What Can Economists Learn From Happiness Research

7/30/2019 What Can Economists Learn From Happiness Research

1/34

Journal of Economic LiteratureVol. XL (June 2002), pp. 402435

Frey and Stutzer: Happiness ResearchJournal of Economic Litera ture, Vol. XL (June 2002 )

What Can Economists Learnfrom Happiness Research?

BRUNO S. FREY and ALOIS STUTZER1

1. Why Study Happiness?

HAPPINESS IS generally considered

an ultimate goal of life; virtuallyeverybody wants to be happy. TheUnited States Declaration of Indepen-dence of 1776 takes it as a self-evidenttruth that the pursuit of happiness isan unalienable right, comparable tolife and liberty. It follows that economicsisor should beabout individual hap-piness; in particular, how do economicgrowth, unemployment and inflation, andinstitutional factors such as governanceaffect individual well-being?

In addition to this intrinsic interest,there are important reasons for econo-mists to consider happiness research. Thefirst is economic policy. At the micro-level, it is often impossible to make aPareto-improving proposal, because asocial action entails costs for some indi-

viduals. Hence an evaluation of the neteffects, in terms of individual utilities,

is needed. On an aggregate level, eco-nomic policy must deal with trade-offs,especially those between unemploy-ment and inflation. Data for twelveEuropean countries for the period197591 show that a 1-percentage-pointincrease in the unemployment rate is

marginally compensated for by a 1.7-percentage-point decrease in inflation(Rafael Di Tella, Robert MacCulloch,and Andrew Oswald 2001). This resultsignificantly deviates from the miseryindex that, for lack of information, hassimply been defined as the sum of thepercent unemployment rate and thepercent annual inflation rate. Anothertrade-off that can be calculated on thebasis of estimated happiness functionsis the compensating variation for beingunemployed rather than holding a job.

For the same European countries re-ferred to above, a move from the lowestincome quartile to the highest incomequartile would not be enough to offsetthe adverse effect of unemployment,suggesting that unemployed people suf-fer high non-pecuniary costs. Happinessresearch can thus usefully informeconomic policy decisions.

Another reason why happiness is ofrelevance to economists is the effect ofinstitutional conditions such as thequality of governance and the size ofsocial capital on individual well-being.Research for 49 countries in the 1980sand 1990s suggests that there are sub-stantial well-being benefits from factorssuch as improved accountability, effec-tiveness and stability of government,the rule of law, and control of corrup-tion. The data show that the effects

flowing directly from the quality of 402

1 University of Zurich, Institute for EmpiricalResearch in Economics. We wish to thank a largenumber of scholars and three referees for theirdetailed comments on this and previous versions

of the paper.


2/34

institutions are often much larger thanthose that flow through productivity andeconomic growth (John Helliwell 2001).

Happiness research can also help usunderstand the formation of subjective

well-being. This sheds new light onbasic concepts and assumptions of eco-nomic theory, such as whether peoplecan successfully predict their own fu-ture utilities (George Loewenstein, TedODonoghue, and Matthew Rabin 2000)or whether individual self-assessmentsof predicted, instant, and rememberedutility are consistent (Daniel Kahne-man, Peter Wakker, and Rakesh Sarin1997). It may also help to solve empiri-cal puzzles that conventional economictheories f ind difficult to explain. Aparadox needing explanation, for exam-ple, is that in several countries since

World War II real income has drasti-cally risen but self-reported subjective

well-being2 of the population has notincreased or has even fallen slightly. Inthe United States between 1946 and1991, per-capita real income rose by a

factor of 2.5 ( from approximately $11,000 to $27,000 in 1996 US$), butover the same period, happiness onaverage remained constant.3 At a givenpoint in time, higher income is posi-tively associated with peoples happiness,

yet over the life cycle, happiness staysmore or less unchanged. Another para-dox is that, since ancient times, workhas been considered a burden for indi-

viduals to bear, but empirical research onhappiness strongly suggests that being un-

employed, even when receiving the sameincome as when employed, depressespeoples well-being markedly.4

Many happiness research findingsadd new knowledge to what have be-

come standard views in economics,while other results challenge thoseviews. One finding is the consistentlylarge influence of nonfinancial variableson self-reported satisfaction. This doesnot mean that economic factors such asincome, employment, and price stabilityare unimportant, but does suggest thatthe recent interest in issues such as goodgovernance and social capital is well-founded. The findings also enrich ourknowledge of discrimination concerninggender, ethnicity and race, and age.

Section 2 discusses the relationshipbetween happiness and utility, and ar-gues that reported subjective well-being is a satisfactory empirical approxi-mation of individual utility. Sections 3to 5 report on how the economic vari-ables of income, unemployment, andinflation affect happiness. Section 6

shows that, in addition to current eco-nomic conditions, institutional factors,in particular the type of democracy andthe extent of government decentraliza-tion, systematically influence how satis-fied individuals are with their lives.Section 7 provides a summary and dis-cusses the implications for economicpolicy and theory.

2. Happiness and Utility2.1 Historical Sketch

For a long time, the study of happi-ness was the province of psychology.5

2 Subjective well-being is the scientific term inpsychology for an individuals evaluation of her ex-perienced positive and negative affect, happiness,or satisfaction with life. They are separable con-structs, and the precise terminology will be used

whenever empirical research is cited. Otherwise,the terms happiness, well-being, and life satis-faction are used interchangeably.

3 This is a well-established finding (RichardEasterlin 2001, p. 472, 1974, 1995; David Blanch-flower and Oswald 2000; Ed Diener and Shigehiro

Oishi 2000; and Charles Kenny 1999).

4 In addition to the literature already cited, seeBlanchflower (1996), Andrew Clark and Oswald(1994), Frey and Stutzer (1999), and Liliana

Winkelmann and Rainer Winkelmann (19 98).5 See, for example, Michael Argyle (1987), Die-

ner et al. (1999), Kahneman, Diener, and NorbertSchwarz (1999), Alex Michalos (1991), David My-

ers (1993), Richard Ryan and Edward Deci (2001),

Frey and Stutzer: Happiness Research 403


3/34

Only recently has this psychological re-search been linked to economics. Thepathbreaking contribution by Easterlin(1974) was noted by many economicsscholars, but at the time found few

followers. General interest in themeasurement and determinants of sub-

jective reported well-being was raisedby a symposium (Robert Frank 1997;Yew-Kwang Ng 1997; and Oswald1997). Since the late 1990s, economistshave started to contribute large-scaleempirical analyses of the determinantsof happiness in different countries andperiods.

This paper does not intend to providea general survey on happiness research(which has already been done byKahneman, Diener, and Schwarz 1999,and Frey and Stutzer 2002). Rather, we

wish to show which insights may be im-portant, if not necessary, for integratinginto economics.

2.2 Objective and Subjective Utility

Standard economic theory employs

an objectivist position based on ob-servable choices made by individuals.Individual utility only depends on tangi-ble goods and services and leisure. It isinferred from behavior (or revealedpreferences), and is in turn used to ex-plain the choices made. This modern

view of util ity has been influenced bythe positivistic movement in philoso-phy. Subjectivist experience (e.g., cap-

tured by surveys) is rejected as beingunscientific, because it is not objec-tively observable. Most importantly,cardinality of utility and interpersonalcomparability are not necessary forpositive demand theory, which, follow-ing Occams razor, constitutes a great

advantage (Lionel Robbins 1932; JohnHicks 1934; Roy Allen 1934). The axi-omatic revealed-preference approachholds that the choices made provide allthe information required to infer the

utility of outcomes. The axiomatic ap-proach is not only applied to deriveindividual utility, but also to measuresocial welfare. To do so, social welfarecomparison is based on the consump-tion behavior of households (DanielSlesnick 1998; for a critical analysis Ng1997, 2001).

The positivistic view still dominatesin economics. Amartya Sen (1986, p.18) observes that the popularity of this

view in economics may be due to a mix-ture of an obsessive concern with ob-servability and a peculiar belief thatchoice . . . is the only human aspectthat can be observed. Its dominance isreflected in the contents of micro-economics textbooks. Not all contempo-rary economists, however, subscribe tothis view. Numerous scholars have chal-lenged standard economic theory from

different angles. There are countlessexamples of nonobjectivist theoreticalanalyses in economics. They incorporateemotions (Jon Elster 1998) self-signaling(self-esteem), goal completion, masteryand meaning (Loewenstein 1999) orstatus (Frank 1985). In order to explainhuman behavior, interdependent utilityfunctions are considered, rather thaninterpersonally independent ones (Clarkand Oswald 1998). This challenges es-tablished welfare propositions (MichaelBoskin and Eytan Sheshinski 1978;Heinz Hollnder 2001; and RichardLayard 1980). The vast literature onanomalies in decision making (RichardThaler 1992) questions whether utilitycan generally be derived from observedchoices. Finally, consequentialism, of

which util itarianism is a special case, isnot the only aspect relevant for behav-

ior, for procedural utility should also be

and Fritz Strack, Argyle, and Schwarz (1991).There are also contributions by sociologists, inparticular Ruut Veenhoven (for example 1993),and political scientists (Ronald Inglehart 1990;

Robert Lane 2000).

404 Journal of Economic Literature, Vol. XL (June 2002)


4/34

considered (Sen 1995, 1997; Marc LeMenestrel 2001). The exclusive relianceon an objectivist approach by standardeconomic theory is thus open to doubt,both theoretically and empirically. In

any case, it restricts the possibility ofunderstanding and influencing human

well-being.The subjective approach to utility of-

fers a fruitful complementary path tostudy the world. Firstly, subjective

well-being is a much broader conceptthan decision utility; it includes experi-enced utility as well as procedural util-ity, and is for many people an ultimategoal. That is not the case for other things

we may want, such as job security, status,power, and especially money (income).

We do not want them for themselves,but rather to give us the possibility ofmaking ourselves happier. Secondly,the concept of subjective happiness al-lows us to capture human well-being di-rectly. This creates a basis for explicitlytesting fundamental assumptions andpropositions in economic theory.

2.3 Measuring Utility

A subjective view of utility recognizesthat everybody has their own ideasabout happiness and the good life andthat observed behavior is an incompleteindicator for individual well-being.Nevertheless, individuals happiness canbe captured and analyzed: people canbe asked how satisfied they are withtheir lives. It is a sensible tradition ineconomics to rely on the judgement ofthe persons directly involved. Peopleare reckoned to be the best judges ofthe overall quality of their lives, and itis a straightforward strategy to ask themabout their well-being. With the help ofone or more questions on global self-reports, it is possible to get indicationsof individuals evaluations of their lifesatisfaction or happiness. Behind the

score indicated by a person lies a cogni-

tive assessment of to what extent theiroverall quality of l ife is judged in afavorable way (Veenhoven 1993). Peo-ple evaluate their level of subjective well-being with regard to circumstances and

comparisons to other persons, past ex-perience, and expectations of the future.Measures of subjective well-being6 canthus serve as proxies for utility.

Individuals happiness and life satis-faction can be captured in large sur-

veys. A prominent example of a single-item question on a three-point scale isin the General Social Surveys (JamesDavis, Tom Smith, and Peter Marsden2001), which asks: Taken all together,how would you say things are thesedayswould you say that you are veryhappy, pretty happy, or not too happy?Life satisfaction is assessed on a scalefrom one (dissatisfied) to ten (satisfied)in the World Values Survey (Inglehartet al. 2000), which asks: All things con-sidered, how satisfied are you with

your li fe as a whole these days? TheEurobarometer Surveys, covering all

members of the European Union, asks asimilar question: On the whole, are

you very satisf ied, fairly satisfied, notvery satisfied, or not at all satisfied withthe life you lead? Among the multiple-item approaches, the most prominent isthe Satisfaction With Life Scale (Wil-liam Pavot and Diener 1993), composedof five questions, rated on a scale fromone to seven.7

6 Subjective well-being is an attitude consistingof the two basic aspects of cognition and affect.Affect is the label attached to moods and emo-tions. Affect reflects peoples instant evaluation ofthe events that occur in their lives. The cognitivecomponent refers to the rational or intellectualaspects of subjective well-being. It is usually as-sessed with measures of satisfaction. It has beenshown that pleasant affect, unpleasant affect, andlife satisfaction are separable constructs (Lucas,Diener, and Suh 1996).

7 For a survey on various measures of subjectivewel l-being, see Frank And rews and John Rob inson

(1991).



5/34

As subjective survey data are basedon individuals judgements, they areprone to a multitude of systematic andnonsystematic biases. It therefore needsto be checked whether people are in-

deed capable and willing to give mean-ingful answers to questions about their

well-being. Reported subjective well-being may depend on the order of ques-tions, the wording of questions, scalesapplied, actual mood, and the selectionof information processed. The rele-

vance of these errors, however, dependson the intended usage of the data.Often, the main use of happiness mea-sure is not to compare levels in an abso-lute sense but rather to seek to identifythe determinants of happiness. For thatpurpose, it is neither necessary to as-sume that reported subjective well-being is cardinally measurable nor thatit is interpersonally comparable. Thesubjective data can be treated ordinallyin econometric analyses so that higherreported subjective well-being reflectshigher well-being of an individual.

Whether happiness measures meet thiscondition has been widely assessed inpsychological evaluation studies.8 Dif-ferent measures of happiness correlate

well with one another (MeredithFordyce 1988). Factor analyses of self-and nonself-reports of well-being haverevealed a single unitary constructunderlying the measures suggestingtheir validity (Ed Sandvik, Diener, andLarry Seidlitz 1993). Reliability studiesindicate that reported subjective well-being is moderately stable and sensitiveto changing life circumstances (JoopEhrhardt, Willem Saris, and Veenhoven2000; and Bruce Headey and Alexander

Wearing 1991). Consistency tests reveal

that happy people smile more oftenduring social interactions (Jos-MiguelFernndez-Dols and Mara-Angeles Ruiz-Belda 1995), and are less likely to com-mit suicide9 (Honkanen Koivumaa et al.

2001) and that changes in brain electri-cal activity and heart rate account forsubstantial variance in reported nega-tive affect (Richard Davidson et al.2000). Thus, Diener (1984) in an earlysurvey concluded that the measuresseem to contain substantial amounts of

valid variance (p. 551).Provided that reported subjective

well-being is a valid and empiricallyadequate measure for human well-being, it can be modeled in a micro-econometric happiness function Wit=+Xit+it that is estimated by or-dered probit or logit. Thereby, true

well-being serves as the latent variable .X=x1, x2, ..., xn are known variableslike sociodemographic and socioeco-nomic characteristics, as well as institu-tional constraints on individual i at time

t. The model allows us to analyze each

factor that is correlated with reportedsubjective well-being separately. Thisapproach has been successfully appliedin numerous studies on the correlatesof happiness. Advanced methods havebeen used recently in order to addressnonrandom measurement errors.

Measurement errors, as well as unob-served characteristics, are captured inthe error term . They are the source ofpotential biases as discussed in the fol-lowing sections on unobserved person-ality traits and correlated measurementerrors. Many mistakes in peoples an-swers, however, are random and thus do

8 Comprehensive discussions of measurementproblems are provided in Andrews and Robinson(1991), Michalos (1991), Randy Larsen and Bar-bara Fredrickson (1999), Schwarz and Strack

(1999) and Veenhoven (1993).

9 Suicide is sometimes considered a more validmeasure of happiness because it refers to revealedbehavior. But suicide only captures the tail end ofthe distribution of mental well-being. While thisis less of a problem in studying the determinantsof low human well-being, it inhibits meaningfulstatements about average well-being and thus

welfare comparisons.



6/34

not bias the estimation results. This alsoholds true for the order of questions,the wording of questions, actual mood,and such.

Nonsampling errors are not always

uncorrelated with the variables of inter-est. A measurement error perspective(Marianne Bertrand and Sendhil Mul-lainathan 2001; Martin Ravallion andMichael Lokshin 2001) suggests thatthe inferences can be clouded by unob-served personality traits that influenceindividuals sociodemographic and socio-economic characteristics, as well as howthey respond to subjective well-beingquestions. For instance, people doing

voluntary work report higher life satis-faction (Argyle 1999). But volunteeringdoes not necessarily make people hap-pier. If extraverted people volunteer moreoften, and it is taken into considerationthat extraverts tend to report higher sat-isfaction scores (Kristina DeNeve andHarris Cooper 1998), then the observedcorrelation is biased. Idiosyncratic ef-fects that are time-invariant, however,

can be controlled for if the same indi-viduals are resurveyed over time. In alongitudinal or panel analysis, it ispossible to consider a specific baseline

well-being for each individual. The sta-tistical relationship between socioeco-nomic status and reported subjective

well-being is then identified by peo-ple who change their socioeconomicstatus.10

For some questions, a further reason

for biases in microeconometric happi-ness functions may be relevant: the cor-relation of measurement errors with in-dividual characteristics. For example,

young people often report lower li fe

satisfaction scores than old people. Onthe one hand, this could mean that

young people in fact experience lowerwell-being. On the other hand, it is pos-sible that age has an influence on howpeople react and respond to questionsabout their subjective well-being. Anobserved statistical relationship couldthen reflect only a spurious correlation.This kind of bias can hardly be overcomeby econometric techniques.11 It can bemitigated, however, by carefully devel-oped psychological tests and generationof the data.

In addition to the statistical precon-ditions to study the determinants ofhappiness discussed so far, further con-ditions have to be met if welfare com-parisons are undertaken on the basis ofreported subjective well-being. Theseconditions are cardinality and interper-

sonal comparability of the individualstatements of well-being. Economistsare likely to be skeptical about bothclaims.12 Both of them may be less of aproblem on a practical level than on atheoretical level (Kahneman 1999).13

Happy people, for example, are rated ashappy by friends and family members(Heidi Lepper 1998; and Sandvik, Die-ner, and Seidlitz 1993), as well as byspouses (Paul Costa and Robert

10 In addition to an unbiased estimation of par-tial correlations, the question of causality arises.In the example mentioned above, the conclusion isthat volunteering makes people happy. However,there is evidence that happier people are more

wil ling to contribute to other people s well-being(Myers 1993). Therefore, the observed partial cor-relation could also mean that happier people domore voluntary work. The direction of causalitycannot easily be identified even in a panel dataanalysis. Additional information from qualitativestudies, or in the form of instrumental variables, is

necessary.

11 Correlated measurement errors are usually noproblem for the inclusion of aggregate variables likeinflation or democratic rights in microeconomichappiness functions.

12 It should be noted that this skepticism coex-ists with well-established propositions in the lit-erature on income inequality and poverty, taxa-tion, and risk that implicitly accept cardinal utilitymeasurement and interpersonal comparability.

13 Ng (1996) develops a method that yields hap-piness measures that are comparable interperson-ally, intertemporally, and internationally based on

the concept of just perceivable increments.



7/34

McCrae 1988). Ordinal and cardinaltreatments of satisfaction scores gener-ate quantitatively very similar results inmicroeconometric happiness functions(Frey and Stutzer 2000). This is consis-tent with validation results of the in-come evaluation approach, which fo-cuses on the translation of verbalevaluations into numerical figures in acontext-free setting (Bernard van Praag1991). The meaning of a sequence of

verbal labels is about the same for allpeople in the sample, and the verbal scaleis efficiently used, as the underlying in-tervals are of about equal length. Theexisting research suggests that, for manypurposes, happiness or reported subjec-

tive well-being is a satisfactory empiri-

cal approximation to individual utility.It is thus possible and worthwhile tostudy economic effects on happiness.

3. Effects of Income on Happiness

In this section, three aspects of therelationship between income andhappiness are discussed:

Are persons with high income at agiven point in time happier than those

with low income (section 3.1)? Does an increase in income over time

raise happiness (section 3.2)? Are persons in rich countries happier

than those in poor countries (section

3.3)?

2.5

7000

Figure 1. Happiness and Equivalence Income in the United States

Equivalence income in 1996 US$

197274 199496 Trend 197274

Averagehappiness

Source: General Social Survey, National Opinion Research Center.

30000 40000 50000 6000020000100000

2.3

2.1

1.9

Trend 199496



8/34

3.1 Happiness and Differencein Income between Persons

Persons with higher income havemore opportunities to achieve whatthey desire: in particular, they can buymore material goods and services.Moreover, they have a higher status insociety. Higher income therefore yieldshigher utility, and conversely the poorare unhappy. This relationship betweenincome and happiness at a particularpoint in time and place (country) hasbeen the subject of a large empirical lit-erature. As a robust and general result,it has been found that richer people, on

average, report higher subjective well-being.14 The relationship between in-come and happiness, both in simple re-gressions and when a large number ofother factors are controlled for in multi-ple regressions, proves to be statistically(normally highly) significant. In thissense, income does buy happiness.

For the United States, figure 1 showsthe strong positive relationship between(equivalent) real income and happinessin 197274 and in 199496 (using theGeneral Social Survey data). Averagehappiness is calculated based on thescores ofnot too happy being equal to1, pretty happy equal to 2, and veryhappy equal to 3. Table 1 shows thecorresponding data in more detail. Inboth periods, the mean happiness rating(the higher it is, the happier peopleare) rises with income. In the lowest

decile of income, the mean happinessscores are 1.92 (for 197274) and 1.94(for 199496), for the fifth decile thescore is 2.19, and for the tenth andhighest decile it is 2.36. In the UnitedStates, people with higher income arehappier.

Data for Europe from the Euro-barometer Survey Series (197591) re-

veal a similar picture. For example, 88percent of persons in the upper quartileof the income bracket rate themselves

to be fairly satisfied or very satis-fied, while 66 percent of those in thelowest income quartile do likewise(see the data presented in Di Tella,MacCulloch, and Oswald 1999).

Additional income, however, does notraise happiness ad infinitum, and notfor certain. As may be seen in figure 1,the relationship between income andhappiness seems to be nonlinear; thereis diminishing marginal utility with ab-solute income. The data in table 1 alsoindicate that the same proportional in-crease in income yields a lower increasein happiness at higher income levels.

Within the bottom five deciles, dou-bling income increases reported happi-ness, on average, by 0.05 score points in199496; but only by 0.03 score pointsfor the top five deciles. Evidence fordiminishing marginal utility is also pro-

vided by three successive waves of theWorld Values Survey covering the years198082, 199091, and 199597, andincluding between eighteen and thirtycountries (a total of 87,806 observa-tions). It is estimated that for a personmoving from the fourth to the fifthdecile in the distribution of family in-come, subjective well-being rises by0.11 (on a ten-point scale with 1.0 indi-cating the lowest, and 10.0 the highestlevel of satisfaction). In contrast, mov-ing from the ninth to the tenth decileincreases subjective well-being by only0.02 (Helliwell 2001, p. 14).

Differences in income explain only alow proportion of the differences inhappiness among persons. In theUnited States, for example, the simplecorrelation is 0.20 (Easterlin 2001, p.468). Sometimes these findings are

misleadingly interpreted to mean that

14 Blanchflower and Oswald (2000) and Easter-lin (1995, 2001) for the United States; Di Tella,MacCulloch, and Oswald (2001) for the membercountries of the European Union; and Frey and

Stutzer (2000) for Switzerland.



9/34

income is not relevant for individualhappiness. The relevance of income,

however, is assessed with regard to thesize of the coefficient in a multivariateanalysis. A low correlation coefficientmight indicate that other factors arealso important in explaining why somepeople are happier than others.15 Inparticular, other economic (in particularunemployment) and noneconomic (inparticular health but also personality)factors exert strong influences beyondthe indirect consequences on income. Arelevant personality factor that mightintervene is that individuals who prizematerial goods more highly than other

values in life tend to be substantiallyless happy (Joseph Sirgy 1997). Simi-larly, people whose goals are intrinsic,i.e., those who define their values bythemselves, tend to be happier than

those with extrinsic goals, i.e., thoseoriented toward some external re-

ward, such as financial success or so-cial approval (Tim Kasser and Ryan2001).

Correlations do not establish causa-tion. It may well be that it is not neces-sarily higher income that makes peoplehappier, but rather that happier peopleearn higher income, e.g., because theylike to work harder and are more enter-prising. In order to test the direction ofcausation, the effect of windfalls deter-mining income were analyzed. Britishlottery winners and people receiving aninheritance reported higher mental

well-being in the following year. An un-expected transfer of 50,000 is esti-mated to raise subjective well-being bybetween 0.1 and 0.3 standard deviations(Jonathan Gardner and Oswald 2001;also Stephen Smith and Peter Razzell1975; and Philip Brickman, Dan Coates,

and Ronnie Janoff-Bulman 1978). This

TABLE 1HAPPINESS AND EQUIVALENCE INCOME IN THE U.S.A.

Equivalence Incomea

(1996 US$)

Mean HappinessRatingb

Mean EquivalenceIncome

Number ofObservations

197274 199496 197274 199496 197274 199496

Full sample 2.21 2.17 17434 20767 4214 5171

Decile1 1.92 1.94 2522 2586 421 4992 2.09 2.03 5777 5867 419 5283 2.17 2.07 8694 8634 417 4974 2.22 2.15 11114 11533 416 5425 2.19 2.19 13517 14763 391 5126 2.29 2.29 15979 17666 460 5007 2.24 2.20 18713 21128 393 5278 2.31 2.20 22343 25745 447 529

9 2.26 2.30 28473 34688 427 47210 2.36 2.36 46338 61836 423 565

Source: General Social Survey, National Opinion Research Center. Variables 34, 157, and 1028. Dont know andno answer responses are omitted.a Total household income divided by the square root of the total number of household members.b Based on score of not too happy = 1, pretty happy = 2, and very happy = 3.

15 The low correlation may, of course, simply

reflect substantial random disturbances.



10/34

suggests that causation indeed runsfrom income to happiness.

There may be many different reasonswhy higher income does not simplytranslate into higher happiness. With-

out doubt, one of the most importantones is that individuals compare them-selves to other individuals. It is not theabsolute level of income that mattersmost but rather ones position relativeto other individuals. This idea of rela-

tive income is part of the more generalaspiration level theory. Concepts of in-terdependent preferences due to com-parisons with relevant others (see GaryBecker 1974; Frank 1985; and RobertPollak 1976) supplement ideas focus-sing on preference changes due to com-parison with ones past consumptionlevel or expected future income.

In economics, Easterlin (1974, 1995,2001) uses the concept of aspirations asa frame of reference to explain happi-ness. He acknowledges that people withhigher income are, on average, happier,but raising everybodys income does

not increase everybodys happiness, be-cause in comparison to others incomehas not improved. This interpretation ofthe data is supported by laboratory find-ings showing the importance of relative

judgements for happiness (RichardSmith, Diener, and Douglas Wedell 1989,and Amos Tversky and Dale Griffin1991).

Many economists in the past havenoted that individuals compare them-selves to significant others with respectto income, consumption, status, or util-ity. Thorstein Veblen (1899) coined thenotion of conspicuous consumptionserving to impress other persons. Therelative income hypothesis was formu-lated and econometrically tested byJames Duesenberry (1949), who positedan asymmetric structure of externali-ties. People look upward when making

comparisons. Aspirations thus tend to

be above the level reached. Wealthierpeople impose a negative external ef-fect on poorer people, but not vice

versa. As a result, savings rates dependon the percentile position in the income

distribution, and not solely on the in-come level, as in a traditional savingsfunction.

A major line of research known asindividual welfare functions, or theLeyden approach, was opened by vanPraag and Arie Kapteyn (1973) and as-sociates (for a recent survey, see vanPraag and Paul Frijters 1999). A cardi-nal relationship between income and

welfare is established by asking indi-viduals to add income ranges to a num-ber of qualitatively characterized in-come levels.16 Answering this incomeevaluation question, they should takeinto account their own situation withrespect to family and job. Up to ninequalitative descriptions ranging fromexcellent to very bad are groupedalong an interval scale from 0 to 1. Thebounded scale reflects that the individ-

ual welfare function only measures rela-tive welfare as perceived by the individ-ual. Every individual evaluates theirincome by comparing it with the worstpossible situation and the best possiblesituation of complete satiation. Fromthis information, it is possible to esti-mate for each individual (i) the incomethat is required to reach a mean welfarelevel and (ii) the sensitivity of reportedeconomic welfare to ex ante incomechanges.

Individual welfare functions havebeen estimated for several countries

with good results, particularly for theNetherlands and Belgium (Floor van

16 For example, Please try to indicate what youconsider to be an appropriate amount for each ofthe following cases. Under my/our conditions, I

would call a net household income per [month] of:about . . . very bad; about . . . very good. Pleaseenter an answer on each line. . . . (van Praag

1993).



11/34

Herwaarden, Kapteyn, and van Praag1977). A particularly interesting aspectis the connection established betweenthe parameter of what people considersufficient income and their actual in-

come, which measures the preferencedrift due to a change in income. Apositive correlation suggests that the expost evaluation of a higher income issmaller than its ex ante evaluation. So

what rich people consider a sufficientincome, for example, is higher than

what poor people consider a suffi-cient income. It is found that thepreference shift destroys about 6080 percent of the welfare effect of anincrease in income, so that somewhatless than a third remains.17 Individualsanticipate substantial gains in terms ofsatisfaction from higher income, butin retrospect are often disappointedabout the small size of the effects ofthe gains.

Fred Hirsch (1976), in his book So-cial Limits to Growth, emphasizes therole of relative social status by calling

attention to positional goods that, bydefinition, cannot be augmented be-cause they rely solely on not beingavailable to others. This theme wastaken up by Frank (1985, 1999) who ar-gues that the production of positionalgoods in the form of luxuries, such asexceedingly expensive watches or

yachts, is a waste of productive re-sources, as overall happiness is therebydecreased rather than increased.

There is little doubt that people com-pare themselves to other people and donot use absolute judgements. But it iscrucial to know with what other peoplesuch a comparison is being made. In astudy of five thousand British workers,

Clark and Oswald (1996) formed thereference group comprising persons

with the same labor-market charac-teristics. They conclude that the higherthe income of the reference group, the

less satisfied people are with their jobs.Social comparisons within the familyare studied by David Neumark and An-drew Postlewaite (1998) in order to testthe role of relative income for utility.They find that the decision of a womanto go for paid work depends on whetherher sisters and sisters-in-law are em-ployed and how much they earn at their

jobs.The effect of the distribution of in-

come on happiness has so far beenrarely addressed, mainly due to the lackof suitable data. A fascinating findingby Alberto Alesina, Di Tella, and Mac-Culloch (2001) suggests that there is alarge negative and statistically signifi-cant effect from inequality on happinessin Europe, but not in the United States.This may mean that Europeans have astronger aversion to inequality than

Americans have. Alternatively, upwardsocial mobility is perceived to begreater in the United States than inEurope, and therefore being low in theincome distribution in the UnitedStates is not seen as limiting futureincome as much.

Most of the research on the relation-ship between individual income andhappiness has been undertaken for ad-

vanced industrial countries. Yet CarolGraham and Stefano Pettinato (2001a,b)show that the results essentially carryover to both developing countries andcountries in transition. The evidence isconsistent with a positive relationshipbetween individual income and happi-ness within a society at a given point intime. It also emphasizes the relevanceof the relative position in the incomedistribution rather than the absolute

level of income.

17 If interdependent preferences are taken intoconsideration in addition to habit formation, thepreference drift seems to outbalance 100 percentof the welfare effect of income gains (Huib van de

Stadt, Kapteyn, and Sara van de Geer 1985).



12/34

3.2 Income and Happiness over Time

Several scholars (Blanchflower andOswald 2000; Diener and Oishi 2000;Myers 2000; Kenny 1999; Lane 1998;and Easterlin 1974, 1995) have identi-fied a striking and curious relationship:per-capita income in western countrieslike the United States, the United King-dom, and Belgium, as well as Japan,has risen sharply in recent decades,

whereas average happiness has stayedvirtually constant or has even de-clined over the same period. Graphi-cally, the development of income andhappiness diverges like open scissors.Consider figure 2 for Japan. Between1958 and 1991, income per capita inJapan rose by a factor of six. This is

probably the most spectacular growth in

income since World War II. The risewas reflected in almost all householdshaving an indoor toilet, a washingmachine, a telephone, a color televi-sion, and a car (Easterlin 2000). Theopen-scissors figure also shows that thistremendous rise in material well-being

was not accompanied by an increase inaverage satisfaction with life. In 1958,average life satisfaction rated on a four-point scale was 2.7. In 1991, after morethan thirty years of increasing afflu-ence, average life satisfaction stillscores 2.7 points.

The same relationship is revealed forthe United States in table 1 above.Between 197274 and 199496, overallmean equivalent real income in thesample increased from $17,434 to

$20,767 (19 percent). But the overall

15000

Figure 2. Satisfaction with Life and Income Per Capita in Japan between 1958 and 1991

Year

RealGDPpercapitainconstant$

Averaelifesatisfaction

Sources: Penn World Tables and World Database of Happiness.

1958

6000

9000

12000

3000

0 1

1.5

2

2.5

3

3.5

4

1962 1966 1970 1974 1978 1982 1986

Life satisfaction

Real GDPper capita

1990



13/34

mean happiness rating decreasedslightly, from 2.21 to 2.17. Income inall deciles (except the third) increased,

yet mean happiness ratings fell orstayed constant in eight of the ten

deciles.What can be inferred from these two

cases? One position that can be taken isto disregard the descriptive evidence (i)because other western countries likeDenmark, Germany, and Italy experi-enced substantial real per-capita in-come growth as well as a (small) in-crease in reported satisfaction with lifein the 1970s and 1980s (Diener andOishi 2000) and (ii) because whetherthere is a small increase or decrease inreported subjective well-being dependson the observation period. Moreover,the relationships presented between in-come and happiness over time are notanalyzed ceteris paribus. For theUnited States, however, a negative timetrend occurs when individual charac-teristics are controlled for (Blanch-flower and Oswald 2000). For twelve

European countries between 1975 and1991 there is no correlation betweenreal GDP per capita and life satisfac-tion, provided individual characteristicsas well as the unemployment rate, infla-tion rate, and income distribution arecontrolled for (Alesina, Di Tella, andMacCulloch (2001).18

Another position that can be taken isto accept that there is no clear-cuttrend, positive or negative, in self-reported subjective well-being overperiods of twenty to thirty years in richcountries. The missing correlation isnot due to a changing population. For

the United States, average happiness ofa cohort also remains constant over thelife cycle, despite considerable growthin income (Easterlin 2001).

The results can be taken as an indica-

tion that there is more to subjectivewell-being than just income level. Oneof the most important processes peoplego through is that of adjusting to pastexperiences. Human beings are unableand unwilling to make absolute judge-ments. Rather, they are constantlydrawing comparisons from the past orfrom their expectations of the future.Thus, we notice and react to deviationsfrom aspiration levels.

Additional material goods and ser-vices init ially provide extra pleasure,but it is usually only transitory. Higherhappiness with material things wearsoff. Satisfaction depends on change anddisappears with continued consump-tion. This process, or mechanism, thatreduces the hedonic effects of a con-stant or repeated stimulus, is calledadaptation. And it is this process of

hedonic adaptation that makes peoplestrive for ever-higher aspirations.

Adaptation level theory is well-grounded in psychology (in particularHarry Helson 1964; Brickman and Don-ald Campbell 1971; Allen Parducci 1995;and, for a modern discussion, ShaneFrederick and Loewenstein 1999), as isthe concept of aspiration levels (FrancisIrwin 1944). According to aspirationlevel theory, happiness is determinedby the gap between aspiration andachievement (Michalos 1991 and Ingle-hart 1990, ch. 7). In economics, thetheories of preference change have con-centrated on habit formation (AlfredMarshall 1890; Duesenberry 1949;Franco Modigliani 1949; Robert Pol-lack 1970; and more recently Chris-topher Carroll and David Weil1994). There are three important

consequences:

18 A more fundamental objection could questionwhether it is in principle possible to capturetrends on a closed scale. Valuable complementaryevidence could be provided by measures of mental

wel l-being like the General Health Questionnaire(David Goldberg 1972), where much less framing

in terms of categories is to be expected.



14/34

(1) The upward adjustment of aspira-tions induces human beings toaccomplish more and more. Theyare never satisfied. Once they have

achieved something, they want toachieve even more. The theory ofrising aspirations does not holdonly for material goods and servicesbut also for many immaterialachievements. A promotion, for ex-ample, makes for temporary happi-ness, but at the same time raisesthe expectation and aspiration forfurther promotions.

(2) Wants are insatiable. The more onegets, the more one wants. The mar-ginal utility of income is thus not de-fined anymore in this framework, asthe utility function changes with theincome level.

(3) Most people think that they felt lesshappy in the past, but expect to bemore happy in the future (Easterlin2001). This asymmetry can be ex-plained by changing aspirations, as

will be illustrated below.

The effects of changes in income affect-ing aspiration levels are illustrated infigure 3 (see Easterlin 2001).

Initially, people have a certain as-

piration level Al so that income Y1produces happiness H1. Raising income,say from Y1 to Y2, raises happinessfrom H1 to H2. If it rises further, sayto Y3, happiness is further increased toH3. The points a, b, and c trace a curve

with decreasing marginal util ity of in-come, as normally assumed in eco-nomic theory. This curve holds for aparticular point in time. It suggests thathigher income indeed makes peoplehappier.

But, over time, aspiration adjusts tothe higher income level. The aspirationlevel curve Al shifts downward to Am.Ex post, the rise in income from Y1 toY2 does not produce any increase inhappiness if the aspiration curve indeedshifts as much downward as assumed inthe graph. If the increase in income

jacks up aspirations even higher, say to

the aspiration curve Ah, income Y2

Happi-ness

Figure 3. Happiness, Income, and the Role of the Aspiration Level

H3

H2H4

H1

H0

Y1 Y2 Y3 Y4 Income Y

a

b e

f

h

g

d

c

A1

Am

Ah

aspirationlevels(A1


15/34

produces even less happiness than thelower income Y1.

Aspiration level theory suggests thatincreases in income and aspiration lev-els are closely connected. The increase

in happiness expected on the basis of agiven aspiration curvefor examplealong the points a, b, and c on aspira-tion curve Aldoes not materialize.Rather, an increase in income is accom-panied by a downward shift in the aspi-ration curve. In equilibrium, one mayobserve, for example, that the sequenceof pointsa, e, and f materializes. Asthe curves are drawn, higher incomematches higher happiness, but an in-crease in income produces a muchsmaller increase in happiness than withgiven aspiration levels.

As indicated above, the figure helpsto explain the asymmetry in evaluationsof happiness referring to the past and tothe future. A person with income Y3

judges her past happiness on the basisof the current aspiration level Am. As in-come has risen, say from Y2 to Y3, the

previous income is evaluated to haveproduced happiness H1 at point d,

which is lower than todays happinesslevel H4, as given at point e. Currenthappiness is thus taken to be higherthan in the past. In actual fact, whenthe individuals actually received in-come Y2, they had a lower aspirationlevel and therefore that income actuallyproduced happiness H2 in the past,

which in our figure is even higher thantodays happiness H4.

Future income is also evaluated onthe basis of the current aspiration level.Lets take the case of a person situatedat point e with income Y3 and happinessH4. The person anticipates that an in-crease in income from Y3 to Y4 pro-duces a well-being along curve Am, sothat happiness H3 at point f is to be ex-pected. But the person does not take

into account that the aspiration level

also rises and that the aspiration curvewill therefore shift downward, say tocurve Ah. In actual fact, therefore,

when the higher income Y4 is indeedreached, the level of happiness is only

H4 at point h, and not H3 as would havebeen the case if the aspiration level hadstayed constant at point f. The actualhappiness of the increase in income isthus systematically lower than expectedbeforehand (in our figure 3, happinesseven stays constant).

3.3 Income and Happiness betweenDifferent Countries

Various studies provide evidencethat, on average, persons living in richcountries are happier than those livingin poor countries (for example Diener,Marissa Diener, and Carol Diener1995; Inglehart 1990). The differencesin income between the countries aremeasured by using exchange rates andpurchasing power parities to control forthe international differences in the costof living. Data on happiness are usually

from the World Values Survey, thebest source available today for interna-tional comparisons of life satisfaction(Inglehart et al. 2000).

Figure 4 illustrates the relationshipbetween income per capita and averagelife satisfaction in 51 countries for datafrom the two waves of the World ValuesSurvey in the 1990s. The figure showsthat reported subjective well-beingseems to rise with income. Some of theauthors identify a concave relationship:income provides happiness at low levelsof development but once a threshold(around $10,000) is reached, the aver-age income level in a country has littleeffect on average subjective well-being.

A visual inspection of the relationshipbetween income and happiness acrosscountries, however, is of limited value.The positive correlation may be pro-

duced by factors other than income



16/34

alone. In particular, countries withhigher per-capita incomes tend to havemore stable democracies than poorcountries. So it may well be that theseemingly observed positive associationbetween income and happiness is in re-ality due to more developed democraticconditions (see section 6).

In addition to democracy, there maybe other conditions correlated with in-come that produce the observed posi-tive correlation between income andhappiness. To mention two more: thehigher the income is, the better theaverage health and the more secure

the basic human rights. Thus, both

health and basic human rights maymake happiness rise with income.

Moreover, the positive correlationmay come as a surprise in light of evi-dence presented for happiness overtime, where no robust relationship be-tween income per capita and happinessis found. There are two possible strate-gies to further address the role of abso-lute income in happiness in and be-tween countries. First, cross-sectionaldata for several periods can be com-bined in order to allow a control oftime-invariant country-specific charac-teristics. These characteristics could

comprise stable cultural differences,

9

30000

Figure 4. Life Satisfaction and Income Levels Across the World in the 1990s

GNP per capita in PPP 1995 US$

Satisfactionwithlife

Sources:World Values Survey 199093/199597 (ICPSR 2790) and World Development Report 1997.

15000 20000 250001000050000

8

7

6

5

4

3



17/34

systematic distortions due to languagedifferences, and so on. Such anapproach is followed in the studymentioned earlier combining data for49 countries from the first three waves

of the World Values Survey. Insteadof country specific effects, separatebase levels for six groups of countriesare taken into consideration in theestimation equation: industrial coun-tries, Scandinavia, countries of theformer Soviet Union, other countriesof Eastern Europe, Latin Americancountries, and other developing coun-tries.19 The study found that nationalincome per capita (measured in per-centage of the value for 1997 in theUnited States) has a very small effect onreported subjective well-being. A 10-percent increase in per-capita incomesin a country with half the level ofthe United States (and unchanged in-come distribution) raises average satis-faction with life by only 0.0003 scorepoints on a scale from one to ten, andthe gain disappears even before the

U.S. 1997 level of real per-capita incomeis achieved (Helliwell 2001, p. 15).

This evidence is more in line with thefindings for income and happiness overtime than with previous results fromcross-section studies that neglectedcountry- or region-specific determi-nants of reported subjective well-being.It may be argued that poor countriesare not adequately represented in thedata pool. As the relationship betweenincome and happiness seems relevant,especially from a development perspec-tive, a second strategy could directly ad-dress subjective well-being in develop-ing countries. But extended time seriesabout reported subjective well-being indeveloping countries with strong eco-nomic growth are lacking so far. Prom-

ising projects are a socioeconomic panelin Russia (Ravallion and Lokshin 2001)and repeated surveys in seventeen LatinAmerican countries (Graham and Petti-nato 2001b). First evidence for Peru

and Russia indicates that economic de-velopment is accompanied by extensivesocial mobility and, for some people

with rapidly increasing aspirationlevels, that may depress overall well-being gains from increased economic

wealth. While perceived past mobilityand prospects of upward mobility havea positive effect on reported subjective

well-being, there is a fraction of frus-trated achievers that, in spite of objec-tive mobility, reports negative per-ceived mobility and low satisfaction

with li fe (Graham and Pettinato 2001a).Another aspect to consider is

whether, when income and happinessbetween countries are compared, cau-sality runs from income to well-being,as implicitly assumed so far. The re-

verse causality can well be imagined(see Kenny 1999). It could be argued

that the more satisfied the popula-tion is with its life, the more it is in-clined to work hard, and therefore thehigher is its per-capita income. In other

words, happy people may be more crea-tive and enterprising, leading again tohigher income. This line of argumenthas not yet been well studied butshould be seriously considered in thefuture.

The available evidence suggests thatacross nations income and happinessare correlated but that the effects aresmall and diminishing. This indicatesthat, on the one hand, other factors maybe more important in explaining differ-ences in reported subjective well-beingbetween countries; on the other hand,the notion that people in poor countriesare happier because they live under morenatural and less stressful conditions is

a myth.

19 The strategy of constructing groups of coun-tries is chosen in order to allow more degrees of

freedom in the statistical analysis.



18/34

4. Effects of Unemploymenton Happiness

Two questions and their ramificationswill be discussed in this section:

What is the level of happiness of anunemployed person (section 4.1)?

How does general unemployment inan economy affect happiness (section4.2)?

4.1 Personal Unemployment

Reports on subjective well-being helpto identify the level of utility of unem-ployed people. How particular people

are affected when they become unem-ployed has been studied with individualdata for twelve European countries overthe period 197591, employing Euro-barometer data on satisfaction with lifeon a four-point scale (Di Tella, MacCul-loch, and Oswald 2001). The analysis,

which controls for a large number ofother determinants of happiness, suchas income and education, finds that the

self-proclaimed happiness of unem-ployed persons is much lower than em-ployed persons with otherwise similarcharacteristics. The loss of subjective

well-being experienced through un-employment amounts to 0.33 units inthe satisfaction scale, which rangesfrom 1 (not at all satisfied) to 4 (verysatisfied).

Many other studies for differentcountries and time periods have simi-

larly found that personally experiencingunemployment makes people very un-happy.20 In their s tudy for Britain,Clark and Oswald (1994, p. 655) sum-marize their result as joblessness de-pressed well-being more than any other

single characteristic, including impor-tant negative ones such as divorce andseparation. Some analyses offer addi-tional results for particular groups ofunemployed people. While the picture

is not totally consistent, many studiesfind that unemployment on average

weighs more heavily on men than onwomen. Younger and older employeessuffer less when hit by unemploymentthan employees in the middle of their

working lives. For Germany, unemploy-ment does not reduce satisfaction withlife of women over fifty (Knut Gerlachand Gesine Stephan 1996). People withhigh education experience a larger de-crease in their subjective well-beingdue to unemployment than employees

with low education (Clark and Oswald1994).

All these results refer to the pureeffect of being unemployed. The in-come loss, as well as other indirecteffects which may, but need not, go

with personally being unemployed, arecontrolled for.

These findings could be interpretedquite differently. While the negativecorrelation between unemployment andhappiness is well established, it may

well be that the causality runs in theopposite direction: unhappy people donot perform well and therefore get laidoff. Happy persons are fitter for work-ing life, which makes it less likely thatthey will lose their jobs. The questionof reverse causality due to selectionbias has been analyzed in many studies

with longitudinal data, before and afterparticular workers lose their jobs (forexample due to plant closure). There isevidence that unhappy people do notperform well on the labor market, butthe main causality seems clearly to runfrom unemployment to unhappiness(see Winkelmann and Winkelmann1998 for German panel data; Gary

Marks and Nicole Fleming 1999 for

20 Anders Bjrklund and Tor Eriksson (1998) forScandinavian countries; Blanchflower and Oswald(2000) for the United Kingdom and the UnitedStates; Thomas Korpi (1997) for Sweden; Raval-lion and Lokshin (2001) for Russia; and William

Darity and Arthur Goldsmith (1996) for a survey.



19/34

Australian panel data, with detailedconsideration of various effects onmental health).21

As the lower subjective well-being ofunemployed people can be explained

neither by the lower income level northe self selection of intrinsically less-happy people, unemployment has to berelated to non-pecuniary costs. Thedrop in happiness may be attributed toa large extent to psychological and so-cial factors (see the survey by NormanFeather 1990):

Psychic Cost. Unemployment producesdepression and anxiety, and results in

a loss of self-esteem and personal con-trol. Especially for persons very in-

volved in their work, not having a jobis a heavy blow. Numerous studieshave established (see Goldsmith,Jonathan Veum, and Darity 1996) thatthe unemployed are in worse mental(and physical) health than workingpeople. As a result, they are subject toa higher death rate, more often com-mit suicide22 and are more prone toconsuming large quantities of alcohol.Their personal relationships are alsomore strained.

The psychic cost is considerably higherfor those made redundant for the firsttime. In contrast, persons who havebeen unemployed more often in thepast suffer less, that is, to some extentthey get used to being unemployed.

This finding may to some extent ex-plain persistent unemployment (Clark,Yannis Georgellis, and Peter Sanfey2001).

Social Cost. Being unemployed has a

stigma attached to it, particularly in aworld in which ones work essentiallydefines ones position in life. This as-pect will be further discussed in thefollowing section.

4.2 General Unemployment

People may be unhappy about unem-ployment even if they are not them-selves out of work. They may feel badabout the unfortunate fate of those un-employed and they may worry about thepossibility of becoming unemployedthemselves in the future. They may alsofeel repercussions on the economy andsociety as a whole. They may dislike theincrease in unemployment contributionsand taxes likely to occur in the future;they may fear that crime and social ten-sion will increase; and they may evensee the threat of violent protests and

uprisings.The study of twelve European coun-

tries over the period 197591 men-tioned above (Di Tella, MacCulloch,and Oswald 2001) finds thatkeepingall other influences constanta 1-percentage-point increase in the gen-eral rate of unemployment from 9percent (the European mean) to 10 per-cent reduces stated life satisfaction by0.028 units on the four-point scale ap-plied. This effect is considerable. Thissmall rise in unemployment is equiva-lent to shifting more than 2 percent ofthe population downward from one life-satisfaction category to another, for ex-ample from not very satisfied to notat all satisfied.

The overall effect of unemploymenton social well-being can be calculatedby adding the loss experienced by the

unemployed persons to the overall

21 Studies in social psychology also identify ef-fects of unemployment and re-employment onmental well-being. Studies that explicitly controlfor individual heterogeneity with a longitudinaldesign are Mary Dew, Evelyn Bromet, and LiliPenkower (1992), and Brian Graetz (1993). For arecent survey, see Gregory Murphy and JamesAthanasou (1999).

22 A recent study using longitudinal data for theUnited States for 197291 found that state unem-ployment rates are significantly positively relatedto the number of suicides. A 1-percentage-pointincrease in state unemployment rates predicts anincrease of suicides by 1.3 percent (Christopher

Ruhm 2000).



20/34

effect of unemployment. Consider againa 1-percentage-point increase in unem-ployment. In the previous section wediscussed that the unemployed experi-ence a decrease of 0.33 in their happiness

scale. This figure must be multiplied bythe 1 percent of the population unfortu-nate enough to actually become unem-ployed: 0.33 0.01 = 0.0033. Added tothe general effect of a 1-percentage-point unemployment increase of 0.028,it leads to a total decrease of 0.0313 (DiTella, MacCulloch, and Oswald 2001).

This calculation must be taken with agrain of salt. It is at best able to ap-proximately gauge the effects of unem-ployment on happiness. One reason forcaution is that there may be various in-teractions between personal and moregeneral unemployment that may in turnaffect the evaluation of happiness.

An important interaction refers toreference groups. As is the case for in-come, individuals tend to evaluate theirown situation relative to other persons.For most persons, unemployment low-

ers their happiness less if they are notalone in their fate. When unemploy-ment is seen to hit many persons oneknows or hears of, both the psychic andthe social effects are mitigated. Self-esteem is better preserved because itseems that being out of a job is lessones own fault and more due to generaldevelopments in the economy. Stigmaand social disapproval are less prevalentif unemployment hits many others atthe same time.

In order to empirically test the effectof reference groups on reported well-being, happiness scores have been re-gressed on three types of explanatory

variables: personal unemployment; unem-ployment among a reference group; and aninteraction variable combining personaland reference group unemployment.

Using as a reference group the em-

ployment state of ones partner or the

region an individual lives in, Clark(2000) has estimated such a happinessfunction for British data over the period199196, again keeping all other influ-ences constant. As in virtually all pre-

vious studies, the unemployed are muchmore dissatisfied than people with jobs,and the general level of unemploymentlowers happiness. In contrast, the un-employed indeed suffer less when thepartner and/or a larger proportion ofother people in their region are also outof work. The same result holds whengeneral unemployment in the economyis taken as the point of reference (PeterKelvin and Joanna Jarrett 1985).

Unemployed peoples well-being,moreover, depends on the strength ofthe social norm to work. Social interac-tion of unemployed people with othercommunity members, the referencegroup forced upon themselves, has theeffect of showing them how they are ex-pected to behave, and norm-conformingbehavior is enforced through socialsanctions. An estimation across Swiss

communities shows that the strongerthe social norm to live off one s ownincome, the lower the unemployedpeoples reported satisfaction with life(Stutzer and Rafael Lalive 2001).

Reference groups are of major impor-tance for showing the extent to whichpeople are distressed by their own un-employment. The group one refers to isnot given, but can to some extent bechosen (Armin Falk and Markus Knell2000). People out of work tend to asso-ciate with other people out of work,partly because they have time to do so,and partly because they retreat fromnormal community life. It is also clearthat marriages and partnerships have ahigh risk of breaking down when onepartner is unemployed. In all thesecases, the definition of the referencegroup adjusts to ones labor market

status. Causation then does not run



21/34

unambiguously from the reference groupto the evaluation of unemployment interms of happiness.

5. Effects of Inflation on Happiness

In combined time-series and cross-section studies, the development ofinflation in several countries over thecourse of time can be analyzed. Of mostinterest is the study of twelve Europeancountries over the period 197591 (DiTella, MacCulloch, and Oswald 2001).The mean rate of inflation was 7.5 per-cent per year. Based on an econometricestimate that keeps individual socio-economic characteristics and the unem-ployment rate constant, increasing theinflation rate by one percentage pointsay from the mean rate of 8 to 9 percentper yearappears to reduce averagehappiness by 0.01 units of satisfaction,i.e., from an average level in the sampleof 3.02 to 3.01. (Average satisfaction iscalculated from a cardinal interpreta-tion of the 4-point scale that attributes

to not at all satisfied a value of 1, notvery satisfied a value of 2, etc.). Corre-spondingly, increasing the inflation rateby 5 percentage points (which histori-cally is quite a likely event) reducessubjective well-being by 0.05 units. Thisis a substantial effect. It means that 5percent of the population is shifteddownward from one life-satisfactioncategory to the next lower one, e.g.from being very satisf ied to fairlysatisfied.

In order to study the trade-off be-tween inflation and unemployment, theresults reported on the effect of unem-ployment on happiness and the effectsof inflation can be combined (Di Tella,MacCulloch, and Oswald 2001). Thequestion is by how much, on average,must a country reduce its inflation inorder to tolerate a rise of 1 percentage

point in unemployment? Over the rele-

vant range, happiness is assumed todepend linearly on these two eco-nomic factors, and the estimate controlsfor country fixed-effects, year effects,and country-specific time trends. A 1-

percentage-point increase in the unem-ployment rate is compensated for by a1.7-percentage-point decrease in the in-flation rate. Thus, if unemploymentrises by 5 percentage points (say from 3to 8 percent), the inflation rate mustdecrease by 8.5 percentage points (sayfrom 10 to 1.5 percent per year) to keepthe population equally satisfied. The so-called misery index, which simplyadds the unemployment rate to the in-flation rate, distorts the picture by at-tributing too little weight to the effectof unemployment, relative to inflation,on self-reported happiness.

6. Institutional Effects on Happiness

6.1 Basic Constitutional Rules

Peoples happiness is influenced bythe kind of political system they live in.

It should be expected that people livingin constitutional democracies are hap-pier because the politicians are moremotivated to rule according to their in-terests. If they disregard the wishes ofthe population, the politicians and par-ties in a democracy fail to be reelectedand hence lose their power. Democraticinstitutions, in particular the right toparticipate in elections and vote onissues, thus contribute to citizenshappiness.

Researchers on happiness havelooked at the interaction between de-mocracy and happiness. The extent to

which a constitution is democratic andallows its citizens to make decisions ac-cording to their own preferences can becaptured by various indices of freedom.

Figure 5 presents a graphical repre-sentation of a comprehensive measure

of freedom, combined with a 4-point



22/34

measure of happiness, in 38 mainly de-veloped nations at the beginning of the1990s (Veenhoven 2000). A visual testreveals that freedom and happiness arepositively related. The comprehensiveindex of the constitutional set-up usedin this figure refers to the following

three areas: Political freedom measures the possi-

bility for citizens to engage in thedemocratic process or, conversely, therestrictions on political participation.It is composed of two sub-indices, thefirst relating to civil rights, such asfreedom of speech (with eleven items),and the second to political rights (nineitems).

Economic freedom measures the op-

portunity for individuals to engage inthe free exchange of goods, services,and labor. It is based on sub-indices(each in turn composed of a number ofitems), referring to the security ofmoney, free enterprise, freedom fromexcessive taxation, and the possibility

of undertaking monetary transfers. Personal freedom measures how free

one is in ones private life, for exam-ple, to practice ones religion, totravel, or to get married.

To combine the sub-indices, averagez-scores are calculated. All threepolitical, economic, and personal free-domare strongly and statistically sig-nificantly correlated with happiness

(Veenhoven 2000). Controlling for

3.5

2

Figure 5. Freedom and Happiness across Nations

Economic, political and personal freedom

Happiness

Source:Veenhoven (2000), Appendix 1.

3

3

2.5

22 1 0 1



23/34

differences in per-capita income, thecorrelation with economic, but not po-litical and personal, freedom remainsstatistically significant. Analyses with sub-samples suggest that economic freedom

contributes to happiness particularly inpoor countries with a low level of gen-eral education, while political freedomis more strongly correlated with sub-

jective well-being in rich countrieswith a high level of education. Inboth cases, differences in income percapita are controlled for (Veenhoven2000).

Such studies are certainly illuminat-ing, but they can only inform us in alimited way about the influence of vari-ous constitutional conditions on subjec-tive well-being. Countries differ fromeach other in many ways, and it is notsufficient simply to control for unequalper-capita incomes to capture the influ-ence of democracy. Moreover, the cross-section studies only report correlationsand do not deal with causation. Even if

we ignore the other problems, it re-

mains open whether democracy fostershappiness, or whether happiness is aprecondition for democracy. It has beenargued, for instance, that high satisfac-tion with life in a population increasesthe legitimacy of the political regime inpower and it may thus foster democracy(Inglehart 1990, 1999). For Latin Amer-ica and Russia, one study (Graham andPettinato 2001b) indeed identified amutual dependence of pro-democracyand pro-market attitudes with well-being: both raise happiness, but happierpeople are also more likely to have pro-democratic and pro-market attitudes.

With due caution, it may be hypothe-sized that, for the respective respon-dents, there is a virtuous circle in whichattitudes favorable to democracy, to themarket, and to life satisfaction reinforceeach other.

In the following, we concentrate on

specific institutions of democracy in oneparticular country. This removes manyproblems, which may arise due to thedifficulties of internationally comparinghappiness scores and their determinants

across countries.6.2 Referenda

The possibility for citizens to directlyparticipate in politics is an importantfeature of democracy. The constitutionsof many countries allow popular refer-enda, but they are sometimes only usedas a device to inform the government

when it no longer knows what to do.Often it is used as a plebiscite in whichthe voters are asked to support the gov-ernments policy. In many cases, it isrestricted to local and sometimes trivialissues, while the decisions on importantissues are reserved for the professionalpolitical actors in parliament and gov-ernment. In the United States, thereare many popular referenda at the locallevel as well as in some states (espe-cially in California), but the constitu-

tion does not allow them at the nationallevel. The only country with an exten-sive set of direct political participationrights at all levels of government, and

with respect to all issues, is Switzer-land. Of the 728 referenda made in the

world at the national level between1900 and 1993, 357 (or almost 50percent) were held in Switzerland(see David Butler and Austin Ranney1994).

A referendum, in which all the citi-zens have the possibility to participate,meets the crucial requirement that itgives decision-making power to peopleoutside of the group of (professional)politicians. The constitutional settingdetermines to a large extent what issuesare put on the political agenda, and

what issues are prevented from appear-ing. In representative democracies,

politicians are often very skilled at not



24/34

letting issues that are to their disadvan-tage be discussed in the democraticinstitutions. For example, they usuallysucceed in not having their privileges(e.g., their incomes and pensions) dis-

cussed in open parliamentary sessions.In direct democracies, in which the citi-zens may put any issue on the ballot,the agenda is much less under thecontrol of the classe politique.

The effect of direct democracy onvarious aspects of society has been care-fully analyzed in a number of econo-metric studies for the United States.Findings include:

Government expenditure and revenuesare lower in institutions with directdemocracy (John Matsusaka 1995);

Per-capita debt is substantially lowerwith a referendum requiring a quali-fied majority (William McEachern1978);

Land prices are higher because peoplefind it attractive to live and work insuch communities (Rexford Santerre

1986); Public expenditures for education arehigher when a referendum is possible(Santerre 1989).

The following insights have beengained on the basis of econometricstudies for Switzerland:

A comparison of Swiss communes withdifferent degrees of institutionalizedforms of participation in political deci-sions reveals that the outcomes corre-spond more closely to the voters pref-erences, the more directly democraticthey are (Werner Pommerehne 1990);

The growth of public expenditure ismore strongly determined by demandfactors (i.e., by the citizens willing-ness to pay) than by supply factors (inparticular by the politicians and bureau-crats own interests) (Pommerehne and

Friedrich Schneider 1978);

Public supply is less costly, the moredirect the democratic institutions are(Pommerehne 1978);

Tax morale is higher than in repre-sentative democracies (Pommerehne

and Hannelore Weck-Hannemann 1996;Frey 1997);

Per-capita incomes in cantons withmore-developed direct participationpossibilities for citizens are signifi-cantly higher than in cantons with less-developed forms of direct participation(Lars Feld and Marcel Savioz 1997).

All these results control for a greatnumber of variables unrelated to directdemocracy. They provide strong evi-dence that deviation from citizenspreferences is indeed significantlylower in a referendum compared to arepresentative democracy.

The influence of direct democracy onhappiness has been analyzed using dataon reported subjective well-being forSwitzerland in 199294 (Frey andStutzer 2000). The major explanatory

variable is the institutionalized right ofindividual political participation viapopular referenda, which varies consid-erably between the 26 Swiss cantons.The estimates reveal that the extent ofdirect democratic participation possi-bilities exerts a statistically significant,robust, and sizeable effect on happinessover and above the demographic andeconomic determinants normally takeninto account. When the full variation inthe institutional variable is considered,i.e., when individuals in the canton withthe highest democracy index (Basel Land)are compared to citizens in the canton

with the lowest direct-participationrights (Geneva), the former state withan 11-percentage-points higher prob-ability that they are completely satis-fied. This effect is larger than livingin the top rather than in the bottom

income category.



25/34

6.3 Federalism

The decentralization of decision mak-ing is an alternative means for betterfulfillment of the voters preferences:individuals tend to leave dissatisfying

jurisdictions, while they are attracted tothose caring for the population s prefer-ences at low cost. The possibility to

vote with ones feet (Charles Tiebout1956; see also James Buchanan 1965;and Albert Hirschman 1970) tends toundermine regional cartels by politicians.The division of competence betweencommunities and the state government,or the extent of fiscal decentralization,

is thus another constitutional factor sys-tematically influencing happiness. Thestudy on Switzerland (Frey and Stutzer2000) measured the extent of localautonomy by an index based on surveyresults. Chief local administrators in1,856 Swiss municipalities reported onhow they perceived their local auton-omy using a 10-point scale. The esti-mate reveals a statistically significant

positive effect of decentralization onsubjective well-being. For local auton-omy, the proportion of people who indi-cate being completely satisfied with lifeincreases by 2.6 percentage points,compared to a situation in which thecommunes are one standard deviationless autonomous vis--vis their canton.

7. Summary and Implications

7.1 What Economists Can LearnThe insights gained from research on

happiness throw new light on importantissues analyzed in economics. Most im-portant, they enlarge the scope of em-pirical measurement and provide newtests for theories.

Happiness is not identical to the tra-ditional concept of utility in economics.It is, however, closely related. On the

one hand, the concept of subjective

happiness is a valuable complementaryapproach, which covers many moreaspects of human well-being than thestandard concept of utility. On theother hand, subjective well-being can

be considered a useful approximation toutility, which economists have avoidedmeasuring explicitly. This allows us toempirically study problems that pre-

viously were analyzed only on an ab-stract theoretical level. Moreover, theanalysis of data on subjective well-beingmay allow for discrimination betweencompeting explanations for empiricalfindings in behavior (for an application,see Stutzer and Lalive 2001). The op-portunities offered by information on

well-being and affect may not only en-rich field research, but also laboratoryresearch in experimental economics(Gary Charness and Brit Grosskopf2001 and James Konow and Joseph Ear-ley 1999). These extensions represent aconsiderable step forward toward asocial science able to provide usefulinformation.

7.2 Implications for Economic Policy

The insights gained about happinessare in many respects useful for economicpolicy undertaken by governments. Someexamples suffice to illustrate the point:

The use of measures of happiness al-lows for a new way of evaluating theeffects of government expenditure. Alltoo often, the effect is measured bythe cost incurred by the state: themore spent, the better. This is obvi-ously not always the case, and in someinstances lower expenditure wouldbe better. The problem has beenapproached scientifically by usingbenefit-cost analysis. The benefits arethe recipients marginal willingness topay, which is best measured in surveysby contingent valuation analysis. This

method can be complemented by



26/34

simulations using micro-econometrichappiness functions with a large num-ber of determinants that allow for theevaluation of the widespread effects ofextensive expenditure policies.

Welfare policy faces the question ofhow much economic destitution is re-sponsible for peoples unhappiness. To

what extent can persons with low in-come be helped by financial support?If low income is due to unemploy-ment, the research results suggest thatproviding people with higher incomescan only compensate for the pecuniaryeffect. In order to improve well-being,policy should rather be directed towardproviding appropriate employment.

An important part of antipoverty pol-icy deals with the question of whatpoverty is. Traditionally, the defini-tion relies on disposable income. Hap-piness research allows the problem tobe approached more fundamentally byconsidering reported levels of subjec-tive well-being. Such complementarymeasurement also allows equivalence

scales to be established (Erik Plug andvan Praag 1995). They indicate the in-crease in income necessary to compen-sate for a larger family, while main-taining the subjective well-being of thefamily.

Tax policy must consider to what ex-tent various income groups are af-fected. Is it possible to achieve socialgoals by redistributing income, or arethe negative effects on subjective well-being prohibitive? Recently it hasbeen argued that the fight for relativepositions is socially wasteful, and thatthe high-income recipients, as winnersof these status races, should be moreheavily taxed (Frank 1999; more gen-erally Layard 1980). This proposal hasbeen influenced by the findings ofhappiness research, which suggest thatpeople derive more satisfaction from

their position in comparison to other

income recipients than from the in-come level as such. If the redistribu-tive tax policy is able to make this raceless attractive, subjective well-beingmay be positively affected. But, for an

overall evaluation, this proposal mustconsider many additional aspects, inparticular what possibilities the high-income recipients have to evadeincreased taxes.

Happiness functions have sometimesbeen looked at as the best existing ap-proximation to a social welfare functionto be maximized (explicitly Di Tella,MacCulloch, and Oswald 2001, p. 340).The optimal values of the determinantsthus derived areaccording to this

viewthe goals economic policy shouldachieve. It seems that, at long last, theso far empirically empty social welfaremaximization of the quantitative theoryof economic policy (Jan Tinbergen1956, and Henry Theil 1964) is given anew lease on life.

Such an endeavor is still confronted

with fundamental problems of socialwelfare maximization (Frey 1983, pp.18294). While the shortcoming of em-pirical emptiness has been overcome(provided one is prepared to accepthappiness functions as a reasonable ap-proximation to a social welfare func-tion), the government still has little orno incentive to pursue such a policy.Only a benevolent dictator govern-ment would do so (Geoffrey Brennanand Buchanan 1985). Empirical analy-ses in public choice (Dennis Mueller1997) suggest that governments are notbenevolent and do not simply follow the

wishes of the population, even in well-functioning democracies, not to men-tion authoritarian and dictatorial gov-ernments. Hence, to maximize thehappiness function neglects the crucialincentive aspect. Therefore, the insights

from empirical analyses should serve



27/34

mainly as information on favorable eco-nomic and institutional conditions. Ifthey are considered to be convincing bypolitical entrepreneurs and citizens,they are taken up and are proposed in

the political process.7.3 Implications for Economic Theory

Happiness research adds consid-erable new insights to well-known theo-retical propositions. This has beenshown with the example of how income,unemployment, inflation, and democracyaffect reported individual well-being.

Effects of income. Most economiststake it as a matter of course that higherincome leads to higher happiness. Ahigher income expands individuals andcountries opportunity set, i.e., moregoods and services can be consumed.The few people not interested in morecommodities need not consume them;they are free to costlessly dispose of anyunwanted surplus. It therefore seemsobvious that income and happiness gotogether (provided, of course, that the

two are correctly measured).But there are also some economists

who do not subscribe to the idea thathigher income produces higher happi-ness. One of them is John Kenneth Gal-braith who, in his famous book The

Affluent Society (1958), pointed out thelimited use of higher private income

while the public sector is starving. Inthe first serious economic study of thedata on happiness, Easterlin (1974)concluded that money does not buyhappiness. Another author claimingthat the most cherished values cannotbe bought on markets is Tibor Scitovsky

with his Joyless Economy: The Psychol-ogy of Human Satisfaction (1976).Scitovsky even argues that a high levelof wealth brings continuous comfortsand thereby prevents the pleasure thatresults from incomplete and intermit-

tent satisfaction of desires. More re-

cently, Frank, in his Luxury Fever(1999), emphasizes that ever-increasingincome and consumption do not bringhigher happiness.

The empirical research on happiness

has clearly established that at a givenpoint in time, and within a particularcountry, persons with higher incomeare happier. Over time, however, happi-ness in western countries and Japandoes not systematically increase, de-spite considerable growth in real per-capita income. This can be attributed tothe rise in aspiration levels going withincreases in income. Betwee

What Can Economists Learn From Happiness Research

Documents

Transcript of What Can Economists Learn From Happiness Research