at
to future consumption. The main result which the reference point adjusts to current
emerges when we consider the marginal util consumption. When the rate of adaptation to
ity of future consumption: consumption changes, p, is high, individuals
essentially only compare to their own recent
consumption, and when p is small (although
(14) ^- = ?(ct+1)-a^ta\
dct+l always positive), adjustment is slow and con
sumption in the distant past remains impor
tant. Again, the value of this critical variable,
When a < 1, we obtain the intuitive result the speed of adjustment, can in principle be
discussed above: the marginal utility from measured in happiness regressions by the
future consumption is lower when ci+1 is coefficients of past consumption on current
higher, and therefore when there are general happiness.
productivity increases. The externality from Carroll, Overland, and Weil (2000) also
higher future consumption reduces savings, show that when adaptation is slow, it makes
which makes the case for subsidies on sav sense for individuals to save more as eco
ings. Again though, the equilibrium under nomic growth increases, despite the fact
a < 1 cannot be solved analytically because that future reference consumption levels
of the coordination issue mentioned above. will be higher. The intuition is that under
Abel thus mainly concentrates on cases with slow adjustment, individuals wish to "smooth
a > 1, under which general productivity out income increases" more than under fast
increases, leading to higher future reference adjustment. On the contrary, if the refer
incomes ct+i, actually increase the future ence position adjusts quickly, individuals
marginal return to consumption, yielding a essentially want to enjoy the status benefit of
case for taxes on savings. higher productivity immediately. This model
Christopher D. Carroll, Jody Overland, is used to rationalise the empirical regular
and David N. Weil (1997, 2000) make a dif ity that high-growth countries also have high
ferent point by adopting a utility function savings rates. It is difficult to reconcile this
where the reference position only depends empirical fact with a model without reference
on the consumption of the individual herself positions affecting utility: in the latter, the
in the past: marginal utility of future consumption is
always lower as economic growth increases
(15) Ut =1?-* f*
? a \c
because future consumption is higher, lead
ing to lower savings rates. Why save now if
we are all going to be rich tomorrow anyway?
Carroll et al.'s answer is that individuals want
ct =J [?oce-^-% ds. to adjust slowly to ever higher consumption
patterns, and it therefore makes sense to
This is the same utility function as Abel save more now, simply to avoid getting used
(1990, 2005) but with what we called an to high consumption too soon. This provides
"internal reference" point in section 3. the link between savings and the speed of
Carroll, Overland, and Weil also concentrate adaptation of reference consumption in hap
on cases with a > 1 and generically argue piness regressions.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
George M. Constantinides (1990) and crisis the risk premia for holding a financial
Jordi Gali (1994) use a similar utility func asset went up in all of the countries involved
tion to address the equity premium puzzle. when the currency of another country col
Constantinides notes that the existence of lapsed, whereas in standard theories one
internal reference points over consumption would expect capital flight from the affected
provides an additional reason for individu country to the other countries which would
als to only slowly adjust their consumption reduce risk premia. Melisso Boschi and
over time. Angus Deaton (1992) confirms Aditya Goenka (2006), who adopt the util
this prediction by showing in U.S. data that ity function of Cambell and Cochrane (1999)
individuals over-smooth consumption after which is very similar to Abel's (1990) speci
permanent income shocks. Constantinides fication, argue that this increase in risk pre
goes further by arguing that the presence of mia may be due to the fact that if own wealth
ct creates a bias in conventional estimates of comes close to the reference wealth level (due
risk-aversion, which are based on the trad to losses incurred in another country) the
eoffs people make over time: without ct then curvature of the utility function increases.
high risk-aversion would imply that the sav The greater the curvature of the utility func
ings rates of the rich would be greater than tion, the more compensation investors needs
those of the poor. The fact that savings rates to keep investing in a country, meaning that
vary only little across income groups may the risk premia go up which in turn may lead
lead us to conclude that risk aversion is actu to the collapse of a currency. This general
ally quite low. Constantinides shows that idea can also be seen in figure 8: the "steep
the presence of ct breaks this logic, in that ness" reduces faster when reference incomes
even with high risk aversion the presence are relatively high relative to own income
of a moving reference point leads all indi and thus risk aversion is stronger. The same
viduals to smooth income over time in a principle applies with wealth if individuals
similar manner. This provides a rationale for realize that making a loss on current wealth
empirical instances of observed high instan may reduce consumption below reference
taneous risk-aversion, in particular the large consumption. Whilst Boschi and Goenka
premium that individuals are prepared to (2006) claim such increases in risk premia
pay for risk-free assets versus risky higher due to wealth effects cannot be explained by
return assets, i.e., the equity premium puz standard (CARA) utility functions, it may of
zle. John Y. Campbell and John H. Cochrane course be the case that financial contagion
(1999) extend this idea to further types of works via channels other than wealth effects
asset pricing puzzles and argue that models in combination with habit formation. We
including a reference point ct exhibit supe could for instance alternatively argue that
rior predictive power over models without the collapse of one country is informative
such a term. about "fundamentals" in a similar country.
Reference point models have also been This is therefore a good example of a situation
used to explain financial contagion where where reference group effects may explain
reference effects are usually termed "habit an observable outcome but where it is hard to
formation." The main observation that this dismiss other possibilities that do not involve
literature addresses is the Asian financial reference groups; there are indeed dozens
crisis, whereby a whole set of countries saw of other competing theories (see the lengthy
their exchange rates and economies collapse discussion in Boschi and Goenka 2006). This
in sequence (for a description, see Barry underlines the importance of laboratory evi
Eichengreen, Galina Hale, and Ashoka Mody dence on the influence of reference groups
2001). A puzzling feature for more standard for this literature, since only then can we be
models is that during the Asian financial sure that there are no other factors involved

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
and that reference groups really do have interest rate of r) links period 1 consumptio
independent effects on choice behavior. to period 2 income: the higher is period
One particular strand of the empiri income, the higher is period 1 consumption
cal income-happiness literature fits in well because of a reduced need for savings. No
with savings models, namely the so-called imagine that period 2 income is in fac
"tunnel-effect." The original idea is attrib unknown by the individual decisionmaker,
uted to Albert O. Hirschman and Michael who estimates her next-period income from
Rothschild (1973), who argued that individu the incomes of "reference" individuals aroun
als could actually derive utility from others' her who share the same observable char
higher incomes if they consider them to be acteristics (region, education, gender, age
informative about their own future income. etc.). This produces a reduced-form lifetime
Senik (2004) uses this argument to explain happiness function in the first period that
why in Russian panel data (RLMS) individ depends positively on the observed incomes
ual happiness was positively linked to refer of reference individuals. Neither adaptation
ence group income, rather than negatively nor social comparisons are needed to pro
as a relative utility function would suggest. duce a relationship between happiness and
Equally, Senik (2005) finds that higher ref others' income; although equally, under cer
erence group income reduces life satis tain parameterizations, tunnel effects ca
faction in Western Europe, but raises life be observed even if relative income or con
satisfaction in the posttransition countries of sumption matters.
Eastern Europe (and the United States). The The tunnel-effect model provides several
potential importance of the information role pointers as to what we should expect t
is underscored by the finding that reference observe empirically: (i) we should see a
group income is more strongly positively positive relationship between reference grou
correlated with life satisfaction for those in income and own current consumption, over
more uncertain situations (as measured by and above the effect of own income (becaus
the volatility of their income and the prob the higher is others' income, the lower are
ability of losing their job, for example). one's own savings); and (ii) we should observ
The simplest model in which a tunnel happiness being positively correlated with
effect can arise is a two-period model where reported expected future income, and that
individuals only derive utility from their own the positive effect of reference group incom
consumption, but face the problem of saving on one's own happiness transits via expected
in period 1 in order to consume in period 2. future income. These tunnel-effect predic
For example, consider: tions are yet to be empirically tested.
5.7 Migration
(16) U = u(cl) + u(c2)
Consider the decision whether to migrate
, C2 , !/2 or not. Without a U2(yt | y*t) term in the utility
1 1 + r i/1 1 + r function, all those who find more attractiv
income and leisure combinations in another
dc country will leave. This conclusion changes if
w'(ci) = (1 + r)u'(c2)->-r?> 0, comparisons, and consider that
dy2 we allow for
migration might lead to changes in y*t. For
where lifetime utility U is now simply the
instance, if y* equals average income in the
local neighborhood or the average income
sum of happiness in period 1, which depends
of people like yourself at your workplace,
only on period 1 consumption, and happiness
in period 2, which depends only on period
then 2someone who fears ending up with low
consumption. The budget constraint (with an income in another country might no
relative

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
migrate, even if both leisure and the abso high-status assimilators, who Stark identi
lute income she could earn there are higher. fies with the brain-drain phenomenon. The
Stark and J. Edward Taylor (1991) appeal to networks of these early migrants may lead to
this idea to explain why the elites in poor more migration of the same variety, but the
countries do not emigrate: the elite are at the fast assimilation of early migrants implies
top of the income distribution in the coun that they are not preoccupied by this, as
try where they currently live, but may well their networks will not primarily consist of
not be so if they emigrate. Stark and Taylor other migrants from the same home country.
further introduce the notion of different eco The second migration pattern is very differ
nomic migrant types. In terms of equation ent, and may well result from exceptionally
(1), they argue that we would expect those low consumption in the home country (i.e.,
with relatively high earnings potential in by U^Cf)). For example, we can think of the
another country to move abroad, e.g., those Irish in the United States being driven by
whose skills are relatively undervalued in the the potato famine back home, forced migra
country of origin. Furthermore, we expect tion in general, or the "guest worker phe
those who can keep individuals in their home nomenon" of the European Union where
(poorer) country as their reference group to whole villages were essentially transplanted
be more likely to emigrate than those whose to other countries in the 1960s. These indi
reference income adapts to the new country. viduals will by design be less likely to assim
The former can become high-status in their ilate, and have strong incentives to carry
home country by earning more in the host on comparing themselves to individuals in
country, yet they remain in a status sense part the home country. This results from their
of the home country. This line of thinking can high wages relative to those in the home
help to explain why many migrants continue country and low wages relative to the host
to visit their home country: this is when they country. These migrants may for the same
can cash in as relatively high earners com reason try to attract more low-skill individu
pared to those in the home country, making als from the home country, as these latter
it irrelevant whether those same incomes are do not detract from their own status but
considered as high in the host country.34 rather increase it by reducing the reference
This kind of analysis yields two distinct income they face in the host country. The
possible migration dynamics. In the first, two types of immigration, associated with
those who migrate do so voluntarily in two different parts of the income/happiness
spite of adaptation in the U2(yt\y*) element gradient, will therefore have very different
of their utility function. That is, they have cultural and economic implications for the
exceptionally high skills and can become host country. Countries such as Canada and
high-status even in the host country and Australia, which operate a points system
want to compare themselves to persons in whereby potential migrants have to offer
the host country anyway. The first group of something exceptional to the host country
voluntary migrants is therefore made up of in order to attain a visa, arguably attempt to
34 It could be argued that the same observation could it is really an issue of prices, there would be no inherent
be rationalized by a standard utility function without reason for migrants to return to their home country: they
relativity in a situation where prices are low in the home could equally go to any other low-price country. We can,
country, but wages are high in the host country. Migrants of course, object to this line of thinking by saying that
going home for the holidays would then simply be taking migrants return to their home country because of specific
advantage of the higher purchasing power of their income ties with family or others that further lowers the price of
in the home country. That would then, however, beg the some goods for them. It is possible to derive alternative
question as to why the nonmigrants of that host country rationalizations of return migration that do not depend on
do not also visit the "home" country in large numbers. If relative considerations.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
attract the first type of economic migrant. and Rabin (1998): Should we only care about
Countries in the European Union that are what Self 0 (who makes the decisions) wants
introducing legislation trying to stop "fam or should we also care about what Self 1 (in
ily reunions" could be interpreted as try the future) experiences? Individuals who care
ing to reduce the second kind of economic only about their desires at time 0 (i.e., Self 0)
migrant. More empirical work on the rela will take decision utility as the normative ref
tionship between happiness and reference erence point. If instead we were to take the
incomes, especially looking at the differ experiences of Self 1 as the guiding principle
ences across migrant groups in reference for policy design, we would potentially act
incomes, would greatly inform this debate. against the explicit wishes of an "ignorant"
(Self 0) electorate by taxing activities that
5.8 Normative Implications
lead to only short-run happiness gains. An
Section 5.5 discussed the tax implications interesting political economy question then
of relative utility. These arise due to exter arises of how governments can do so with
nalities between individuals. Somewhat more out being voted out of office by myopic Self
subtle are the policy implications of adapta 0's. This question only arises when individu
tion, which can be thought of as externali als misforecast their degree of adaptation.
ties within individuals. We discussed above However, the empirical literature is still only
how changing reference incomes can affect beginning to grapple with the questions of
intertemporal trade-offs in consumption, differential paces of adaptation to life events.
wage profiles, and the costs and benefits of While there is now growing evidence regard
migration. In general, events to which indi ing the misprediction of adaptation (see Jason
viduals adapt quickly only have a happiness Riis et al. 2005 and the research reviewed in
payoff in the short run, while events where Loewenstein and Peter A. Ubel 2006), the
adaptation is slow (or absent), have long-run information required to advocate paternal
happiness payoffs. The impact of a particular ism is currently far from complete.
circumstance on current happiness is then
only a snapshot of the stream of effects on 6. Conclusions
happiness associated with that circumstance,
and is therefore in principle not necessarily One of the exciting developments within
informative about life-time trade-offs. economics over the last decade has been the
What is often not well understood is that booming "economics of happiness" litera
as long as individuals are rational, the mere ture, which has expanded in both theoreti
presence of adaptation is no reason for policy cal and empirical directions. The basis for
intervention unless it is accompanied by an the empirical work has been the increasing
externality (such as those found in social evidence from both psychologists and econo
comparisons). We may not necessarily want mists that measures of individual well-being
to counteract activities which produce only a collected in surveys contain "valid variabil
short-lived happiness "buzz" at the expense of ity," in the sense that current happiness or
a long-run happiness cost, unless we believe satisfaction is a strong predictor of future
that individuals are unaware of the fact that behavior. The wider acceptance of subjective
the "buzz" may indeed only be ephemeral. well-being measures as a direct proxy for util
When individuals do not in fact realize that ity has consequently opened up a wide range
they will get used to some things (but not to of opportunities to further inform theory
others), a basic paternalistic question arises. and policy design. The happiness literature
This has been well-stated by many, includ has in the last few years began to make major
ing Kahneman, Wakker, and Sarin (1997), inroads in this respect. The analysis of sub
Kahneman, Diener, and Schwarz (1999), jective well-being data provides a valuable

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
alternative, but complementary, approach reasons why we believe the two are related
to the revealed-preference framework that and review evidence consistent with relative
dominates the discipline of economics. utility from nonhappiness sources.
The two specific issues that have generated Going beyond the paradox that initiated
the most interest in the literature are (i) the the literature, this paper has attempted to
effect of labor market status, and especially connect the economics of happiness literature
unemployment, on happiness, and (ii) the rela with theoretical economic models of taxa
tionship between income and happiness. This tion, labor supply, economic growth, savings,
paper has focussed on the latter, motivated wage profiles, migration, and consumption.
by its central importance to economists and We have identified how the outcomes of
policymakers. Our contribution has been to mainstream theoretical models hinge on key
provide a new overview of the theoretical and behavioral parameters that could in princi
empirical literature on income and happiness, ple be identified from the empirical analysis
bringing together the most recent research, of happiness data. Some examples of these
and showing how the traditional utility func parameters include:
tion framework can be readily generalised to
incorporate a range of observed behaviors. 1. The degree of risk-aversion and the com
In particular, our main challenge has been plementarity between own income and
to align the widely cited Easterlin "Paradox," reference income, which are important for
that of increasing real incomes in developed the savings literature;
countries but with no noticeable increase in 2. The malleability of reference groups,
average happiness, with the large empirical which is key to migration decisions and
survey literature that has found that income education decisions;
and happiness are positively related. 3. The kink in utility functions around the
The broad consensus in the literature is reference position, which is important for
that the paradox points to the importance of wage policies and career decisions; and,
relative considerations in the utility function, 4. The existence and extent of material and
where higher income brings both consump nonmaterial status races, which are para
tion and status benefits to an individual. mount for optimal taxation policy.
Comparisons can either be to others or to
oneself in the past. Utility functions of this The interaction between economic the
type can explain the positive slope found in ory and happiness is therefore the next
much of the empirical literature. However, milestone for the developing economics of
since status is a zero-sum game, only the happiness literature. However, it is clear that
consumption benefit of income remains at the empirical literature on happiness still
the aggregate level. Since the consumption faces several challenges, many of which are
benefit approaches zero as income rises, hap shared with other empirical literatures. Two
piness profiles over time in developed coun of the key challenges are to deal with a gen
tries are flat. Carlyle's pitifulest whipster will eral inability of survey data to precisely time
indeed be made happier by higher income, changes in income with changes in happi
but only at the expense of someone else or ness over long time periods, and the diffi
his own future self. culty in mapping incomes into current and
We have appealed to the growing litera expected consumption. It is also the case
ture to show that happiness is indeed nega that most datasets do not contain reliable (if
tively related to others' incomes and to own any) ex ante information regarding the group
past income. We are aware though that it will (the reference point) to which individuals
never be possible to prove that happiness compare themselves. Similarly, no dataset
measures utility. We thus also discuss the can contain all the variables of importance,

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
so that researchers will continue to face the Alpizar, Francisco, Fredrik Carlsson, and Olof Johans
son-Stenman. 2005. "How Much Do We Care about
issue of endogeneity with respect to income Absolute versus Relative Income and Consump
and other variables such as marriage, educa tion?" Journal of Economic Behavior and Organi
zation, 56(3): 405-21.
tion, and the reference group. Finally, natural
Angner, Erik. 2005. "The Evolution of Eupathics: The
experiments producing exogenous variation Historical Roots of Subjective Measures of Well
in income are only rarely observed, making Being." Unpublished.
the issue of establishing the causal effect of Barker, David J. P. 2005. "The Developmental Origins
of Well-Being." In The Science of Well-Being, ed.
income on happiness a major challenge. F. Huppert, N. Baylis, and B. Keverne. Oxford and
Our final conclusion is that taking relative New York: Oxford University Press, 59-73.
Bertrand, Marianne, and Sendhil Mullainathan. 2001.
income seriously is an important step toward
"Do People Mean What They Say? Implications
greater behavioral realism in Economics, for Subjective Survey Data." American Economic
such that our models and empirical analy Review, 91(2): 67-72.
sis move closer to how real people feel and Blanchflower, David G., and Andrew J. Oswald. 2004.
"Well-Being over Time in Britain and the USA."
behave. Some may not like the insertion of Journal of Public Economics, 88(7-8): 1359-86.
additional arguments into individual utility, Blanchflower, David G., and Andrew J. Oswald. Forth
and remark that any behavior can be ratio coming. "Hypertension and Happiness across
Nations." Journal of Health Economics.
nalized by an appropriate manipulation of Boes, Stefan, Markus Lipp, and Rainer Winkelmann.
the utility function. While this is formally 2007. "Money Illusion under Test." Economics Let
ters, 94(3): 332-37.
true, it does not apply wholesale to the issue
of relative income. As we have tried to dem Boschi, Melisso, and Aditya Goenka. 2006. "Habit
Formation and the Transmission of Financial Cri
onstrate, utility functions including relative ses." University of Essex Working Paper, no. 608.
income terms produce a wide variety of test Boskin, Michael J., and Eytan Sheshinski. 1978. "Opti
mal Redistributive Taxation When Individual Wel
able predictions regarding both well-being fare Depends upon Relative Income." Quarterly
(measured by survey or neurologically) and Journal of Economics, 92(4): 589-601.
observable behaviors: it is not true that "any Brickman, Philip, and Donald T. Campbell. 1971.
"Hedonic Relativism and Planning the Good Soci
thing goes." To our mind, this is precisely why ety." In Adaptation-Level Theory: A Symposium, ed.
we need to appeal to both direct measures M. H. Appley. New York: Academic Press, 287-302.
of utility and observed behavior in order to Brickman, Philip, Dan Coates, and Ronnie Janoff-Bul
man. 1978. "Lottery Winners and Accident Victims:
obtain a better idea of what the utility func Is Happiness Relative?" Journal of Personality and
tion looks like, and make policy recommen Social Psychology, 36(8): 917-27.
Brosnan, Sarah F, and Frans B. M. De Waal. 2003.
dations in the best interest of society. Testing
"Monkeys Reject Unequal Pay." Nature, 425(6955):
these predictions not only allies theory and 297-99.
empirical analysis in economics, it also spills Brown, Gordon D. A., Jonathan Gardner, Andrew J.
across many disciplines in the social and nat Oswald, and Jing Qian. Forthcoming. "Does Wage
Rank Affect Employees' Wellbeing?" Industrial
ural sciences; it is arguably the most impor Relations.
tant and the most promising of the research Burchardt, Tania. 2005. "Are One Man's Rags another
avenues open to this thriving literature. Man's Riches? Identifying Adaptive Expectations
Using Panel Data." Social Indicators Research,
74(1): 57-102.
References Campbell, John Y., and John H. Cochrane. 1999. "By
Force of Habit: A Consumption-Based Explanation
Abel, Andrew B. 1990. "Asset Prices under Habit For of Aggregate Stock Market Behavior." Journal of
mation and Catching Up with the Joneses." Ameri Political Economy, 107(2): 205-51.
can Economic Review, 80(2): 38-42. Cappelli, Peter, and Peter D. Sherer. 1988. "Satisfac
Abel, Andrew B. 2005. "Optimal Taxation When Con tion, Market Wages, and Labor Relations: An Airline
sumers Have Endogenous Benchmark Levels of Study." Industrial Relations, 27(1): 57?73.
Consumption." Review of Economic Studies, 72(1): Carroll, Christopher D., Jody Overland, and David N.
21-42. Weil. 1997. "Comparison Utility in a Growth Model."
Alesina, Alberto, Rafael Di Telia, and Robert J. Mac Journal of Economic Growth, 2(4): 339-67.
Culloch. 2004. "Inequality and Happiness: Are Carroll, Christopher D., Jody Overland, and David N.
Europeans and Americans Different?" Journal of Weil. 2000. "Saving and Growth with Habit Forma
Public Economics, 88(9-10): 2009-42. tion." American Economic Review, 90(3): 341?55.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Carroll, N., Paul Frijters, and Michael A. Shields. Primate Striatum." Experimental Brain Research,
Forthcoming. "Quantifying the Costs of Drought: 162(4): 520-25.
New Evidence from Life Satisfaction Data." Journal Daly, Mary C, and Daniel J. Wilson. 2006. "Keeping
of Population Economics. Up with the Joneses and Staying Ahead of the Smiths:
Chen, M. Keith, Venkat Lakshminarayanan, and Laurie Evidence from Suicide Data." Federal Reserve Bank
R. Santos. 2006. "How Basic Are Behavioral Biases? of San Francisco Working Paper, no. 2006-12.
Evidence from Capuchin Monkey Trading Behavior." Darity, William A., Jr., and Arthur H. Goldsmith.
Journal of Political Economy, 114(3): 517-37. 1996. "Social Psychology, Unemployment and Mac
Cherkas, Lynn, Abraham Aviv, Ana Valdes, Janice roeconomics." Journal of Economic Perspectives,
Hunkin, Jeff Gardner, Gabriela Surdulescu, 10(1): 121-40.
Masayuki Kimura, and Tim Spector. 2006. "The Davidson, Richard J. 2004. "Well-Being and Affective
Effects of Social Status on Biological Aging as Style: Neural Substrates and Biobehavioural Corre
Measured by White-Blood-Cell Telomere Length." lates." Philosophical Transactions of the Royal Soci
Aging Cell, 5(5): 361-65. ety B: Biological Sciences, 359(1449): 1395-1411.
Clark, Andrew E. 1996a. "Job Satisfaction in Brit Deaton, Angus. 1992. Understanding Consumption.
ain." British Journal of Industrial Relations, 34(2): Clarendon Lectures in Economics. Oxford; New
189-217. York; Toronto and Melbourne: Oxford University
Clark, Andrew E. 1996b. "L'utilit? est-elle relative? Press, Clarendon Press.
Analyse ? l'aide de donn?es sur les m?nages." Econo Deci, Edward L., and Richard M. Ryan. 2000. "The
mie et Pr?vision, 121: 151-64. 'What' and 'Why' of Goal Pursuits: Human Needs
Clark, Andrew E. 1997. "Job Satisfaction and Gender: and the Self-Determination of Behavior." Psycho
Why Are Women So Happy at Work?" Labour Eco logical Inquiry, 11(4): 227-68.
nomics, 4(4): 341-72. Di Telia, Rafael, and Robert J. MacCulloch. 2005.
Clark, Andrew E. 1999. "Are Wages Habit-Forming? "Gross National Happiness as an Answer to the
Evidence from Micro Data." Journal of Economic Easterlin Paradox?" Unpublished.
Behavior and Organization, 39(2): 179-200. Di Telia, Rafael, and Robert J. MacCulloch. 2006.
Clark, Andrew E. 2001. "What Really Matters in a Job? "Some Uses of Happiness Data in Economics." Jour
Hedonic Measurement Using Quit Data." Labour nal of Economic Perspectives, 20(1): 25-46.
Economics, 8(2): 223-42. Di Telia, Rafael, John Haisken-DeNew, and Robert
Clark, Andrew E. 2003. "Unemployment as a Social J. MacCulloch. 2007. "Happiness Adaptation to
Income and to Status in an Individual Panel." NBER
Norm: Psychological Evidence from Panel Data."
Journal of Labor Economics, 21(2): 323-51. Working Paper, no. 13159.
Clark, Andrew E., Fabrice Etil?, Fabien Postel-Vinay, Di Telia, Rafael, Robert J. MacCulloch, and Andrew J.
Claudia Senik, and Karine Van der Straeten. 2005. Oswald. 2003. "The Macroeconomics of Happiness."
"Heterogeneity in Reported Well-Being: Evidence Review of Economics and Statistics, 85(4): 809?27.
from Twelve European Countries." Economic Jour Diener, Ed, Marissa Diener, and Carol Diener. 1995.
nal, 115(502): C118-32. "Factors Predicting the Subjective Well-Being of
Clark, Andrew E., Yannis Georgellis, and Peter Sanfey. Nations." Journal of Personality and Social Psychol
1998. "Job Satisfaction, Wage Changes, and Quits: ogy, 69(5): 851-64.
Evidence from Germany." In Research in Labor Diener, Ed, and Richard E. Lucas. 1999. "Personal
Economics. Vol. 17, ?d. S. W. Polachek. Stamford, ity and Subjective Well-Being." In Well-Being: The
Conn, and London: JAI Press, 95-121. Foundations of Hedonic Psychology, ed. D. Kahne
Clark, Andrew E., Yannis Georgellis, and Peter Sanfey. man, E. Diener, and N. Schwarz. New York: Russell
2001. "Scarring: The Psychological Impact of Past Sage Foundation, 213-29.
Unemployment." Econ?mica, 68(270): 221-41. Duesenberry, James S. 1949. Income, Saving, and the
Clark, Andrew E., David Masclet, and Marie-Claire Theory of Consumer Behavior. Cambridge and Lon
Villeval. 2006. "Effort and Comparison Income: don: Harvard University Press.
Experimental and Survey Evidence." IZA Discus Dupor, Bill, and Wen-Fang Liu. 2003. "Jealousy and
sion Paper, no. 2169. Equilibrium Overconsumption." American Eco
Clark, Andrew E., and Andrew J. Oswald. 1996. "Satis nomic Review, 93(1): 423-28.
faction and Comparison Income." Journal of Public Easterlin, Richard A. 1974. "Does Economic Growth
Economics, 61(3): 359-81. Improve the Human Lot? Some Empirical Evidence."
Cohen, Sheldon, William J. Doyle, Ronald B. Turner, In Nations and Households in Economic Growth:
Cuneyt M. Alper, and David P. Skoner. 2003. "Emo Essays in Honor of Moses Abramovitz, ed. R. David
tional Style and Susceptibility to the Common Cold." and M. Reder. New York: Academic Press, 89-125.
Psychosomatic Medicine, 65(4): 652?57. Easterlin, Richard A. 1995. "Will Raising the Incomes
Constantinides, George M. 1990. "Habit Formation: A of All Increase the Happiness of A\\?" Journal of Eco
Resolution of the Equity Premium Puzzle." Journal nomic Behavior and Organization, 27(1): 35-47.
of Political Economy, 98(3): 519-43. Easterlin, Richard A. 2001. "Income and Happiness:
Corneo, Giacomo, and Olivier Jeanne. 2001. "Status, Towards an Unified Theory." Economic journal,
the Distribution of Wealth, and Growth." Scandina 111(473): 465-84.
vian Journal of Economics, 103(2): 283-93. Easterlin, Richard A. 2005a. "A Puzzle for Adaptive
Cromwell, Howard C, Oum K. Hassani, and Wol Theory." Journal of Economic Behavior and Orga
fram Schultz. 2005. "Relative Reward Processing in nization, 56(4): 513-21.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Easterlin, Richard A. 2005b. "Feeding the Illusion of Frijters, Paul. 2000. "Do Individuals Try to Maximize
Growth and Happiness: A Reply to Hagerty and Veen General Satisfaction?" Journal of Economic Psy
hoven." Social Indicators Research, 74(3): 429-43. chology, 21(3): 281-304.
Eichengreen, Barry, Galina Hale, and Ashoka Mody. Frijters, Paul, Ingo Geishecker, John P. Haisken
2001. "Flight to Quality: Investor Risk Tolerance and DeNew, and Michael A. Shields. 2006. "Can the Large
the Spread of Emerging Market Crises." In Interna Swings in Russian Life Satisfaction Be Explained by
tional Financial Contagion, ed. S. Claessens and K. Ups and Downs in Real Incomes?" Scandinavian
J. Forbes. Boston; Dordrecht and London: Kluwer Journal of Economics, 108(3): 433-58.
Academic, 129-55. Frijters, Paul, John P. Haisken-DeNew, and Michael A.
Elster, Jon, and John E. Roemer, eds. 1991. Interper Shields. 2002. "Individual Rationality and Learning:
sonal Comparisons ofWell-Being. Studies in Ratio Welfare Expectations in East Germany Post-Reuni
nality and Social Change. Cambridge; New York and fication." IZA Discussion Paper, no. 498.
Melbourne: Cambridge University Press. Frijters, Paul, John P. Haisken-DeNew, and Michael
Falk, Armin, and Markus Knell. 2004. "Choosing the A. Shields. 2004a. "Investigating the Patterns and
Joneses: Endogenous Goals and Reference Stan Determinants of Life Satisfaction in Germany fol
dards." Scandinavian Journal of Economics, 106(3): lowing Reunification." Journal of Human Resources,
417-35. 39(3): 649-74.
Fern?ndez-Dols, Jos?-Miguel, and Maria-Angeles Frijters, Paul, John P. Haisken-DeNew, and Michael
Ruiz-Belda. 1995. "Expression of Emotion versus A. Shields. 2004b. "Money Does Matter! Evidence
Expressions of Emotions: Everyday Conceptions from Increasing Real Income and Life Satisfaction
about Spontaneous Facial Behavior." In Everyday in East Germany Following Reunification." Ameri
Conceptions of Emotion: An Introduction to the Psy can Economic Review, 94(3): 730-40.
chology, Anthropology and Linguistics of Emotion, Frijters, Paul, and Andrew Leigh. 2005. "Materialism
ed. J. Russell et al. Dordrecht: Kluwer, 505-22. on the March: From Conspicuous Leisure to Con
Ferrer-i-Carbonell, Ada. 2005. "Income and Well spicuous Consumption?" CEPR Discussion Paper,
Being: An Empirical Analysis of the Comparison no. 495.
Income Effect." Journal of Public Economics, 89(5 Gali, Jordi. 1994. "Keeping Up with the Joneses: Con
6): 997-1019. sumption Externalities, Portfolio Choice, and Asset
Ferrer-i-Carbonell, Ada, and Paul Frijters. 2004. "How Prices." Journal of Money, Credit, and Banking,
Important Is Methodology for the Estimates of the 26(1): 1-8.
Determinants of Happiness?" Economic Journal, Galizzi, Monica, and Kevin Lang. 1998. "Relative
114(497): 641-59. Wages, Wage Growth, and Quit Behavior." Journal
Fershtman, Chaim, Kevin M. Murphy, and Yoram of Labor Economics, 16(2): 367-91.
Weiss. 1996. "Social Status, Education, and Growth." Gardner, Jonathan, and Andrew J. Oswald. 2006. "Do
Journal of Political Economy, 104(1): 108-32. Divorcing Couples Become Happier by Breaking
Fliessbach, Klaus, Bernd Weber, Peter Trautner, Up?" Journal of the Royal Statistical Society: Series
Thomas Dohmen, Uwe Sunde, Christian Elger, A (Statistics in Society), 169(2): 319-36.
and Armin Falk. 2007. "Social Comparison Affects Gardner, Jonathan, and Andrew J. Oswald. 2007.
Reward-Related Brain Activity in the Human Ven "Money and Mental Wellbeing: A Longitudinal
tral Striatum." Unpublished. Study of Medium-Sized Lottery Wins." Journal of
Frank, Robert H. 1985. Choosing the Right Pond: Health Economics, 26(1): 49-60.
Human Behavior and the Quest for Status. Oxford Georgellis, Yannis, John Sessions, and Nikolaos Tsit
and New York: Oxford University Press. sianis. 2007. "Pecuniary and Non-pecuniary Aspects
Frank, Robert H. 1999. Luxury Fever: Why Money Fails of Self-Employment Survival." Quarterly Review of
to Satisfy in an Era of Excess. New York: Free Press. Economics and Finance, 47(1): 94-112.
Frank, Robert H., and Robert M. Hutchens. 1993. Glomm, Gerhard, and B. Ravikumar. 1994. "Growth
"Wages, Seniority, and the Demand for Rising Con Inequality Trade-Offs in a Model with Public Sec
sumption Profiles." Journal of Economic Behavior tor R&D." Canadian Journal of Economics, 27(2):
and Organization, 21(3): 251-76. 484-93.
Frederick, Shane, and George F. Loewenstein. 1999. Goedhart, Theo, Victor Halberstadt, Arie Kapteyn,
"Hedonic Adaptation." In Well-Being^. The Founda and Bernard M. S. Van Praag. 1977. "The Pov
tions of Hedonic Psychology, ed. D. Kahneman, E. erty Line: Concept and Measurement." Journal of
Diener, and N. Schwarz. New York: Russell Sage Human Resources, 12(4): 503-20.
Foundation, 302-29. Graham, Carol, Andrew Eggers, and Sandip Sukhtan
Freeman, Richard B. 1978. "Job Satisfaction as an kar. 2004. "Does Happiness Pay? An Exploration
Economic Variable." American Economic Review, Based on Panel Data from Russia." Journal of Eco
68(2): 135-41. nomic Behavior and Organization, 55(3): 319-42.
Frey, Bruno S., and Alois Stutzer. 2002a. Happiness Graham, Carol, and Andrew Felton. 2006. "Inequality
and Economics: How the Economy and Institutions and Happiness: Insights from Latin America." Jour
Affect Well-Being. Princeton and Oxford: Princeton nal of Economic Inequality, 4(1): 107-22.
University Press. Graham, Carol, and Stefano Pettinato. 2002. Happi
Frey, Bruno S., and Alois Stutzer. 2002b. "What Can ness and Hardship: Opportunity and Insecurity in
Economists Learn from Happiness Research?" Jour New Market Economies. Washington, D.C.: Brook
nal of Economic Literature, 40(2): 402-35. ings Institution Press.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Grund, Christian, and Dirk Sliwka. 2007. "Reference Kahneman, Daniel, Jack L. Knetsch, and Richard H.
Dependent Preferences and the Impact of Wage Thaler. 1986. "Fairness as a Constraint on Profit
Increases on Job Satisfaction: Theory and Evidence." Seeking: Entitlements in the Market." American
Journal of Institutional and Theoretical Economics, Economic Review, 76(4): 728-41.
163(2): 313-35. Kahneman, Daniel, Jack L. Knetsch, and Richard H.
Guth, Werner, Rolf Schmittberger, and Bernd Thaler. 1991. "The Endowment Effect, Loss Aver
Schwarze. 1982. "An Experimental Analysis of Ulti sion, and Status Quo Bias: Anomalies." Journal of
matum Bargaining." Journal of Economic Behavior Economic Perspectives, 5(1): 193-206.
and Organization, 3(4): 367-88. Kahneman, Daniel, and Alan B. Krueger. 2006.
Hagenaars, Aldi J. M. 1986. The Perception of Poverty. "Developments in the Measurement of Subjective
Amsterdam: Elsevier, North-Holland. Well-Being." Journal of Economic Perspectives,
Hamermesh, Daniel S. 1977. "Economic Aspects of 20(1): 3-24.
Job Satisfaction." In Essays in Labor Market Analy Kahneman, Daniel, Peter P. Wakker, and Rakesh
sis, ed. O. Ashenfelter and W. Oates. New York: Sarin. 1997. "Back to Bentham? Explorations of
Wiley, 53-72. Experienced Utility." Quarterly Journal of Econom
Headey, Bruce, and Mark Wooden. 2004. "The Effects ics, 112(2): 375-405.
of Wealth and Income on Subjective Well-Being and Kapteyn, Arie, Bernard M.S. Van Praag, and Floor G.
Ill-Being." Economic Record, 80: S24-33. Van Herwaarden. 1976. "Individual Welfare Func
Helliwell, John F. 2003. "How's Life? Combining tions and Social Reference Spaces." Leyden Univer
Individual and National Variables to Explain Sub sity, Economic Institute Report, no. 76.01.
jective Well-Being." Economic Modelling, 20(2): Kimball, Miles, and Robert Willis. 2006. "Utility and
331-60. Happiness." Unpublished.
Helliwell, John F., and Haifang Huang. 2005. "How's Knight, John, and Lina Song. 2006. "Subjective
the Job? Well-Being and Social Capital in the Work Well-Being and Its Determinants in Rural China."
place." NBER Working Paper, no. 11759. Unpublished.
Hirschman, Albert O., and Michael Rothschild. 1973. Knutson, Brian, Charles M. Adams, Grace W. Fong,
"The Changing Tolerance for Income Inequality in and Daniel Hommer. 2001. "Anticipation of Increas
the Course of Economic Development; with a Math ing Monetary Reward Selectively Recruits Nucleus
ematical Appendix." Quarterly Journal of Econom Accumbens." Journal of Neuroscience, 21(16):
ics, 87(4): 544-66. RC159.
Huppert, Felicia A. 2006. "Positive Emotions and Cog Koivumaa-Honkanen, Heli, Risto Honkanen, Heimo
nition: Developmental, Neuroscience and Health Viinam?ki, Kauko Heikkil?, Jaakko Kaprio, and
Perspectives." In Affect in Social Thinking and Markku Koskenvuo. 2001. "Life Satisfaction and
Behavior, ed. J. P. Forgas. New York: Psychology Suicide: A 20-Year Follow-Up Study." American
Press, 235-52. Journal of Psychiatry, 158(3): 433-39.
Inglehart, Ronald. 1990. Cultural Shift in Advanced Kristensen, Nicolai, and Niels Westerg?rd-Nielsen.
Industrial Society. Princeton: Princeton University 2006. "Job Satisfaction and Quits?Which Job Char
Press. acteristics Matters Most?" Danish Journal of Eco
Inglehart, Ronald, and Jacques-Ren? Rabier. 1986. nomics, 144(2): 230-48.
"Aspirations Adapt to Situations?But Why Are Lane, Robert E. 2000. The Loss of Happiness in Mar
the Belgians So Much Happier than the French? A ket Democracies. Yale ISPS Series. New Haven and
Cross-Cultural Analysis of the Subjective Quality London: Yale University Press.
of Life." In Research on the Quality of Life, ed. F. Layard, Richard. 1980. "Human Satisfactions and Pub
M. Andrews. Ann Arbor, Mich.: Institute for Social lic Policy." Economic Journal, 90(363): 737-50.
Research, University of Michigan, 1-49. Layard, Richard. 2005. Happiness: Lessons from a
Ireland, Norman J. 1994. "On Limiting the Market for New Science. London: Allen Lane.
Status Signals." Journal of Public Economics, 53(1): Leigh, Andrew, and Justin Wolfers. 2006. "Happiness
91-110. and the Human Development Index: Australia Is
Ito, Tiffany A., and John T. Cacioppo. 1999. "The Psy Not a Paradox." Australian Economic Review, 39(2):
chophysiology of Utility Appraisals." In Well-Being:
176-84.
The Foundations of Hedonic Psychology, ed. D. Lelkes, Orsolya. 2006. "Tasting Freedom: Happiness,
Kahneman, E. Diener, and N. Schwarz. New York: Religion and Economic Transition." Journalof Eco
Russell Sage Foundation, 470-88. nomic Behavior and Organization, 59(2): 173?94.
Johansson-Stenman, Olof, Fredrik Carlsson, and Dinky L?vy-Garboua, Louis, and Claude Montmarquette.
Daruvala. 2002. "Measuring Future Grandparents' 2004. "Reported Job Satisfaction: What Does It
Preferences for Equality and Relative Standing." Mean?" Journal ofSocio-Economics, 33(2): 135-51.
Economic Journal, 112(479): 362-83. Ljungqvist, Lars, and Harald Uhlig. 2000. "Tax Policy
Kahneman, Daniel, Ed Diener, and Norbert Schwarz, and Aggregate Demand Management under Catch
eds. 1999. Well-Being: The Foundations of Hedonic ing Up with the Joneses." American Economic
Psychology. New York: Russell Sage Foundation. Review, 90(3): 356-66.
Kahneman, Daniel, Barbara L. Fredrickson, Charles Loewenstein, George F., Ted O'Donoghue, and Mat
A. Schreiber, and Donald A. Redelmeier. 1993. thew Rabin. 2003. "Projection Bias in Predicting
"When More Pain Is Preferred to Less: Adding a Future Utility." Quarterly Journal of Economics,
Better End." Psychological Science, 4(6): 401-05. 118(4): 1209-48.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
142 Journal of Economic Literature, Vol. XLVI (March 2008)
Loewenstein, George F., and Nachum Sicherman. Women's Employment." Journal of Public Econom
1991. "Do Workers Prefer Increasing Wage Pro ics, 70(1): 157-83.
files?" Journal of Labor Economics, 9(1): 67-84. Ng, Yew-Kwang. 1987. "Relative-Income Effects and
Loewenstein, George F., and Peter A. Ubel. 2006. the Appropriate Level of Public Expenditure."
"Hedonic Adaptation and the Role of Decision and Oxford Economic Papers, 39(2): 293-300.
Experience Utility in Public Policy." Unpublished. Oswald, Andrew J. 1983. "Altruism, Jealousy and the
Lucas, Richard E. 2005. "Time Does Not Heal All Theory of Optimal Non-linear Taxation." Journal of
Wounds: A Longitudinal Study of Reaction and Public Economics, 20(1): 77-87.
Adaptation to Divorce." Psychological Science, Oswald, Andrew J. 1997. "Happiness and Economic Per
16(12): 945-50. formance." Economic Journal, 107(445): 1815-31.
Lucas, Richard E., Andrew E. Clark, Yannis Georgel Oswald, Andrew J., and Nattavudh Powdthavee. 2005.
lis, and Ed Diener. 2003. "Reexamining Adaptation "Does Happiness Adapt? A Longitudinal Study of
and the Set Point Model of Happiness: Reactions to Disability with Implications for Economists and
Changes in Marital Status." Journal of Personality Judges." Unpublished.
and Social Psychology, 84(3): 527-39. Oxoby, Robert J. 2004. "Cognitive Dissonance, Status
Lucas, Richard E., Andrew E. Clark, Yannis Georgel and Growth of the Underclass." Economic Journal,
lis, and Ed Diener. 2004. "Unemployment Alters the 114(498): 727-49.
Set Point for Life Satisfaction." Psychological Sci Palmore, Erdman 1969. "Predicting Longevity: A Follow
ence, 15(1): 8-13. Up Controlling for Age." Gerontologist, 9(4): 247-50.
Luttmer, Erzo F. P. 2005. "Neighbors as Negatives: Park, Yongjin. 2006. "The Second Paycheck to Keep
Relative Earnings and Well-Being." Quarterly Jour Up with the Joneses: Relative Income Concerns
nal of Economics, 120(3): 963-1002. and Labor Market Decisions of Married Women."
Lydon, Reamonn, and Arnaud Chevalier. 2001. "Esti Unpublished.
mates of the Effect of Wages on Job Satisfaction." Patacconi, Andrea, and Florian Ederer. 2005. "Inter
Unpublished. personal Comparison, Status and Ambition in
Lyubomirsky, Sonja, Laura King, and Ed Diener. Organisations." University of Oxford, Economics
2005. "The Benefits of Frequent Positive Affect: Series Working Paper, no. 222.
Does Happiness Lead to Success?" Psychological Pham, Thi Kim Cuong. 2005. "Economic Growth and
Bulletin, 131(6): 803-55. Status-Seeking through Personal Wealth." European
de Mandeville, Bernard. 1924. The Fable of the Bees; Journal of Political Economy, 21(2): 407?27.
or, Private Vices, Publick Benefits. Oxford: Claren Platt, Stephen, Rocco Micciolo, and Mich?le Tansella.
don Press.
1992. "Suicide and Unemployment in Italy: Descrip
Manski, Charles F. 1993. "Identification of Endogenous tion, Analysis and Interpretation of Recent Trends."
Social Effects: The Reflection Problem." Review of Social Science & Medicine, 34(11): 1191-1201.
Economic Studies, 60(3): 531-42. Plug, Erik J. S. 1997. "Leyden Welfare and Beyond."
Marmot, Michael. 2004. Status Syndrome: How Our Unpublished.
Position on the Social Gradient Affects Longevity Pollak, Robert A. 1976. "Interdependent Preferences."
and Health. London: Bloomsbury. American Economic Review, 66(3): 309-20.
Mas, Alexandre. 2006. "Pay, Reference Points, and Pressman, Sarah D., and Sheldon Cohen. 2005. "Does
Police Performance." Quarterly Journal of Econom Positive Affect Influence Health?" Psychological
ics, 121(3): 783-821. Bulletin, 131(6): 925-71.
McBride, Michael. 2001. "Relative-Income Effects on Rabin, Matthew. 1998. "Psychology and Economics."
Subjective Well-Being in the Cross-Section." Jour Journal of Economic Literature, 36(1): 11-46.
nal of Economic Behavior and Organization, 45(3): Rabien, Matthew. 2006. "Endogenous Reference
251-78. Level Formation as a Solution to a Problem of Self
McBride, Michael. 2006. "Money, Happiness, and Control." Unpublished.
Aspiration Formation: An Experimental Study." Ravallion, Martin, and Michael Lokshin. 2002. "Self
Unpublished. Rated Economic Welfare in Russia." European Eco
McDonald, Ian M. 2002. "How Social Preferences Can nomic Review, 46(8): 1453-73.
Improve Our Understanding of the Inflation-Unem Ravina, Enrichetta. 2005. "Keeping Up with the Jone
ployment Relation." Unpublished. ses: Evidence from Micro Data." Unpublished.
McMahon, Darrin M. 2006. Happiness: A History. Rayo, Luis, and Gary S. Becker. 2007. "Evolution
New York: Atlantic Monthly Press. ary Efficiency and Happiness." Journal of Political
Melenberg, Bertrand. 1992. "Micro-econometric Economy, 115(2): 302-37.
Models of Consumer Behaviour and Welfare." Riis, Jason, George F. Loewenstein, Jonathan Baron,
Unpublished. Christopher Jepson, Angela Fagerlin, and Peter A.
Mroczek, Daniel K., and Avron Spiro III. 2007. "Per Ubel. 2005. "Ignorance of Hedonic Adaptation to
sonality Change Influences Mortality in Older Men." Hemodialysis: A Study Using Ecological Momentary
Psychological Science, 18(5): 371-76. Assessment." Journal of Experimental Psychology:
Nettle, Daniel. 2005. Happiness: The Science behind General, 134(1): 3-9.
Your Smile." Oxford and New York: Oxford Univer Ryff, Carol D. 1989. "Happiness Is Everything, or Is
sity Press. It? Explorations on the Meaning of Psychological
Neumark, David, and Andrew Postlewaite. 1998. Well-Being." Journal of Personality and Social Psy
"Relative Income Concerns and the Rise in Married chology, 57(6): 1069-81.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Ryff, Carol D., and Burton H. Singer. 1998. "The Stark, Oded. 2006b. "Status Aspirations, Wealth
Contours of Positive Human Health." Psychological Inequality, and Economic Growth." Review of
Inquiry, 9(1): 1-28. Development Economics, 10(1): 171?76.
Sales, Stephen M., and James S. House. 1971. "Job Stark, Oded, and Lukasz Tanajewski. 2006. "Inducing
Dissatisfaction as a Possible Risk Factor in Coro Overtime Work: A Relative Deprivation Approach."
nary Heart Disease." Journal of Chronic Diseases, Unpublished.
23(12): 861-73. Stark, Oded, and J. Edward Taylor. 1991. "Migration
Sandvik, Ed, Ed Diener, and Larry Seidlitz. 1993. Incentives, Migration Types: The Role of Relative
"Subjective Well-Being: The Convergence and Sta Deprivation." Economic Journal, 101(408): 1163-78.
bility of Self-Report and Non-self-Report Measures." Stark, Oded, and You Qiang Wang. 2005. "Towards a
Journal of Personality, 61(3): 317-42. Theory of Self-Segregation as a Response to Rela
Schor, Juliet B. 1991. The Overworked American: The tive Deprivation: Steady-State Outcomes and Social
Unexpected Decline of Leisure. New York: Harper Welfare." In Economics and Happiness: Framing
Collins, Basic Books. the Analysis, ?d. L. Bruni and P. L. Porta. Oxford
Seabright, Paul. 2004. The Company of Strangers: A and New York: Oxford University Press, 223-42.
Natural History of Economic Life. Princeton and Steptoe, Andrew, Jane Wardle, and Michael Marmot.
Oxford: Princeton University Press. 2005. "Positive Affect and Health-Related Neuro
Seligman, Martin E. P. 2002. Authentic Happiness: endocrine, Cardiovascular, and Inflammatory Pro
Using the New Positive Psychology to Realize Your cesses." Proceedings of the National Academy of
Potential for Lasting Fulfillment. New York: Free Sciences, 102(18): 6508-12.
Press. Stutzer, Alois. 2004. "The Role of Income Aspira
Sen, Amartya. 1983. "Poor, Relatively Speaking." tions in Individual Happiness." Journal of Economic
Oxford Economic Papers, 35(2): 153-69. Behavior and Organization, 54(1): 89-109.
Senik, Claudia. 2004. "When Information Dominates Teulings, Coen, and Joop Hartog. 1998. Corporatism
Comparison: Learning from Russian Subjective or Competition? Labour Contracts, Institutions
Panel Data." Journal of Public Economics, 88(9-10): and Wage Structures in International Comparison.
2099-2123. Cambridge; New York and Melbourne: Cambridge
Senik, Claudia. 2005. "Ambition and Jealousy: Income University Press.
Interactions in the 'Old' Europe versus the 'New' Tremblay, L?on, and Wolfram Schultz. 1999. "Relative
Europe and the United States." PSE Discussion Reward Preference in Primate Orbitofrontal Cor
Paper, no. 2005-14. tex." Nature, 398(6729): 704-08.
Shields, Michael A., and Stephen Wheatley Price. Tversky, Amos, and Daniel Kahneman. 1991. "Loss
2005. "Exploring the Economic and Social Deter Aversion in Riskless Choice: A Reference-Depen
minants of Psychological Well-Being and Perceived dent Model." Quarterly Journal of Economics,
Social Support in England." Journal of the Royal 106(4): 1039-61.
Statistical Society: Series A (Statistics in Society), Urry, Heather L., Jack B. Nitschke, Isa Dolski, Daren
168(3): 513-37. C. Jackson, Kim M. Dalton, Corrina J. Mueller,
Shiv, Baba, and Joel Huber. 2000. "The Impact of Melissa A. Rosenkranz, Carol D. Ryff, Burton H.
Anticipating Satisfaction on Consumer Choice." Singer, and Richard J. Davidson. 2004. "Making
Journal of Consumer Research, 27(2): 202-16. a Life Worth Living: Neural Correlates of Well
Shizgal, Peter. 1999. "On the Neural Computation of Being." Psychological Science, 15(6): 367-72.
Utility: Implications from Studies of Brain Stimula Van de Stadt, Huib, Arie Kapteyn, and Sara Van de
tion Reward." In Well-Being: The Foundations of Geer. 1985. "The Relativity of Utility: Evidence
Hedonic Psychology, ed. D. Kahneman, E. Diener, from Panel Data." Review of Economics and Statis
and N. Schwarz. New York: Russell Sage Founda tics, 67(2): 179-87.
tion, 500-524. Van Praag, Bernard M. S. 1971. "The Welfare Func
Sloane, P. J., and H. Williams. 2000. "Job Satisfaction, tion oflncome in Belgium: An Empirical Investiga
Comparison Earnings, and Gender." Labour, 14(3): tion." European Economic Review, 2(3): 337-69.
473-501. Van Praag, Bernard M. S. 1991. "Ordinal and Cardi
Smith, Dylan M., Kenneth M. Langa, Mohammed U. nal Utility: An Integration of the Two Dimensions
Kabeto, and Peter A. Ubel. 2005. "Health, Wealth, of the Welfare Concept." Journal of Econometrics,
and Happiness: Financial Resources Buffer Subjec 50(1-2): 69-89.
tive Well-Being after the Onset of a Disability." Psy Van Praag, Bernard M. S., and Barbara E. Baarsma.
chological Science, 16(9): 663?66. 2005. "Using Happiness Surveys to Value Intangi
Smith, Vernon L. 1994. "Economics in the Labo bles: The Case of Airport Noise." Economic Journal,
ratory." Journal of Economic Perspectives, 8(1): 115(500): 224-46.
113-31. Van Praag, Bernard M. S., and Paul Frijters. 1999.
Solnick, Sara J., and David Hemenway. 1998. "Is More "The Measurement of Welfare and Well-Being: The
Always Better?: A Survey on Positional Concerns." Leyden Approach." In Well-Being: The Foundations
Journal of Economic Behavior and Organization, of Hedonic Psychology, ed. D. Kahneman, E. Die
37(3): 373-83. ner, and N. Schwarz. New York: Russell Sage Foun
Stark, Oded. 2006a. "Inequality and Migration: dation, 413-33.
A Behavioral Link." Economics Letters, 91(1): Veblen, Thorstein. 1899. The Theory of the Leisure
146-52. Class. London: Macmillan, George Allen and Unwin.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00
Veenhoven, Ruut. 1999. "Quality-of-Life in Individual Woittiez, Isolde, and Arie Kapteyn. 1998. "Social Inter
istic Society: A Comparison of 43 Nations in the Early actions and Habit Formation in a Model of Female
1990s." Social Indicators Research, 48(2): 157-86. Labour Supply." Journal of Public Economics, 70(2):
Vendrik, Maarten CM. 1998. "Unstable Bandwagon 185-205.
and Habit Effects on Labor Supply." Journal of Eco Wu, Stephen. 2001. "Adapting to Heart Conditions: A
nomic Behavior and Organization, 36(2): 235-55. Test of the Hedonic Treadmill." Journal of Health
Weinzierl, Matthew. 2005. "Estimating a Relative Util Economics, 20(4): 495-508.
ity Function." Unpublished. Zizzo, Daniel John, and Andrew J. Oswald. 2001. "Are
Winkelmann, Liliana, and Rainer Winkelmann. 1998. People Willing to Pay to Reduce Others' Incomes?"
"Why Are the Unemployed So Unhappy? Evidence Annales d'Economie et de Statistique, 63-64:
from Panel Data." Econ?mica, 65(257): 1-15. 39-65.

46.193.68.106 on Mon, 18 Dec 2023 22:22:47 +00:00

Session 4 l1

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Session 4 l1

Uploaded by

Copyright:

Available Formats

Relative Income, Happiness, and Utility: An Explanation for the Easterlin Paradox and

Stable URL: http://www.jstor.com/stable/27646948

This content downloaded from

Relative Income, Happiness, and Utility:

This content downloaded from

This content downloaded from

Figure 2. Life Satisfaction in Five European Countries, 1973-2004

This content downloaded from

Figure 3. Life Satisfaction and Income in East Germany, 1991-2002

This content downloaded from

This content downloaded from

This content downloaded from

Figure 4. The Relationship between Income and Happ

This content downloaded from

This content downloaded from

This content downloaded from

that earned over the past three years, we

This content downloaded from

Jump in income at the

Figure 6. Change in Happiness Following an Income Shock

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

4.4 Do the Empirical Correlates of Social comparisons can also be demon

This content downloaded from

27 Nicely summarized in Frank (1999), pp. 140-42.

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

30 A related issue is how hard individuals work once

This content downloaded from

y*t y? + i y*t +2 yi + 3 -^- Time

Incomes decreasing over time

Time -< y* + 3 y*t + 2 y? +1 ?/r

Figure 7. Importance of Kinks in the Utility Function

This content downloaded from

This content downloaded from

This content downloaded from

ginal utility of the consumption of the to over-consumption of the observable

This content downloaded from

Low reference consumption

Low reference consumption

Figure 8. Utility of Consumption with High/Low Reference Consumption, and

current and future status would seem toisbe

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

This content downloaded from

yt y? + i yt +2 yi + 3 -^- Time