Munich Personal RePEc Archive
Economic growth and well-being beyond
the Easterlin paradox
Sarracino, Francesco and O’Connor, Kelsey J.
STATEC Research, Laboratory for Comparative Social Research,
Higher School of Economics
13 September 2019
the Easterlin paradox
Franeso Sarraino
and Kelsey J. O'Connor
y
September 13, 2019
Abstrat
Reentstudiessuggestthateonomigrowthandwell-beingangrowtogether inthe longrunin preseneofgeneroussoialsafetynets,inreasing soialapital anddelining inomeinequality. We put theseonditionsto atest in anattempt to explain the absene of a relation between eonomi growth and well-being in Luxembourg. To this aim we apply an error orretion model to a panel of 15 Western Europeanountries,and we usethe resultsto preditlife satisfation in Luxembourgbetween1991and2015. Wendthattheattrendoflifesatisfation inLuxembourgislikelytheresultoffourforesatinginoppositediretions. This suggeststhattheavailablelistofmoderatingonditions{althoughnotexhaustive {isapromisingstartingpointtodesignnewpoliiestodurablyimprovewell-being.
Key-words: time-series; subjetive well-being; errororretion model; life satis-fation;dynamis; inlusivegrowth.
JEL odes: I31;O11;E6;O21;D60.
Institut national de la statistique et des etudeseonomiques du Grand-Duhe du Luxem-bourg(STATEC),andLCSRNationalResearhUniversityHigherShoolofEonomis,Russian Federation. Email: Franeso.Sarrainostate.etat.lu.
y
The nding of no long-runrelationship between eonomi growth and subjetive well-being is ontroversial. This result, ontraditing a positive ross-setional relation,ametobeknownastheEasterlinparadox(Easterlin,1974). Sine,many sholars have ontributed with diverging views. There are those who oppose (see e.g. Easterlinetal.,2010;Bruniand Stana,2008;Easterlinand Angelesu, 2009; Behetti et al.,2011; Clark et al., 2014; Easterlin,2017) and those who support (seee.g. StevensonandWolfers,2008;Deaton,2008;Saksetal.,2012;Veenhoven and Vergunst, 2013) eonomi growth as a way to improve well-being. Beyond the views onerning simply whether or not growth will aompany inreasing well-being,some reentevidene suggests therelationdependsonsoial,politial, eonomi, ultural,and institutionalonditions: if eonomigrowth isompatible with a ohesive and inlusive soiety, it is reasonable to expet that well-being willimprove(Oishi and Kesebir, 2015;Mikuka etal.,2017; Easterlin, 2013; Ono and Lee, 2013). In ontrast,if eonomi growthleads tolonelinessand inequality, well-being may arguably deline.
Althoughthequestforonditionsof\inlusivegrowth"{agrowththatbenets all the members of a soiety { is still in its infany, past evidene provides a preliminaryexplanationofhowand whenapositiveorrelation between eonomi growthandwell-beinganexistovertime. Thisisimportantbeauseitansuggest ways to promote well-being. Our aim is to distill the evidene on the onditions aeting the eonomi growth - well-being gradient, to explain the at trend of life satisfationin Luxembourg.
We fous on Luxembourg as an example of the lak of orrelation between eonomi growth and well-being in the long run. Panel 1ainFigure 1shows that sine the early 1980s, Luxembourg experiened substantial eonomi growth, at least until the eonomi risis of 2008. Yet, the share of very satised people did not hange substantially over time.
1
The eonomi risis may explain what happened after 2008, but prior to 2008, it is not lear why eonomi growth did not improvepeople's well-being (froma traditionaleonomis perspetive).
Apossibleexplanationisthatlifesatisfationisanunreliablemeasure. Yet,we havereasonstobelievethatlifesatisfationisreliablebasedonavailabledata and previousliterature. Panel1breportstheshareofverysatisedpeopleaordingto Eurobarometer(EB),EuropeanValuesSurvey(EVS)andEuropeanQualityofLife Survey(EQLS).Fortheyearswhenthedataarejointlyavailable,thethreesurveys
1
(GNI)perapita(Panel B) inLuxembourg inthe period1981-2015. The samples onsist onlyof native born individuals.
40000
60000
80000
100000
120000
140000
GNI per capita (constant 2010 US$)
25
35
45
55
65
Share of very satisfied people (%)
1980
1990
2000
2010
2020
% very satisfied
GNI per capita (constant 2010 US$)
(a)Lifesatisfation(Eurobarometerdata)andrealGrossNational Inome perapita(World Development Indiators).
25
35
45
55
65
Share of very satisfied people (%)
1980
1990
2000
2010
2020
EB
EVS
EQLS
(b) Life satisfation in Luxembourg. For the years when three dierent data-set are jointly available, they provide omparable information.
Note: Werestritouranalysisonnativesonlytoensuretheomparabilityofdata onlife satisfationovertime. Eurobarometerdataolletedbefore1994provideddataonnationals only,andafter1994theyinludedinformationonimmigrantsfromotherEuropeanountries. WedisussthisissueinmoredetailinsetionData.
provided evidene supporting the reliability and validity of life satisfation as a measure of subjetive and objetive well-being (Blanhower and Oswald, 2004; VanReekum et al., 2007; Shimmak et al.,2010; Kahneman and Krueger, 2006; Layard, 2005).
Anotherpossibleexplanationisthatthetrendsoflifesatisfationarealwaysat - atleast amongthe rihest and most developed ountries inthe world. However, the evidenedoesnotsupport this view. Figure2shows the trendsof theshare of verysatised peoplein Belgium,Frane,Germany and the Netherlands, i.e. aset ofWesternEuropeanountrieswhihare losetoLuxembourg. Thepitureshows thatthe trendsoflife satisfationare not alwaysat: althoughaverage levelsmay dier, the trends in Frane and Netherlands are monotonially positive, whereas the trend is rather at in Belgiumand it follows a `J' urve in Germany. This is onsistent with previousstudies (Sarraino, 2012).
Figure 2: Trends of the share of very satised people in a sample of Western European ountries.
10
20
30
40
50
10
20
30
40
50
1980
1990
2000
2010
20201980
1990
2000
2010
2020
Belgium
France
Germany
Netherlands
Share of very satisfied people (%)
Graphs by Country
together beause four fators ated in opposite diretions for well-being, namely inreasing soial apital and soial expenditures { whih are expeted to have a positive impat on well-being { and inreasing inome inequality and unemploy-ment{ whih,on the otherhand, ould have anegativeimpat.
Unfortunately, we do not have miro data providing long time-series for Lux-embourg: the European Value Study inludes individual data olleted in 1999 and 2008, the European Soial Survey was administered in 2002 and 2004, and the European Quality of Life Survey provides data every four years sine 2003. Thus, weadoptamaro-eonomiperspetive. Speially,weapply anerror or-retionmodeltoapanelof15WesternEuropeanountriestoexplainountry-year levelsof lifesatisfationusing theset of potentialexplanatoryfatorsidentiedin previous literature. The results are then used to predit life satisfation and to assesswhether andtowhatextenttheexplanatoryfatorsexplainthe trendoflife satisfationin Luxembourg.
We build our argument in two steps: in setion 2 we review the literature onthe Easterlin paradox and on the fators moderatingthe relationship between eonomi growth and well-being. Subsequently, we apply a time-series approah (setion3) toa marodata-set onsisting of 15WesternEuropean ountries (se-tion 4). Setion5illustratesthe resultsof the model,whihwe use topreditand explain the trend of life satisfationin Luxembourg. The lastsetion onludes.
2 The Easterlin paradox and moderating fators
inreasing subjetive well-being (Oishi and Kesebir, 2015; Mikuka et al., 2017). The literature identied three fators whih plausibly aet the relation between eonomi growth and well-beingover time: inome inequality (Oishi andKesebir, 2015; Mikuka et al., 2017), soial apital (Uhlaner, 1989; Helliwell, 2003, 2008; Bartoliniet al., 2013; Clark et al., 2014), and soial poliy (Easterlin, 2013; Ono and Lee, 2016).
Conerning inomeinequality,the evidene about the ross-setional relation-ship with well-being is mixed (e.g. Alesina et al., 2004; Clark and D'Ambrosio, 2015). Theseontraditionsmayarisebeausethe relationshipbetweeninequality and well-being depends on a ountry's level of development (Jiang et al., 2012; Iniguez-Montiel, 2014). However, previous studies found that inreasing inome inequalityisonsistentlynegativelyrelatedtowell-being(BartoliniandSarraino, 2015; Oishi and Kesebir, 2015; Mikuka et al., 2017). By widening the possibil-ities to establish soial omparisons, growing inome inequality undermines the positiveeet ofinomegrowth forwell-being. Raisinginomeinequalityan also undermine well-being by reduing feelings of fairness and trust in others (Oishi et al.,2011) or by weakening soiallinkages and feelings of ooperation (Graham and Felton, 2006; Oishi etal.,2011).
SoialapitalisdenedbytheOECD(2001)as\networkstogetherwithshared norms, values and understandings that failitate o-operation within or among groups". A well-established literature shows that soial apital orrelates pos-itively with subjetive well-being at both the individual (Helliwell et al., 2017; Clark et al., 2014; Behetti et al., 2009) and aggregate level, over time within ountries (Bartoliniet al.,2013; Bartoliniand Sarraino,2015; Brokmann etal., 2009;Easterlinetal.,2012)andinountrypanels(BartoliniandSarraino,2014). Helliwelland Aknin(2018)disussindetailtherelationshipbetween soialapital and subjetive well-being.
averagingabout 25perentabove the 1990svalue (Easterlin, 2009, p. 142). In China, life satisfation exhibited a similar pattern of ollapse and reovery following the transition, all the while growing at an average annual rate of more than9.0%. Brokmannetal.(2009),Easterlinetal.(2012),Easterlinetal.(2017) andBartoliniandSarraino(2015)disusspossibleexplanationsforthesestartling fats. Eah work partially attributes the deline in life satisfation to inreased soial omparisons, espeially failitated by rising inome inequality. Bartolini and Sarraino (2015) doument the importane of soial apital, estimating that nearly 19.0% of the well-being loss in China is related to a derease in soial apital. Easterlin et al. (2012) and Easterlin et al. (2017) instead emphasize the role of rising unemployment
3
, whih was inversely related to life satisfationover the full yle from 1990 to 2010 (while inequality, in ontrast, rose throughout the period). And, like in the European transitionountries, with unemployment ame not onlyinome losses, but also the eliminationof soialbenets. The loss of these benets arguably signiantly exaerbated the eets of unemployment. Soialsafetynetsarepositivelyrelatedtolifesatisfationingeneral(DiTellaetal., 2003; Rothstein, 2010; Paek and Radli, 2008; Boarini et al., 2013; Easterlin, 2013; Ono and Lee, 2016; O'Connor, 2017), not just intransition eonomies, and the assoiation is not limited to those diretly aeted (e.g., unemployed) (Carr and Chung, 2014). In sum, the deline inChinese well-being an be explained by (1) inreasing inome inequality whih failitated inreasing soial omparisons, (2) delining soial apital, and (3) inreasing unemployment aompanied by a severely redued soial safety net. The reent reovery appears to be driven by improvements in trust, employment, and the soial safety net (Easterlin et al., 2017).
Previous studies investigating the Easterlin paradox and its moderating fa-tors fousedmainlyon ross-ountry studies oronountries providing\negative" examples, i.e. ountries in whih eonomi growth and inreasing well-being do not gotogether. The aseof Japan standsout asa\positive"example: aountry where eonomi reforms in the early 1990s shifted the ountry from a pattern of rampanteonomigrowthandstagnantwell-being,tooneofmoderategrowthand inreasing well-being (see gure 3). The question then beomes: what made this hange possible?
Bytheend ofthe1980s,Japanwasinthemiddleoftworisis: ononeside, the demographirisis;onthe other,the delineinthe viabilityof thetraditionaland
2
It is possiblethat asymmetri responses to eonomi ollapseand positive inome growth ould explain why life satisfation did not fully reover at the same time as GDP (e.g., from lossaversionDe Neve et al.(2018)),but that is insuÆient toexplain thepattern in China as disussedinthenextparagraph.
3
Japan between 1981 and 2010.
6
6.2
6.4
6.6
6.8
7
Average life satisfaction
25000
30000
35000
40000
45000
GDP per capita
1980
1990
2000
2010
Year survey
GDP per capita
life satisfaction
Note: Life satisfation dataare fromthe WVS, whereas GDP gures,presented inreal dollarswith base year set to2010, are issued fromthe World Development Indiators of the World Bank. The trends in lifesatisfation fromWVS are roughly onsistent withthose issued from other soures.
eonomistagnation andinternationalompetition,putpressure onthesheme of soial safety nets whih traditionally relied on intergenerational support and on generousbenetsfortheemployeesoflargeorporations. Forinstane,theshareof three-generation-familyhouseholds went from54 perent in1975 to 13perentin 2013(Ministry ofHealth, Labourand Welfare,2014),whereas the shareof elderly people livingalone nearly doubled. Atthe same time, eonomi onditions fored ompanies to limit the benets granted to their employees, and in partiular to newly hired personnel. Moreover, the likelihoodof lifetime employment delined (Ono,2010). Theshareofworkersinnonstandardemploymentmorethandoubled from 15 to 38 perent between 1984 and 2016 (Ministry of Health, Labour and Welfare, 2014). Consequently, the population inneed of soial protetion greatly expanded during the 1990s, as wellas inomeinequality (see gure 4).
Figure 4: Evolution of the Gini index of inomein Japan.
25
30
35
40
Gini index
1980
1990
2000
2010
SWIID Post
WIID
Note: Lowess smoothed urves. The two lines inthe hart referto measures of Giniissued fromtwo dierent souresof data: the Standardized World Inome Inequality Database, and the World Inome Inequality Database. Together, the two series of data provide onsistent evidene that inomeinequality inJapan inreased sine 1980.
Soure: Solt (2016) and UNU-WIDER (2018).
ies were introdued in the mid-1990s with the aim of improving the living and health onditions of elderly people, alleviating the osts of having hildren, and failitating women aess to the job-market. Figure 5 shows the trend of welfare state generosity inJapan (Sruggs etal.,2017).
Figure 5: Average life satisfation and Generosity of Welfare State (Japan 1981-2010).
6.5
6.6
6.7
6.8
6.9
Life satisfaction
1
1.02
1.04
1.06
Index (1990 = 100)
1980
1990
2000
2010
Gen. of Welfare State Policy
Life satisfaction
Note: Lowess smoothed urves. Soure: Sarraino et al.(2019).
Available studies indiate that: i. soial apital, soial safety nets, and inome inequality aet the relationship between eonomi growth and well-being over time; ii. poliy-makers an adopt poliies to promote well-being in the long run. Our aim is to assess whether the fators disussed above an help explain the at trend of life satisfation in Luxembourg. This test is important to evaluate the reliabilityof available knowledge about the onditions to promote well-being in the long run, and to identify possible areas of intervention for poliy-makers. Additionally,inpresentworkweextendthelistofmoderatingonditionstoinlude unemployment. Itiswellestablishedthatunemploymentisoneofthemajorauses of ill-being. Thus it is possible that the hanges in unemployment ontribute to explainingthe trendof life satisfation.
3 Method
We use an error orretion model (ECM) to analyze the fators that ontribute to life satisfation in the long-run. The main reason is that ECMs allows us to estimate onsistent long-run relations between the explanatory variables and dependent variable. Additional reasons are more tehnial. First, explanatory variables in levels (e.g., GDP p) often exhibit unit root properties, whih ould lead to the estimation of spurious relations (Engle and Granger, 1987). First-dierening thevariablesanbeusedtoaddresssuhspuriousrelations,but rst-dierening limits the interpretation of the results to short-run hanges. ECMs separately estimate the short- and long-run relations to avoid spurious relations (underertainonditionsdisussedbelow). Also,theestimatedlong-runrelations, referred to as long-run eets in the time-series literature, are onsistent in the preseneofshort-runreverseausality(ChudikandPesaran,2015;Pesaran,2015). Before presenting the ECM, we begin with our assumed data generating pro-ess, represented by Equation 1. LS
it
represents life satisfation for ountry i at time t, the vetor X
i;t
inludes the explanatory variables, and xed ountry harateristis are represented by
i . LS it = i LS i;t 1 + 0 i0 X i;t + 0 i1 X i;t 1 + i +" it (1)
a large hange in X i;t
, the response in the LS i;t
might overshoot the long-run equilibriumrelationship. When this happens, the error orretion term,, serves tobring the relationship bak to the long-runone.
For an error orretion parameterization to be appropriate, (1) the error or-retiontermshouldbestatistiallysigniant,negative,but greaterthan negative two, and(2)theremust be along-runointegrating relationshipbetween the level variables. Condition 2 isneessary for the term (LS
i;t 1
0 i
X i;t
) tobe station-ary, whih is neessary for the error term to be stationary when is statistially signiant. Condition 1 is tested diretly in the regression analysis. To hek ondition 2 we test the regression residuals for stationarity using panel unit root tests (Fisher type augmented Dikey Fuller tests). The results indiate that the residualsare indeedstationary and ondition 2 ismet.
WeadjustEquation 2todevelop our nalspeiation(presented below). No-tietheoeÆientsareindexedbyiinEquation2,indiatingthattheyareallowed to vary aross ountries. We allow the short-run relations to be heterogeneous in order to apture exibly any reverse ausality. However, inour nal speiation we assume the long-run eets () are ommon aross ountries. Without this assumption, we ould not simultaneously assess eah of the fators disussed in Setion 2 that may explain the trend of life satisfation in Luxembourg. This speiation is referred toas apooled mean groupmodel(PMG).
LS it = i (LS i;t 1 0 X i;t )+ 0 i X i;t 1 +' i LS t +Æ i X t + i +" it (3) 4 Data
Individual life satisfation data are from repeated ross-setional Eurobarometer (EB) surveys (European Commission, 2018). Life satisfation is measured on a sale from1to4usingthe responsestothe question,\On the whole,are youvery satised, fairly satised, not very satised or not at all satised with the life you lead?". This question was rst asked in1973 in seven ountries and ontinues to-day formore thanthe present 28EU ountries. In eahyear, multiplesurveysare ondutedthat ask about lifesatisfation. Annualobservationsof lifesatisfation wereonstrutedineahountryastheweightedproportionofnative-born individ-ualsreportingthetopresponseategory,\verysatised". Foreign-bornindividuals were neessarily exluded, beause prior to 1994, the Eurobarometer target pop-ulation only inluded native-born individuals,and in 1994, the target population expanded but stillexluded individualsborn in non-EU ountries (Shmitt etal., 2009, p. 56).
TheexplanatoryvariablesinludethenaturallogofrealGrossNationalInome perapita(GNI),unemploymentrate,theGinioeÆientofinome,soial expen-ditures,and trust inothers. GNI and unemploymentdata (nationalestimate)are from the World Development Indiators(WDI) (World Bank, 2018). We use the Gini oeÆient of inequality in equivalent household disposable (post-tax, post-transfer) inomefrom the Standardized World Inome Inequality Database (Solt, 2016).
4
Dataforsoialexpendituresperapitaareavailableeveryveyearsfrom 1980-2015 and 2016 from the OECD Soial Expenditures Database (OECD, 2018). The variable inludes all publi soial expenditures on ative labour market pro-grammes, family, health, housing, inapaity related, old age, other soial poliy areas, survivors, and unemployment. We adjusted the variable to more losely represent the generosity of the welfarestate poliy. Coneptually subjetive well-beingrelatestothegenerosityofpoliiesnottoexpenditures;thatisbeausesoial
4
reases. Indeed O'Connor (2017) nds, soial expenditures relate to subjetive well-being, but the relation beomes statistially insigniant when exluding a ontrol for the old age dependeny ratio. In the present analysis, we ould also inlude the old age dependeny ratio, but given the small number of degrees of freedom,wehose insteadtopartialoutthe oldage dependenyratioandalsothe unemployment rate. Speially, weused the residualsfrom aregression of soial expenditures on the old age dependeny ratio and unemployment rate. We then use linear interpolationto failitateannual analysis.
Trust in others is based on responses to the question, \Generally speaking, wouldyousaythatmostpeopleanbetrusted,orthatyououldnotbetooareful in dealing with people?" Individual responses are obtained from Eurobarometer surveys in the years 1986, 2004, 2009, 2010, and 2014. These responses are then aggregatedatthe ountrylevelforeahyear asthe portion ofpeoplefeelingmost people an be trusted. However, omparison over time is limited by dierenes in the response sales. The largest hange ourred beginning in 2009, when the sale went from two disrete hoies to a sale ranging from 1 to 10. In order to produe annual estimates and aount for the hange insale, we imputetrust in the followingsteps:
1. The weighted perentage of people stating most people an be trusted is alulated by ountry year. For the years 2009, 2010, and 2014 responses 7-10 are reorded as most people an be trusted;
2. These sores are de-meaned by subtrating the average levelof trust within a ountry over the years 1986 and 2004 (the years based on the previous response sale);
3. The de-meaned trust values are then linearly interpolated and extrapolated over the sample period with anexeption { trust is not extrapolated to the years before2004 if it is unobserved in1986;
4. Additionaldata from the World Values Survey (WVS, 2014) and European ValuesStudy (EVS,2011)isusedtoprovideadditionalinformationontrust. The two surveys provide dihotomous answers to a question asking respon-dentswhetherpeopleanbetrusted. Asinstep1,theanswersarede-meaned (withinountry) and extrapolated;
intera-and EVS/WVS trust; and
6. Last, the ountry means from step 2 were added bak to obtain our nal predition of trust.
Our sample of ountries inludes the rst fteen European Union member states (EU15) beause only these ountries have suitably long enough series to be inluded. The period of analysis inludes the years 1991 { 2016. Our sample for regression analysis begins in 1991 to ensure there were at least ten ountries observed in eahyear. Prior to 1991 data for fewer ountries were available when inluding lags. It is important to use as many ountries as possible beause the analysis inludesross-setional means in eah period. We would prefer tobegin the sample with more than ten ountries, but data for all 15 ountries are not available until 1995, and with lags, that would signiantly redue the time di-mension. Table 1 presents the sample harateristis and average variable values for eah ountry.
Table 1: Desriptivestatistis
Country First Last Lifesatisfation GrossNationalInome Giniindex Unemployment Trust Adj.SoialExpenditures year year (%veryhappy) ln(US$perapita) (0-100) rate(%) (%anbetrusted) ln(US$perapita)
Austria 2001 2016 23.95 10.74 27.58 5.10 47.78 0.42
Belgium 1991 2016 27.35 10.62 25.75 7.99 35.10 0.16
Denmark 1991 2016 65.82 10.92 23.85 6.16 78.45 0.51
Finland 2001 2016 33.11 10.73 25.37 8.63 70.02 0.27
Frane 1991 2015 15.90 10.57 28.69 9.83 29.14 0.15
Germany 1991 2015 17.85 10.57 27.38 7.86 45.06 0.13
Greee 1991 2016 8.29 10.07 33.58 12.98 37.04 -0.67
Ireland 1991 2016 36.02 10.51 31.13 9.48 37.67 -0.22
Italy 1991 2015 11.41 10.45 32.99 9.83 30.93 -0.07
Luxembourg 1991 2015 43.39 11.25 26.67 3.80 35.68 0.89
Netherlands 1991 2016 47.47 10.73 26.14 4.96 64.32 0.17
Portugal 1992 2015 4.01 9.93 34.03 8.03 31.60 -0.73
Spain 1992 2016 18.17 10.24 32.93 17.38 35.53 -0.44
Sweden 2001 2016 46.83 10.86 25.20 6.77 71.07 0.40
UnitedKingdom 1991 2016 35.30 10.49 33.83 6.74 44.17 -0.24
Sampleaverage 28.76 10.57 29.22 8.48 45.06 0.02
5 Results
Luxembourg.
24
25
26
27
28
29
Gini index
25
35
45
55
65
Share of very satisfied people (%)
1980
1990
2000
2010
2020
Share of very satisfied people
Gini index of eqv. hh. disp. income
(a) Gini index of equivalent household disposableinome.
0
2
4
6
8
Unemployment (% of tot. labor force)
25
35
45
55
65
Share of very satisfied people (%)
1980
1990
2000
2010
2020
Share of very satisfied people
Unemployment (% of total labor force)
(b) Unemployment as a perentage of total laborfore.
Soure: authors'ownelaboration.
The inreases in soial trust and soial expenditures, on the other hand, are expetedtohavepositivelyontributed tolifesatisfation. Sine1980the shareof people who feelthat others an betrusted nearly doubled (see Panel 7ain Figure 7),whereassoialexpendituresrose from8,190US$perapita(baseyear2013)in 1980 to23,880 in2015, i.e. a nearly 3 foldinrease (Panel 7b). It is possible that the eets onlife satisfation of inreasing inome inequality and unemployment, on one side, and inreasing soial trust and expenditures, on the other, o-set eah other. To test this hypothesis formally, we turn to the results of the error orretion model.
Table 2 presents the results of the error orretion model. The rst ve rows present the long-run relations orresponding to the in equation 2; ECT orre-sponds to the ; the middle rows present the short-run-hange relations orre-spondingtothes,andthe nalrows the ross-setionalmeansoflifesatisfation and the independent variables in levels. The rst ve olumns use one explana-toryvariableatatime,olumn6reportstheresultsfromthemodelinludingeah explanatory variable, and 7, standardized oeÆients of the long-run eets from the full model.
The long-runeets generally orrespond with our expetations, with the ex-eption of the Gini oeÆient, whih is statistiallysigniant and positive. Per-manent inreases in GNI p, inome inequality, soial trust, and adjusted soial expenditures are positively related to life satisfationin the long run, and unem-ployment, negatively.
5
Luxembourg.
20
30
40
50
% people trusting others
25
35
45
55
65
Share of very satisfied people (%)
1980
1990
2000
2010
2020
Share of very satisfied people
Pred. % People Can Be Trusted
(a)Shareofpeopledelaringthatothers an be trusted.
5
10
15
20
25
Social expenditures p.c. (2013 US$/1000)
25
35
45
55
65
Share of very satisfied people (%)
1980
1990
2000
2010
2020
Share of very satisfied people
Social Expend. pc Interp./1000
(b)Soialexpendituresperapita.
Soure: authors'ownelaboration.
reduedmodels (ols. 1-5)andthe fullmodel(ol. 6). The magnitudesand signif-ianeof GNI pand soialexpenditures are redued. Indeed soialexpenditures are no longer statistially signiant. This nding is surprising in light of the positive relations found in ross-setional evidene; however, insigniane ould bedue tomultiollinearityandlowstatistialpower(realladjustedsoial expen-ditures are positive and signiant in olumn 5). The magnitudes of the other variables (Gini, unemployment rate, and trust) inrease in size. Aross variables, trust has the largest standardized oeÆient. The oeÆient of trust in others is more than two times bigger than the one of GNI and nearly three times larger than the one of inequality or unemployment (in absolute terms). This indiates that trust inothers is the strongest orrelate of the hanges of life satisfationin thelong runamongthe onsideredvariables. Themagnitude,however, issmall: a onestandarddeviationdiereneintrustisrelatedto0.6perentagepointgreater life satisfation. The standard deviationof life satisfation overthe full sampleis 17.5 perentage points.
It issurprising thatthe GinioeÆientis positivelyrelatedtolife satisfation, however, as mentioned in the literature setion, positive relations have been ob-tained in ross-setional studies. The Hirshman tunnel eet ould explain the relation { inreasing inome of a few, leading to greater inequality, may signal that the inomesof everyone are inreasing, thereby raising subjetive well-being. Futureresearhshouldfousfurther oninequalityandrevaluateitinatime-series ontext.
a permanent inrease in GNI in a partiular ountry and year is assoiated with a long-runinrease in life satisfation, when GNI inreases in all ountries in the same year, the impat is negative. A three perent growth rate in eah ountry is assoiated with a derease in life satisfation of 0.5 perentage points per year (basedonolumn6inludingthe longrun eetand ross-setionalmeanbut not the insigniant short-run relations: -0.05 = 0.03 * 10.6 + 0.03 * -27.4). This suggests that GNI positively aets the life satisfation in a partiular ountry if it grows ata signiantly greaterrate than inother ountries: tobreak even, the GNI hange in a ountry needs to be 27.4/10.6 = 2.58 times the average hange aross ountries. For the other four fators, the impats of the signiant ross-setional means are also in the opposite diretion but they are smaller than the long-run eets. Conerning mean life satisfation, the impat of ross-setional life satisfation is positive indiatingpositive spill-overeets. If life satisfation inreases in one ountry, then life satisfation willinrease in the other ountries by approximately one fteenth of that inrease (the oeÆient rounds to 1.0 in olumn6).
Short-run variation invarious fators has theoretially distintimpats onlife satisfationompared to permanent inreases, espeiallyfor inomeas illustrated inBartoliniandSarraino(2014). Forinstane,wewouldexpetshortrunhanges in unemployment to be signiantly (and negatively) orrelated to the hanges of life satisfation. However, the present study annot omment on the short-run relations. The short-run variation is not independent and the relations are generally statistiallyinsigniant.
ountries (1991 -2016).
Lifesatisfation oeÆients
(1) (2) (3) (4) (5) (6) (7)
GNI 38.274 10.648 0.259 (3.990) (6.004) Gini 0.526 0.734 0.190 (0.249) (0.231)
Unemploymentrate -0.437
-0.572 -0.181 (0.249) (0.134)
Trustinothers 0.175
0.494 0.570 (0.082) (0.046)
Soialexpenditures 19.689
0.695 0.023 (3.449) (3.101) ECT -0.745 -0.741 -0.693 -0.610 -0.770 -1.036 (0.049) (0.062) (0.061) (0.075) (0.043) (0.090)
GNI -6.238 8.324
(9.132) (12.123)
Gini -0.424 0.441
(0.538) (1.183)
Unemploymentrate 0.085 0.202
(0.212) (0.309)
Trustinothers 0.162
-0.170
(0.060) (0.171)
Soialexpendituresp) 5.943 -15.673
(16.668) (28.448) MeanLifeSatisfation 0.955
0.869 0.860 0.868 0.926 0.984 (0.139) (0.140) (0.137) (0.149) (0.123) (0.169)
MeanGNI -30.443
-27.442
(5.410) (16.434)
MeanGini -0.475 -2.541
(0.923) (0.877)
MeanUnemploymentrate 0.352
0.319
(0.131) (0.291)
MeanTrustinothers -0.118 -0.399
(0.124) (0.232)
MeanSoialexpenditures -15.231
15.418 (4.621) (11.047)
Constant 14.287 -2.074 -5.714
-5.191 -5.112 222.303 (63.727) (25.398) (2.846) (6.063) (3.342) (173.954)
N 353 353 353 353 353 353
Note:Soialexpendituresareperapita,adjustedandtransformedinlogarithm.GNIisperapita andtransformedinlogarithm.Formoredetails,please,refertosetion4.
Standarderrorsinparenthesis.
p<0:10,
p<0:05,
−20
0
20
40
60
80
Percent Responding Very Satisfied
1990
1995
2000
2005
2010
2015
Pred. % Very Satisfied
% Very Satisfied
Previousliteraturesuggests thattherelationshipbetweenwell-beingandeonomi growth depends ona set of onditions: if eonomi growth isaompanied by ex-tensive soial safety nets, high soial apital, and low inome inequality, than it is likely to be assoiated with inreasing well-being. In this artile we test this view. In partiular,wehek whether the at trend of life satisfationin Luxem-bourg, despite a growing eonomy, an be explained by the onditions identied inprevious literature.
Between 1980 and 2008 { the year of the eonomi risis { the Gross National Inome per apita in Luxembourg grew by 6.35% yearly, while the share of very satised people remained onstant at about 35%. If the evidene from previous studies is orret, we should expet that the onditions mentioned above have a zero net eet on life satisfation. Unfortunately, the lak of long time-series of miro data prevents a miro-eonomi analysis in Luxembourg. We thus adopt a maro-eonomi perspetive, and we apply an error orretion model to a panel of 15 Western European ountries to predit life satisfation in Luxembourg on the basis of a known set of explanatory fators. These are: inome inequality, unemployment,soialexpenditures, and trust inothers {a ommonlyusedproxy of soialapital.
WendthefatorsexplaintheattrendoflifesatisfationinLuxembourg rea-sonably well and broadly onsistently with expetations. Inreases in unemploy-mentosetthe positiveinuenes ofinreasing soialtrust and eonomi growth. On the ontrary, inreases in inome inequality apparently positively aeted life satisfation,and soial expenditures, did not have the statistiallyrobust impat onwell-being that we expeted. Aross fators, the single most impatfulis trust in others. Standardized oeÆients indiate that the long-run eet of trust is nearly twie the eetof eonomi growth. Wealsofound that theross-setional average of GNI per apita attrats a signiant and negative oeÆient. Usu-ally ross-setional averages are inluded in the regression to ontrol bias due to ross-ountry orrelations,suh as the impat of European Union poliies, but in this ase,average GNIplays anunexpeted role. Eonomigrowth inapartiular ountry has aweak, positive,impaton well-being, but when eah ountrygrows ata similar pae the total eet onwell-being is negative;this is reminisentof a \soialomparisons"eet,butthis timearossountries. Asfarasweknow, this is the rst time that suh evidenehas been doumented ina panel of ountries.
ouldhangewith newdata ormethods. Indeed,the availabilityofa newwave of data fromthe EuropeanValue Study willsoonallowustoperformamiro analy-sisovering the period1999-2019, and thereforeto gaina more rened knowledge about what happened tolife satisfationin Luxembourg.
Alesina,A.,DiTella,R.,andMaCulloh,R.(2004).Inequalityandhappiness: are Europeans and Ameriansdierent? Journal of Publi Eonomis,88(9):2009{ 2042.
Bartolini,S.,Bilanini,E.,andPugno,M.(2013).Didthedelineinsoial onne-tions depress Amerians' happiness? Soial Indiators Researh, 110(3):1033{ 1059.
Bartolini,S.andSarraino,F.(2014). Happyforhowlong? Howsoialapitaland eonomigrowth relate tohappiness overtime. EologialEonomis, 108:242{ 256.
Bartolini,S.and Sarraino,F. (2015). Thedarkside ofChinese growth: delining soial apital and well-being in times of eonomi boom. World Development, 74:333{351.
Behetti,L., Giahin Ria,E.,and Pelloni,A.(2009). The60esturnaroundasa test onthe ausal relationshipbetween soiability and happiness. Eonometia Working Papers wp07, Eonometia.
Behetti, L., Trovato, G., and Bedoya, D. (2011). Inome, relationalgoods and happiness. Applied Eonomis,43(3):273{290.
Beja, E. L. (2014). Inome growth and happiness: Reassessment of the Easterlin Paradox. International Review of Eonomis, 61(4):329{346.
Blanhower, D. andOswald, A.(2004). Money, sexand happiness: An empirial study. The Sandinavian Journal of Eonomis,106(3):393{415.
Boarini, R., Comola, M., de Keulenaer, F., Manhin, R., and Smith, C. (2013). Can governments boost people's sense of well-being? The impat of seleted labourmarketandhealthpoliiesonlifesatisfation. Soialindiators researh, 114(1):105{120.
Brokmann, H., Delhey, J., Welzel, C., and Yuan, H. (2009). The China puzzle: fallinghappinessinarisingeonomy. Journal of HappinessStudies,10:387{405.
moderatinginuene of labourmarketpoliiesaross Europe. Journal of Euro-pean Soial Poliy, 24(4):383{399.
Chudik, A. and Pesaran, M. H. (2015). Common orrelated eets estimation of heterogeneous dynami panel data models with weakly exogenous regressors. Journal of Eonometris, 188(2):393{420.
Clark, A. E. and D'Ambrosio, C. (2015). Attitudes to inome inequality: exper-imental and survey evidene. In Handbook of inome distribution, volume 2, hapter13, pages 1147{1208.Elsevier.
Clark, A.E., Flehe, S., Senik, C., etal.(2014). The greathappiness moderation. In Clark, A. and Senik, C., editors, Happiness and Eonomi Growth: Lessons from DevelopingCountries. OxfordUniversity Press, Oxford.
De Neve, J.-E., Ward, G., De Keulenaer, F., Van Landeghem, B., Kavetsos, G., and Norton, M. I. (2018). The asymmetriexperiene of positive and negative eonomi growth: Global evidene using subjetive well-being data. Review of Eonomis and Statistis, 100(2):362{375.
Deaton, A. (2008). Inome, health, and well-being around the world: Evidene fromtheGallupWorldPoll.TheJournalofEonomiPerspetives,22(2):53{72.
Di Tella, R., MaCulloh, R., and Oswald, A. (2003). The maroeonomis of happiness. The Review of Eonomis and Statistis, 85(4):809{827.
Easterlin,R.(1974). Doeseonomi growthimprovethe human lot? Some empir-ialevidene. In David, P. and Melvin, W., editors,Nations and households in eonomi growth, pages 98 {125. CA: Stanford University Press, PaloAlto.
Easterlin,R.(2009). Lostintransition: Lifesatisfationonthe roadtoapitalism. Journal of Eonomi Behavior & Organization,71(2):130{145.
Easterlin,R.(2017).Paradoxlost? ReviewofBehavioralEonomis,4(4):311{339.
Easterlin, R. and Angelesu, L. (2009). Happiness and growth the world over: Timeseries evideneon the happiness-inome paradox. IZA Disussion Paper, 4060.
The happiness-inome paradox revisited. Proeedings of the National Aademy of Sienes, 107(52):1 {6.
Easterlin, R. A., Morgan, R., Switek, M., and Wang, F. (2012). China's life satisfation, 1990{2010. Proeedings of the National Aademy of Sienes, 109(25):9775{9780.
Easterlin,R.A.,Wang, F., andWang,S.(2017). Growthand happinessinChina, 1990-2015. In Helliwell, J. F., Layard, R., and Sahs, J. D., editors, World HappinessReport 2017,pages 48{83. New York.
Engle,R.F. andGranger,C. W.(1987). Co-integrationanderrororretion: rep-resentation, estimation,and testing. Eonometria: journalof the Eonometri Soiety,(55):251{276.
European Commission (2018). Eurobarometer. Tehnial report, GESIS Data Arhive for the SoialSienes.
EVS (2011). European values study 1981-2008, longitudinal data le. Tehnial report, GESIS Data Arhive, Cologne. ZA4804 Data FileVersion 2.0.0.
Graham,C. and Felton,A. (2006). Inequality and happiness: insightsfrom Latin Ameria. Journal of Eonomi Inequality, 4(1):107{122.
Helliwell,J. (2008). Life satisfation and quality of development. Working Paper 14507,National Bureau of EonomiResearh.
Helliwell, J. F. (2003). How's life? Combining individual and national variables toexplain subjetivewell-being. Eonomi Modelling, 20(2):331{360.
Helliwell,J.F.andAknin,L.B.(2018).Expandingthesoialsieneofhappiness. Nature HumanBehaviour, (2):248{252.
Helliwell,J.F., Aknin,L.B., Shiplett,H.,Huang, H.,andWang,S.(2017). Soial apitaland prosoialbehaviour assoures of well-being. Working Paper23761, NationalBureau of Eonomi Researh.
Horioka, C. Y. and Kanda, R. (2010). Revitalizing the Japanese eonomy by soializingrisk. Japanese Eonomy, 37(3):3{36.
and SWIID. The Journal of Eonomi Inequality, 13(4):629{671.
Jiang, S., Lu, M., and Sato, H. (2012). Identity, inequality, and happiness: Evi-denefrom urbanChina. World Development,40(6):1190{1200.
Kahneman, D. and Krueger, A. (2006). Developments in the measurement of subjetive well-being. Journal of Eonomi Perspetives,20:3 { 24.
Layard, R. (2005). Happiness: lessons from a new siene. The Penguin Press, New York.
Mikuka, M., Sarraino, F., and Dubrow, J. K. (2017). When does eonomi growth improve life satisfation? Multilevel analysis of the roles of soial trust andinome inequality in46ountries, 1981{2012. World Development,93:447{ 459.
Ministry of Health, Labour and Welfare (2014). Graphial Review of Japanese Household. Ministry of Health,Labour and Welfare, Tokyo.
O'Connor, K. J. (2017). Happiness and welfare state poliy around the world. Review of Behavioral Eonomis,4(4):397{420.
OECD (2001). The evidene on soial apital. In The well-being of nations: the role of human and soial apital,pages 39 {63.OECD, Paris.
Oishi,S.and Kesebir,S.(2015). Inomeinequality explainswhyeonomi growth does not always translate to an inrease in happiness. Psyhologial siene, 26(10):1630{1638.
Oishi, S., Kesebir, S., and Diener, E. (2011). Inome inequality and happiness. Psyhologialsiene, 22(9):1095{1100.
Ono, H. (2010). Lifetime employment in Japan: Conepts and measurements. Journal of the Japanese and international eonomies, 24(1):1{27.
Ono,H. andLee, K.S.(2013). Welfarestatesand the redistributionof happiness. Soial Fores,92(2):789{814.
Ono, H. and Lee, K. S. (2016). Redistributing Happiness: How Soial Poliies Shape Life Satisfation. Praeger, Santa Barbara.
Press.
Rothstein, B.(2010). Happinessandthe welfarestate. Soial Researh: An Inter-national Quarterly, 77(2):441{468.
Saks, D. W., Stevenson, B., and Wolfers, J. (2012). Subjetive well-being, in-ome, eonomi development and growth. In Booth, P., editor, The pursuit of happiness, hapter 3, pages pp. 59 { 98. The Institute of Eonomi Aairs, London.
Sarraino, F. (2012). Money, soiability and happiness: Are developed oun-tries doomed to soial erosion and unhappiness? Soial Indiators Researh, 109(2):135{188.
Sarraino, F., O' Connor, K. J., and Ono, H. (2019). Making eonomi growth andwell-beingompatible: evidenefromJapan. Workingpaper, MPRApaper n.93010.
Shimmak, U., Krause, P., Wagner, G., and Shupp, J. (2010). Stability and hange of well-being: an experimentally enhaned latent state-trait-error anal-ysis. Soial Indiators Researh, 95(1):19{ 31.
Shmitt,H., Sholz,E., Leim,I., and Moshner,M. (2009). The Mannheim Euro-barometer Trend File Codebook and Unweighted Frequeny Distributions.
Sruggs, L., Detlef, J., and Kati, K. (2017). Comparative welfare entitlements dataset 2. Tehnial report, University of Connetiut & University of Greif-swald. Version 2017-09.
Solt,F.(2016).Thestandardizedworldinomeinequalitydatabase. SoialSiene Quarterly, 97(5):1267{1281. SWIID Version 7.0, July 2018.
Stevenson,B. andWolfers,J.(2008). Eonomigrowth andsubjetivewell-being: reassessingtheEasterlinParadox.BrookingsPapersonEonomiAtivity,(1):1 {87.
Uhlaner, C. (1989). Relationalgoodsand partiipation: inorporatingsoiability intoa theory of rationlation. Publi Choie,62:253 { 285.
UNU-WIDER (2018). World inomeinequality database.
uation speed and psyhologial well-being. Journal of Cognitive Neurosiene, 19(2):237{248.
Veenhoven, R. and Vergunst, F. (2013). The Easterlin illusion: eonomi growth doesgo with greaterhappiness. EHERO working paper 2013/1.
WorldBank(2018). Worlddevelopmentindiators. Tehnialreport,WorldBank. AessedMay 2018.
