Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Adaptation, assets, and aspiration. Three essays on the economics of subjective well-being
(USC Thesis Other)
Adaptation, assets, and aspiration. Three essays on the economics of subjective well-being
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
ADAPTATION, ASSETS, AND ASPIRATIONS.
THREE ESSAYS ON THE ECONOMICS OF SUBJECTIVE WELL-BEING.
by
Anke C. Zimmermann
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
May 2007
Copyright 2007 Anke C. Zimmermann
Acknowledgements
First of all I want to thank my advisor, Richard A. Easterlin for introducing me to a fascinating
topic and guiding my research throughout the years. I would also like to thank the members
of my qualifying committee and my dissertation committee, Timur Kuran, Kelly Musick, Jeffrey
Nugent, Merril Silverstein and Guofu Tan for helpful advice and guidance. I also received helpful
advice from John Ham and John Strauss and the participants of numerous seminars.
Laura Angelescu, Onnicha Sawangfa and Olga Shemyakina have been great officemates and
friends during the years.
Last but not least, Vincent Plagnol deserves many thanks for continuously increasing my own
subjective well-being, and his advice as a mathematician.
ii
Contents
Acknowledgements ii
List Of Tables v
List Of Figures vii
Abstract ix
1 The Economics of Subjective Well-Being: A Brief Overview 1
2 Happily Ever After? Cohabitation, Marriage, Divorce, and Happiness in Ger-
many 7
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Model, data, and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3 Financial Satisfaction over the Life Cycle: The Influence of Assets and Liabil-
ities 29
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Life cycle financial satisfaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Data and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 A model of financial satisfaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.5 Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4 Chasing the Good Life: Life Cycle Aspirations and Attainments 65
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Data and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3 Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.1 Material Goods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.1.1 Necessities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.1.2 Luxuries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.1.3 Lots of money . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.3.2 Family . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.2.1 A happy marriage . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.2.2 Children . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.3.3 Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
iii
4.3.4 Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.3.5 Relationofaspirationsandattainmentstodemographicvariablesotherthan
age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5 Epilogue: Policy Implications 85
Bibliography 90
Appendix A
Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
A.1 German Socio Economic Panel data . . . . . . . . . . . . . . . . . . . . . . . . . . 97
A.2 Cohabitation and Divorce . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Appendix B
Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
B.1 National Survey of Families and Households data . . . . . . . . . . . . . . . . . . . 101
B.2 NSFH Income Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
B.2.1 Income measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
B.2.2 Life cycle income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Appendix C
Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
C.1 Roper Survey data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
iv
List Of Tables
2.1 Characteristics of the German Socio-Economic Panel (GSOEP 1984-2004) popula-
tion and first marriage samples
a
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Hierarchicallinearmodelingregressionoflifesatisfactiononcohabitation,marriage,
divorce, and specified control variables (results with robust standard errors) . . . . 19
2.3 Fixed effects regression of life satisfaction on cohabitation, marriage, divorce, and
specified time-variant control variables . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Fixed effects regression of life satisfaction on cohabitation, marriage, divorce, and
specifiedtime-variantcontrolvariables,withinteractiontermsfordivorcesubgroup
(Div) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1 Answers of respondents who answered the question “My standard of living will get
much worse when I retire”, NSFH, wave 2 and 3 . . . . . . . . . . . . . . . . . . . 32
3.2 Descriptive statistics, full sample, NSFH, wave 2 (income, liabilities and assets in
$1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3 Descriptive statistics, waves 2 and 3, balanced panel (income, liabilities and assets
in $1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Household income by sample characteristics, NSFH wave 2 balanced sample (in-
come in $1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.5 Percentage of respondents who have certain types of debt . . . . . . . . . . . . . . 50
3.6 Weighted least squares regression on financial satisfaction, wave 2 (dollar amounts
in $1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.7 Ordered logit regression on financial satisfaction, wave 2 (dollar amounts in $1993) 58
3.8 Fixed-effects and ordered probit regressions on financial satisfaction . . . . . . . . 60
4.1 Descriptive Statistics: Mean percentage of people stating that an item is part of
the good life and mean percentage stating that they have the item, 1978− 2003
a
. 70
4.2 Descriptive Statistics, 1978− 2003 . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
v
4.3 Mean percentage of people stating that a happy marriage is part of the good life
by marital status, 1978− 2003 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.4 Regression coefficients on birth cohort and time . . . . . . . . . . . . . . . . . . . . 81
4.5 Regression coefficients on gender, race and education . . . . . . . . . . . . . . . . . 82
A.1 Prevalence of cohabitation and divorce in each sample period . . . . . . . . . . . . 100
B.1 Comparisonofmedianhouseholdincomes, CPSandNSFH.Householdincomesare
converted to index form with age 45-54 = 100.00 . . . . . . . . . . . . . . . . . . . 104
B.2 Comparison of mean household incomes, CPS and NSFH. Household incomes are
converted to index form with age 45-54 = 100.00 . . . . . . . . . . . . . . . . . . . 104
vi
List Of Figures
2.1 Prevalence of cohabitation in the sample of first marriages, GSOEP 1984-2004 . . 13
2.2 Added life satisfaction before and after marriage for persons in first marriages . . . 22
3.1 Lowess estimation: Financial satisfaction . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Lowess estimation: Log household income . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Financial Satisfaction in the U.S., General Social Surveys, 1972-2004, ages 18-89 . 33
3.4 Satisfaction with household income and standard of living in West Germany, Ger-
man Socio Economic Panel, 1991-2003, ages 18-90 . . . . . . . . . . . . . . . . . . 33
3.5 Financial satisfaction, NSFH, wave 2 and wave 3 . . . . . . . . . . . . . . . . . . . 36
3.6 Debt as an indicator of the discrepancy between financial aspirations and financial
means . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.7 Lowess estimation: Log financial assets . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.8 Lowess estimation: Log home value . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.9 Lowess estimation: Log credit card debt . . . . . . . . . . . . . . . . . . . . . . . . 50
3.10 Lowess estimation: Log debt on homes . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.11 Lowess estimation: Log other debt . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.12 Lowess estimation: Log net wealth . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.13 Lowess estimation: Children in household . . . . . . . . . . . . . . . . . . . . . . . 53
3.14 Lowess estimation: Other adults in household . . . . . . . . . . . . . . . . . . . . . 53
3.15 Lowess estimation: Self-rated health . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.1 Aspirations and attainments: A home you own, ages 19-78 . . . . . . . . . . . . . . 73
vii
4.2 Aspirations and attainments: A car, ages 19-78 . . . . . . . . . . . . . . . . . . . . 73
4.3 Aspirations and attainments: A swimming pool, ages 19-78 . . . . . . . . . . . . . 74
4.4 Aspirations and attainments: Travel abroad, ages 19-78 . . . . . . . . . . . . . . . 74
4.5 Aspirations and attainments: A vacation home, ages 19-78 . . . . . . . . . . . . . . 75
4.6 Aspirations and attainments: A lot of money, ages 19-78 . . . . . . . . . . . . . . . 75
4.7 Aspirations and attainments: A happy marriage, ages 19-78 . . . . . . . . . . . . . 77
4.8 Aspirations and attainments: One or more children, ages 19-78 . . . . . . . . . . . 78
4.9 Aspirations and attainments: Children’s college education, ages 19-78 . . . . . . . 79
4.10 Aspirations and attainments: An interesting job, ages 19-52 . . . . . . . . . . . . . 79
4.11 Aspirations and attainments: Health, ages 19-78, survey year 2003 . . . . . . . . . 80
viii
Abstract
Thethreeessaysinthisdissertationshareincommonanattempttostudytheeffectonsubjective
well-being of psychological processes along with objective conditions traditionally emphasized in
economics.
Do people adapt rapidly and completely to marriage as a recent study of German data con-
cludes? I analyze the role of adaptation – i.e whether people fairly quickly revert to their previous
level of well-being – following the formation of partnerships using the German Socio-Economic
Panel. On average, marriage has a lasting impact on life satisfaction, equal in magnitude to the
effect of cohabitation prior to marriage. The findings do not support the popular notion in psy-
chology that individuals always revert to a set level of happiness no matter how life circumstances
change.
Does financial satisfaction only depend on income? In the second essay, I explore the de-
terminants of financial satisfaction, including not only income but also the impact of assets and
liabilities. Financial satisfaction steadily increases over the life cycle, whereas household income
shows an inverted U-pattern with a peak at midlife. While income has the expected positive rela-
tion,increasingfinancialsatisfactionatolderagecanbepartlyexplainedbydecreasesinliabilities
and increases in financial assets. In addition, reduction in the dependency burden at old age leads
to increased financial satisfaction while the deterioration of health has a negative impact.
Is adaptation different with regard to economic and non-economic circumstances? In the
last essay I analyze differences in the life cycle patterns of aspirations and attainments in the
pecuniaryandnon-pecuniarydomains. Pecuniaryaspirations–i.e. aspirationsformaterialgoods
ix
– continue to increase over the life course, whereas nonpecuniary aspirations – i.e. aspirations for
family, work and health – remain constant or decline. The implication is that the steady increase
in pecuniary aspirations can undermine the pursuit of happiness. In contrast, aspirations with
regards to marriage do not increase; a finding which supports the results of the first essay that
marriage has a lasting effect on well-being. The empirical analysis is based on responses to Roper
surveys on the “good life”.
x
Chapter 1
The Economics of Subjective Well-Being: A Brief Overview
The economics of subjective well-being is the study of the nature and determinants of well-being,
as measured by subjective indicators. Most, if not all, people want to be happy and the pursuit
of happiness functions as an important motivational device for individuals. But why should
economists be concerned with the study of well-being?
1
Teachers in introductory undergraduate
classes in economics usually begin the first class with the question “What is economics?” Most
textbooksofferadefinitionwhichmoreorlessstatesthateconomicsisthestudyofhowindividuals
make decisions given limited resources to best satisfy their unlimited wants. Alternatively, one
canalsosaythatindividualswanttomaximizetheirutility givenscarceresources. Researchabout
the nature and determinants of this utility provides policy makers with valuable information, and
also allows the individual to evaluate his decisions and goals more accurately.
Utility–alsoreferredtoaspreferences,happiness,well-beingorsatisfaction–canbemeasured
in different ways. Economists often infer utility or preferences from the observable choices that
people make (revealed preferences), but this objective well-being approach relies heavily on the
assumption that people act rationally – an assumption which does not hold in many situations as
has been shown by numerous behavioral economists. For instance, individuals usually base their
decisions on the utility that they expect to derive from their choices, but this expected utility
1
The terms subjective well-being, happiness and satisfaction – though not necessarily the same – are often used
interchangeably in the literature.
1
often does not match the utility that individuals actually experience once they have made their
choice (Kahneman, 2000). Hence, if we assume bounded rationality, choices often do not reflect
true utility.
Another way to measure utility is to ask individuals directly how happy or satisfied they are
and thus rely on subjective measures of well-being. A typical subjective well-being question asks
the respondent to rate his satisfaction or happiness on a numerical scale. Subjective measures al-
low the respondent to assess his utility himself, and therefore take into account that the influence
of objective measures, such as income or employment status, is often mediated by psychological
processes, such as social comparison and hedonic adaptation. Social comparison is a psycholog-
ical process through which people evaluate themselves in comparison to the standard of others
in their peer group (also called reference group, frame of reference, norms, or standards; see, for
example, Duesenberry, 1949; Pollak, 1970; Hirsch, 1976; Frank, 1985, 1997). Hedonic adaptation,
also known as habituation, involves a different form of comparison; in this concept the individ-
ual’s standard for comparison is his own past (e.g. Modigliani, 1949; Pollak, 1976; Frederick and
Loewenstein, 1999). Both social comparison and hedonic adaptation mediate the influence of
objective indicators on well-being. For instance, in the case of income the initial positive effect
of an increase in income on well-being diminishes often quickly due to hedonic adaptation – the
individual habituates to the new level of consumption choices and experiences increased material
aspirations subsequently (see chapter 4). Moreover, if the individual moves to a more affluent
neighborhood after an increase in income, his frame of reference will shift upwards because he is
suddenly surrounded by individuals with higher incomes (see Luttmer, 2005, for a study on the
effect of neighbors’ incomes on well-being).
A considerable advantage of subjective well-being measures is that the respondent himself
evaluates which aspects are important for his well-being. Some objective measures, e.g. the gross
domestic product (GDP), are not truly objective because at some point someone had to decide
whatthismeasureshouldencompass(Sirgyetal.,2006). TheuseofindicatorssuchastheGDPas
2
measures of well-being is based on the assumption that these measures indeed contain the factors
that are important for one’s well-being. Indicators of subjective well-being allow the respondent
to assess his well-being himself, but research using these measures is only meaningful if people’s
concept of well-being is similar. In fact, individuals in different nations tend to list the same
factors when asked about their concerns, thus indicating that people’s concept of well-being does
notvarysubstantiallybetweenindividualsornations(Cantril,1965). Asimilarconcerniswhether
we can compare scores between people. Does a satisfaction level of 4 have the same meaning for
me as it has for my neighbor? The possibility that people rate their well-being differently is a
valid concern, but studies usually compare large groups. For example, we might be interested
in the question wether the unemployed are less happy than the employed. A potential bias due
to different satisfaction scales used by respondents would most likely cancel out between groups,
unless one argues that the unemployed systematically employ a different scale for rating their
well-being than the employed.
Subjective well-being measures are often dismissed by economists on the grounds that they
might not adequately measure true utility. There is a large body of literature concerning this
issue in psychology and sociology. Two recent articles in the Journal of Economic Perspectives
provide an overview of the field and also address the validity and reliability of these measures
(Kahneman and Krueger, 2006; DiTella and MacCulloch, 2006). Concerns about validity refer
to the questions whether measures of subjective well-being indeed measure true utility or merely
reflect momentary mood. Numerous studies have shown strong correlations between self-reports
on well-being and objective indicators of well-being, such as Duchenne smiles – i.e. smiles that
cannot be faked (Ekman et al., 1988). Happiness scores are also correlated with left frontal brain
activity which is said to be connected to true utility, with the left frontal brain being the center of
pleasure and approval (e.g. Davidson and Fox, 1982). Happiness scores also correlate highly with
other measures such as the depression scale in psychology. The consensus among researchers is
3
thatdespitesomenoiseinthedataduetotemporarymood,happinesscanbemeasuredsufficiently
well (DiTella and MacCulloch, 2006).
The other concern refers to the reliability of the measures – i.e. would the same level of
happiness be measured by the same score on two different occasions? Research has shown that
relatively constant variables such as education and income have a test-retest correlation of 0.9.
Composite scores of life satisfaction have a test-retest correlation of 0.77 (Lucas et al., 1996).
Thus, happiness data are quite reliable although there seems to be some influence from mood and
context.
The three essays in this dissertation share in common an attempt to study the effect on
subjective well-being of psychological processes along with objective conditions traditionally em-
phasized in economics. The main questions to be addressed concern topics of considerable debate
in the subjective well-being literature, including the question whether changes in objective life
circumstances can lead to lasting changes in satisfaction.
One of the main goals of this dissertation is to examine the determinants of subjective well-
beingindifferentdomainsoflife. ThedomainsatisfactionapproachpioneeredbyAngusCampbell
(Campbell et al., 1976; Campbell, 1981) assumes that overall life satisfaction is determined by
an individual’s satisfaction with different domains of his life, such as the family, income, work,
and health domains. Satisfaction in each domain depends to a large part on the extent to which
aspirations are met.
Do positive changes in the family domain lead to lasting improvements in subjective well-
being? My analysis indicates that in the case of marriage individuals benefit from forming a
maritalunioninthelong-term(Chapter2). Improvementsinwell-beingarenotmerelytemporary,
as has been previously suggested (Lucas et al., 2003), but last for a considerable time period after
the wedding day. Indeed, most people desire to have a happy marriage, and declining aspirations
over the life cycle for having a happy marriage suggest that improvements in this domain would
lead to lasting increases in well-being (Chapter 4). An analysis of aspirations and attainments in
4
the family domain thus confirms the results of the study on the enduring benefits of marriage.
Some changes in life circumstances – at least in the family domain – therefore seem to lead to
lasting changes in well-being. This result does not confirm the popular notion in psychology that
individuals have a genetically determined setpoint level of well-being to which they return after
temporary deflections from this setpoint due to changes in life circumstances.
Thesecondessayanalyzesthedeterminantsoffinancialsatisfaction–adomainofconsiderable
importance for overall life satisfaction. The same essay also contributes to the understanding of
one of the topics of great interest in subjective well-being research in economics – the influence
of income on well-being. Since Easterlin’s seminal 1974 article (Easterlin, 1974), which is com-
monly accepted as the research paper that initiated the field of subjective well-being research in
economics, there has been a considerable amount of research on income and happiness and in
particular on the influence of psychological processes such as social comparison and habituation.
The impact of income on happiness seems to be largely influenced by relative standards.
People evaluate their own financial situation based on comparisons with others or comparisons
with what they owned previously. The influence of income on overall well-being is mediated by
its impact on financial satisfaction – people often do not evaluate their income rationally and it is
therefore important to account for possible psychological processes influencing their perception of
their financial situation. My findings suggest that income has a considerable impact on financial
satisfaction, but assets and debts as well as financial obligations explain the life cycle profile of
financial satisfaction better than income alone. My research on financial satisfaction contributes
to the ongoing debate on income and well-being because it points out the importance of perceived
needs on satisfaction.
The last essays considers changes in aspirations and attainments over the life cycle and largely
confirmstheresultsofthefirsttwoessays. Incontrasttotheeconomicdomain, aspirationsdonot
increase with age in the nonpecuniary domains, which includes the family domain. Improvements
5
inthenonpecuniarydomainscanthereforeleadtoincreasedsubjectivewell-beingwhereasgreater
attainments in the pecuniary domain are undermined by increasing aspirations.
All three essays thus address questions that are currently of great interest in the subjective
well-beingliterature. Theycontributetotheunderstandingofdomainsatisfactionandthecurrent
debates on the influence of income on well-being, as well as the questions whether people fully
adapt to changes in life circumstances.
6
Chapter 2
Happily Ever After? Cohabitation, Marriage, Divorce, and
Happiness in Germany
A condensed version of this chapter has been published in Population and Development Review,
32(3), pp. 511-528, September 2006, co-authored with Richard A. Easterlin.
2.1 Introduction
One of the domains that has been shown to have a large impact on subjective well-being is the
family domain. The family domain encompasses relationships with partners, children and other
family members. In this chapter I will mostly concentrate on the effect of marriage on well-being
and only briefly mention the effect of children. In particular, I want to analyze whether gains in
well-being after a wedding are only temporary or indeed enduring as expected by everyone who
decides to formalize a union by marriage.
There is a comfortable consensus in the social sciences that marriage has a positive and en-
during effect on well-being (for references in sociology and demography, see Waite, 1995; Waite
and Lehrer, 2003, in economics, Frey and Stutzer, 2002; Layard, 2005). A frequent criticism of
some of these studies is that they are based on cross-sectional data and therefore do not allow any
conclusionsaboutcausality. Ajarringchallengetothisconsensus, however, wasrecentlyproposed
7
in an award-winning article in a leading psychological journal by a group of two psychologists and
two economists (Lucas et al., 2003). In a German panel study covering 15 years they find that
there is a temporary positive “honeymoon period” effect of marriage, but typically people revert
two years after marriage to the same “baseline” level of life satisfaction that prevailed two years
before. Thepsychologists’conclusionisthat, “onaverage, peopleadaptquicklyandcompletelyto
marriage” (p. 536). “Adaptation” here means, not that one adjusts to difficulties encountered in
living with a partner, but that the hedonic gains from forming a union are transient and quickly
disappear. The authors further conclude that the common observation in cross-sectional studies
that married individuals have higher levels of well-being than their single counterparts can be
explained by the high prevalence of newly married couples who are still in their “honeymoon
period”.
The significance of this conclusion goes beyond the issue of whether marriage has lasting
benefits. The “setpoint theory” in psychology sees individuals as adapting fully to all kinds of life
circumstances – job promotion, serious accident, death of a partner, and so on (Kammann, 1983;
Lykken and Tellegen, 1996; Myers, 1992, 2000). Lucas and colleagues are testing the setpoint
model in the family domain. In this theory a person’s subjective well-being tends to center
around a setpoint determined by genetics and personality, and major life transitions and events
merely deflect a person temporarily from this level. David G. Myers (2000:60), a proponent of
setpoint theory, quotes favorably the view expressed by the late Richard Kammann (1983:18):
“Objective life circumstances have a negligible role to play in a theory of happiness.” When
Lucas and colleagues state that “people adapt quickly and completely to marriage”, they mean
that the partners in a marital union fairly quickly return to the happiness level dictated by their
personalitytraitsandgeneticheritage. Adisturbingimplicationofthesetpointmodelisthatlittle
can be done by personal action or public policy to improve individual well-being. Ed Diener and
RichardE.Lucas, twooftheauthorsofthepanelstudy, arequiteexplicitaboutthis. Inanearlier
article they state: “The influence of genetics and personality suggests a limit on the degree to
8
which policy can increase SWB [subjective well-being].... Changes in the environment, although
important for short-term well-being, lose salience over time through processes of adaptation,
and have small effects on long-term SWB” (Diener and Lucas, 1999: 227)
1
. Clearly if, in the
population as a whole, adaptation to life circumstances is typically rapid and complete, then any
measure taken to improve average well-being is fruitless (cf. Easterlin 2003).
The present analysis provides new evidence on the validity of setpoint theory with regards to
marriage. I analyze the same data set as used in the 2003 study by Lucas and colleagues (the
German Socio-Economic Panel), but cover 21 waves (1984-2004) compared with their 15 waves
(1984-1998).
2
The sample is of first marriages among previously unmarried persons who married
during the survey period and for whom data are available for at least two years before marriage
(to establish the premarriage baseline of life satisfaction) and two years after marriage (to test
whether there is a return to baseline satisfaction after the“honeymoon period”). The sample
includes marriagesthatremainintact duringthe sample periodas wellasthose thatdissolve after
two or more years. (Very few marriages dissolve within the first two years.) One might suppose
that broken marriages would be characterized by a baseline-to-postmarriage trajectory that is
significantly different from that of intact marriages, and that a study confined to “successful”
marriages – those still intact at the last date surveyed – would give too favorable a picture of
the effects of marriage. I also examine the effect on subjective well-being of the formation of
cohabiting unions before marriage, and I take account of many of the ways in which a sample
population of individuals who enter first marriages in the survey period differs from the German
Socio-Economic Panel population generally.
1
A gradual retreat by these authors from the view expressed in this 1999 quotation is apparent in later work. In
the 2003 article cited here, Lucas and colleagues find that adaptation to widowhood takes eight years. Elsewhere,
they conclude that unemployment has a lasting effect on well-being, altering the happiness setpoint (Lucas et al.,
2004). Lucas (2005) finds that divorce too reduces life satisfaction, a result seemingly at odds with the finding on
marriage (Lucas et al., 2003), which Lucas continues to defend (Lucas and Clark, 2006). A proposal by Diener
and Seligman (2004) for governmental measurement of subjective well-being suggests numerous ways in which
socioeconomic policy might improve well-being.
2
The data used here were made available to me by the German Socioeconomic Panel Study at the German
Institute for Economic Research (DIW), Berlin.
9
Most studies of the effects of marriage and divorce use objective indicators of well-being,
but subjective measures are gradually finding acceptance in demographic research (Kohler et al.,
2005; Bernhardt and Fratczak, 2005; Horwitz et al., 1996; Marks and Lambert, 1998; Simon,
2002). These measures play a central role in Waite’s collaborative work on the well-being effects
of dissolution of marital unions (Waite and Luo, 2003; Waite et al., 2002).
Previous analyses of the effects of marriage are typically point-of-time studies of the relation-
shipofsubjectivewell-beingtomaritalstatus,withcontrolsforsuchfactorsasincome,health,and
work status. The repeated conclusion of these cross-sectional studies is that being married has a
positiveimpactonlifesatisfaction,whilebeingdivorcedorseparatedhasanegativeeffect(Argyle,
1999; Blanchflower and Oswald, 2004b; Frey and Stutzer, 2002; Stutzer and Frey., 2006). The few
panel studies other than that of Lucas and colleagues have also usually supported the consensus
on the positive effect of marriage (Johnson and Wu, 2002; Mastekaasa, 1995). However, these
panel studies, like those of Waite and her collaborators, have not included a premarriage baseline
period – an important innovation of the Lucas panel study – and therefore do not address the
issue of whether well-being in the postmarriage period returns to the premarriage baseline level.
Theusualexplanationofthebenefitsofmarriage–whethermeasuredwithobjectiveorsubjective
indicators – is in terms of “social support”, that is, the beneficial effects of companionship, emo-
tional support, sustained sexual intimacy, and so on (Blanchflower and Oswald, 2004a; Coombs,
1991; Johnson and Wu, 2002; Laumann et al., 1994; Powdthavee, 2005; Waite and Joyner, 2001).
There is also recognition that in cross-sectional studies of the relationship of well-being to marital
status some selection effect may be at work – that persons, say, with “happier” personalities are
more likely to marry – but such effects are typically considered to be unable to account for much
of the positive association. In contrast to the prevailing view, Lucas and colleagues argue that
their results reject the “social support” or “social role” hypothesis. Rather, they believe that
the positive relationship of marriage to subjective well-being in the cross-section is attributable
10
to selection into marriage on the basis of personality traits, and that cross-section surveys are
capturing some recently married individuals who are still in the temporary “honeymoon”‘ period.
Most studies on marriage in economics discuss the reasons for couples to enter a marital union
and emphasize the pecuniary benefits of marriage. Becker (1973, 1974) postulates a theory of
marriage which assumes that individuals decide to marry based on expected gains from marriage.
In this views both partners search for a mate who will maximize their well-being. Becker (1973)
first proposes a model in which the gains from marriage are mostly pecuniary and depend on the
spouses’ incomes, human capital and wage differential. The current analysis concerns the realized
gains from marriage instead of the motivation for entering a marital union. I also concentrate on
the nonpecuniary benefits of marriage by introducing a control for household income in my model
to account for increased household incomes.
In what follows I describe the model, data, and methodology, and report my findings. I
conclude that the German Socio-Economic Panel data support the conclusion of cross-sectional
analyses that the formation of unions – marital or cohabiting – increases happiness and that the
dissolution of unions decreases it. Subjective well-being two or more years after marriage is not
statistically different from that experienced during cohabitation prior to marriage. It seems that
a union improves well-being regardless of whether it is formalized or not.
2.2 Model, data, and methods
My model consists of an intercept and four terms, which describe different life stages for an
individualwhomarriesduringthesampleperiod. Theinterceptreflectstheaveragelifesatisfaction
of individuals in the sample in the “baseline” period – all noncohabiting years that are at least
one year before marriage (t
1
and before). The first term is a cohabitation term, and it measures,
for those who form a cohabiting union prior to marriage, the average difference in life satisfaction
from one’s baseline value arising from participation in a cohabiting union. The second term, a
11
marriage “reaction” term, measures the average difference in life satisfaction in the first year of
marriage (t
0
) and the year immediately following (t
+1
) from one’s baseline value. A marriage
“adaptation” term measures the average difference in life satisfaction from one’s baseline value in
thesecondyearaftermarriageandallsubsequentyearsofmarriage(t
+2
andthereafter). Boththe
reaction and the adaptation terms are included instead of only one term for “marriage”, because
I want to test whether there is a “honeymoon” effect, that is, whether individuals experience
significantly higher levels of life satisfaction during the year of marriage and the year immediately
following. A single measure for the period of marriage could be biased upwards by elevated levels
of well-being in the early years of marriage. Moreover, it would not allow me to asses whether
well-being declines to its pre-marriage level after only a short period. The final term, a “divorce”
term,measuresforthosewhodivorceorseparateaftertwoormoreyearsofmarriagethedifference
in life satisfaction from one’s baseline value. The model is structured so that I can test for two
key results of the analysis by Lucas and colleagues. The first is whether, two or more years after
marriage, individuals who are still married revert to the baseline level of satisfaction that existed
before marriage. The second is whether a significant increase in life satisfaction occurs around
the time of marriage – a “honeymoon period”. But the present model is considerably broader,
encompassingthelifesatisfactioneffectsoftheformationanddissolutionofunionsmoregenerally.
Thus, it includes a term reflecting the effect on life satisfaction of cohabitation before marriage.
The evidence is considerable that the formation of a cohabitating union has a positive impact
on life satisfaction similar to that of marriage, although the magnitude of the effect is sometimes
not as great (Stack and Eshleman, 1998; Frey and Stutzer, 2002). Because cohabitation is fairly
prevalent among young Germans, it is possible that Lucas et al.’s estimates of life satisfaction
in the baseline period and the year before marriage include a considerable part of the impact of
the benefit of forming unions – thus elevating baseline satisfaction. An analysis which considers
the effects of marriage on well-being should not compare satisfaction during marriage with that
experienced during cohabitation. For most couples, the only difference between these two types
12
of unions lies in the formalization of the union and one should therefore not expect significant
differences in well-being. Setpoint theory refers to a level of well-being determined by personality
and genetics – i.e. the well-being experienced before a change in life circumstances, such as
cohabitation, occurred.
The German Socio-Economic Panel contains questions (given here in Appendix A.1) which,
though varying slightly over time, permit me to examine the extent of premarital cohabitation
in the sample of first marriages. In the year before marriage, 67 percent of respondents were
cohabiting. Among the sample observations two or more years prior to marriage, 29 percent are
for persons who were cohabiting. Figure 2.1 gives an overview of the prevalence of premarital
cohabitation in the sample.
Figure 2.1: Prevalence of cohabitation in the sample of first marriages, GSOEP 1984-2004
Although I cannot identify partners, it seems likely that a large fraction of the cohabiting ob-
servationsareofunionswithone’seventualmarriagepartner,becausesomanyoftheobservations
are for unions in the year preceding marriage. Indeed, most cohabiting couples have the intention
to eventually marry their current partner. In a study by Brown and Booth (1996) three quarters
13
of all cohabitors had plans to get married. The authors further find that cohabitors usually re-
port poorer relationship quality than married couples (also reported by Nock 1995; DeMaris and
Leslie 1984). However, cohabitors with plans to marry do not differ significantly in terms of rela-
tionship quality from their married counterparts. The present analysis only includes partners in
cohabiting unions who will marry during the sample period. An analysis of exclusively cohabiting
unions which do not lead to marriage would be of considerable interest given the similarity of the
benefits of cohabiting and married relationships. Brown (2004) examines whether marriage leads
to an increase in the quality of relationships by comparing cohabitors who marry to cohabitors
who remain in a cohabiting relationship. She finds that cohabitors who marry report higher re-
lationship quality but this finding does not seem to be related to the effects of marriage per se
but to reported commitment to marriage. Cohabitors who remained cohabiting but with plans
to get married were similar to married cohabitors in terms of relationship quality. An analysis
of adaptation to long-term cohabiting relationships would require several years of data for each
cohabiting individual, but most cohabiting relationships end after a few years due to marriage or
the dissolution of the union. The present data do not provide me with a sufficiently large sample
for such an analysis.
My model also includes a term to test for the impact of marital dissolution on life satisfaction.
To see whether life satisfaction follows the same initial course during marriage for those who
eventually divorce or separate as for those who do not, I focus on individuals who divorce or
separate after two or more years of marriage (n = 151) by estimating the effect of a “divorce
group” dummy variable on the intercept and slopes. In this way I obtain comparable estimates
for cohabitation, reaction, and adaptation terms for marriages that remain intact throughout the
survey period and those that break up. Some of the marriages that I call “intact” will, of course,
eventually dissolve. Excluded from the analysis are the small number of first marriages ending in
divorceinthefirsttwoyearsaftermarriage(n=2), marriagesdissolvedbydeathofaspouse(n=
5),andfirstmarriagesofforeign-bornpersonswhosespouseislivinginadifferentcountry(n=6).
14
For brevity of presentation, I refer below to the group who do not remain married as the “divorce
subgroup”, although in my sample of 151 individuals experiencing marriage breakup, only 100
are actually divorced while 51 are separated. Of the 100 divorced, 75 were separated for one or
more years before divorce and 25 divorced without first being separated. I want to emphasize
that these divorces occur relatively shortly after marriage and thus represent only 10% of the
marriages in my sample of first marriages (Table A.1). The divorce rate is significantly higher in
Germany, reaching 43.6% in West Germany and 37.1% in East Germany in 2003 (Bundeszentrale
f¨ ur politische Bildung, 2004). The results for the divorce subgroup have to be taken with a grain
a salt because of the small sample size. Table A.1 in appendix A.2 shows the prevalence of
cohabitation and divorce in each sample period.
The model also takes account of the distinctive socioeconomic characteristics of the first mar-
riage sample. Not surprisingly, this group is younger than the sample population as a whole, by
14 years on average (Table 2.1, cols. 1 and 2). Younger people are more likely to be better edu-
cated, employed, and healthier than average, and this is true for my sample. The first marriage
sample is considerably higher too on religiosity. The divorce subgroup differs somewhat from the
first marriage sample as a whole in having a larger proportion of females and persons of lower
socioeconomic status (col. 3).
Other research has found that life satisfaction tends to vary significantly by sex, age, income,
education, health, employment, and religiosity (Argyle, 1999; Blanchflower and Oswald, 2004b;
Frey and Stutzer, 2002). To assess the specific impact on life satisfaction of the formation and
dissolution of unions, I include controls for these characteristics in the model (except for health
because questions on health status were not asked before 1992). Also, to examine whether the
presence of children affects life satisfaction, I include a variable for the number of children in the
household in the year of marriage and thereafter. (The divorce subgroup and those who remained
marriedhavevirtuallythesamenumberofchildreninthesecondyearaftermarriage.) Acomplete
description of the variables is given in Appendix A.1.
15
Table 2.1: Characteristics of the German Socio-Economic Panel (GSOEP 1984-2004) population
and first marriage samples
a
GSOEP First marriage Divorce
Line Sample size or variable population sample subgroup
1 Number of persons 37,244 1,582 151
2 Mean age, years 42.8 29.0 30.1
3 Education greater than high school, percent 19.0 27.1 21.2
4 Employed, percent
b
68.7 98.5 98.0
5 Mean health (1 = low, to 5)
c
3.47 3.77 3.67
6 Religiosity, percent 33.7 49.6 45.7
7 Household income ine1995 29,729 29,195 25,619
8 Male, percent 49.0 50.3 43.7
9 Children in t+2 – 0.82 0.85
Note: – = not applicable.
a. See Appendix A.1 for description of variables.
b. Percentage of respondents who were employed at least once during the sample period.
c. Sample size is smaller than in line 1.
As noted earlier, the data are from waves 1-21 of the German Socio-Economic Panel, covering
the years 1984-2004 (Haisken-DeNew and Frick, 2005). To my knowledge this is the longest-
running panel study to include a measure of subjective well-being. The specific question asked
is: “How satisfied are you with your life, all things considered?” Responses are ranked on a scale
from0(completelydissatisfied)to10(completelysatisfied). FollowingLucasandcolleagues’2003
study,Icenterlifesatisfactionscoresaroundtheannualmeanofeachpopulationsubsampleinthe
originalGermanSocio-EconomicPanelpopulation. Thiscenteringtechniquewaschosentoadjust
for significant differences in life satisfaction between the subsamples as well as for time trends in
life satisfaction. For instance, East Germans were significantly less satisfied than West Germans
shortly after unification (Easterlin and Zimmermann, 2007). The centering procedure eliminates
these effects.
3
The regression analysis employs hierarchical linear modeling, generally considered
to be the statistical technique most appropriate for analysis of panel data (Luke, 2004; van der
3
It is also common to center the dependent variable around the grand mean, although this strategy would not
account for annual trends or group differences. I reran the analysis using grand mean centering for the dependent
variable and found essentially the same results.
16
Leeden, 1998). This model is also referred to as a random coefficients model in the econometric
literature. The formal model is of the following form: Level 1: (within-subject)
(Life satisfaction)
it
= β 0i
+β 1i
(cohabitation)
it
+β 2i
(reaction)
it
+β 3i
(adaptation)
it
(2.1)
+ β 4i
(divorced)
it
+β 5i
(employed)
it
+β 6i
(age)
it
+β 7i
(income)
it
+r
it
Level 2: (between-subjects)
β 0i
= γ 00
+γ 01
(divorce group)
i
+γ 02
(male)
i
+γ 03
(education)
i
(2.2)
+ γ 04
(religiosity)
i
+u
0i
β 1i
= γ 10
+γ 11
(divorce group)
i
+u
1i
β 2i
= γ 20
+γ 21
(divorce group)
i
+u
2i
β 3i
= γ 30
+γ 31
(divorce group)
i
+u
3i
β 4i
= γ 40
+γ 41
(education)
i
+u
4i
β 5i
= γ 50
+γ 51
(divorce group)
i
+γ 52
(male)
i
+γ 53
(education)
i
β 6i
= γ 60
+γ 61
(divorce group)
i
β 7i
= γ 70
+γ 71
(divorce group)
i
The intercept, cohabitation, reaction, adaptation and divorced variables are entered as random
variables – i.e. they are allowed to differ between individuals. The u
i
error terms capture the
unique effect of each individual on the slopes and intercept. I test the difference in trajectories
for those who remain married and those who will divorce by including a dummy variable in the
second level of the hierarchical model that indicates whether an individual belongs to the divorce
subgroup. Theimpactofbelongingtothegroupofpeoplewhowilleventuallydivorceisestimated
for each slope, except for the “divorced period” slope, which can only be estimated for people
17
who divorce. Sex, religiosity and to some extent education are time-invariant characteristics and
thus are entered as second-level variables. I found little evidence of a different trajectory by sex.
Being male had a significant effect on the intercept and the employment slope, but not on the
remaining slopes (Table 2.2). An individual’s employment status, income, and age vary over time
and might alter life satisfaction independent of one’s marital situation. I therefore include these
time-variant covariates in the first level of the hierarchical model. Age and income are grandmean
centered; hence the intercept value reflects the life satisfaction of a person of mean age with mean
income in the marriage sample. The software used is HLM 6 (Raudenbush et al., 2000), which
uses an iterative maximum likelihood procedure to estimate the model. The detailed regression
results are presented in table 2.2.
The main results here are based on the hierarchical linear model because this is also the
model of choice by Lucas et al. (2003) and I want to ensure that my results are comparable to
theirs. Fixed-effectsregressionanalysis,thepreferredmethodofmanyeconomists,yieldsthesame
results as hierarchical linear modeling if the same time-variant covariants are included (the fixed-
effects framework accounts for time-invariant characteristics). The model above does not fulfill
the standard ordinary least squares (OLS) assumptions of random errors that are independent,
normally distributed and have constant variance. In fact, the error structure of the hierarchical
linear model is rather complex. A simplified model of the form
(Life Satisfaction)
it
= β 0i
+β 1i
(cohab)
it
+r
it
(2.3)
and
β 0i
= γ 00
+γ 01
(male)
i
+u
0i
(2.4)
β 1i
= γ 10
+γ 11
(male)
i
+u
1i
18
can be combined as:
(Life Satisfaction)
it
= γ 00
+γ 01
(male)
i
+γ 10
(cohab)
it
+γ 11
(male)
i
(cohab)
it
(2.5)
+u
0i
+u
1i
(cohab)
it
+r
it
Table 2.2: Hierarchical linear modeling regression of life satisfaction on cohabitation, marriage,
divorce, and specified control variables (results with robust standard errors)
Standard
Model term Coefficient error t-ratio p-value
Intercept β 0i
intercept γ 00 0.102 0.063 1.612 0.107
divorce group γ 01 -0.480 0.176 -2.724 0.007
male γ 02 -0.339 0.077 -4.379 0.000
education γ 03 0.318 0.058 5.512 0.000
religiosity γ 04 0.187 0.048 3.891 0.000
Cohabitation β 1i
intercept γ 10 0.183 0.040 4.627 0.000
divorce group γ 11 -0.031 0.140 -0.219 0.827
Marriage reaction period β 2i
a
intercept γ 20 0.369 0.043 8.651 0.000
divorce group γ 21 0.050 0.141 0.358 0.720
Marriage adaptation period β 3i
b
intercept γ 30 0.173 0.051 3.382 0.001
divorce group γ 31 -0.095 0.169 -0.566 0.571
Divorce period β 4i
c
intercept γ 40 -0.286 0.205 -1.395 0.163
education group γ 41 0.422 0.163 2.595 0.010
Time-variant covariates
Employed β 5i
intercept γ 50 0.077 0.039 1.942 0.052
divorce group γ 51 0.234 0.110 2.130 0.033
male γ 52 0.333 0.071 4.692 0.000
education γ 53 -0.179 0.053 -3.416 0.001
Age β 6i
intercept γ 60 -0.017 0.004 -4.300 0.000
divorce group γ 61 0.004 0.012 0.311 0.756
Household income β 7i
intercept γ 70 4 x 10
− 6
1 x 10
− 6
4.481 0.000
divorce group γ 71 3 x 10
− 6
4 x 10
− 6
0.718 0.473
n 1,568
a. Year of marriage and following year.
b. Second year after marriage and thereafter.
c. Third year after marriage and thereafter.
19
One can quickly see that the errors in equation 2.5 are not independent, but contain com-
ponents u
0i
and u
1i
that are common for every time period nested within an individual. The
results in the present analysis are not sensitive to the choice of model. A fixed-effects model,
which fulfills the standard OLS assumption of independent errors, yields the same results. One
advantage of the fixed-effects model is that it controls for all unobserved individual fixed effects.
This feature of the model can also be seen as a disadvantage because one cannot see the effect of
specific time-invariant controls on the slopes and intercept.
Table 2.3: Fixed effects regression of life satisfaction on cohabitation, marriage, divorce, and
specified time-variant control variables
Standard
Model term Coefficient error t-ratio p-value
Intercept 0.056 0.032 1.76 0.079
Cohabitation 0.143 0.033 4.28 0.000
Marriage reaction period
a
0.350 0.035 10.02 0.000
Marriage adaptation period
b
0.123 0.041 3.02 0.003
Divorce period
c
-0.242 0.081 -3.00 0.003
Time-variant covariates
Employed 0.194 0.024 8.00 0.000
Age -0.012 0.003 -3.87 0.000
Household income 5 x 10
− 6
8 x 10
− 6
6.52 0.000
n of observations 23,632
n of groups 1,582
R
2
within 0.0134
R
2
between 0.0532
R
2
overall 0.0250
a. Year of marriage and following year.
b. Second year after marriage and thereafter.
c. Third year after marriage and thereafter.
I also included interaction terms in the fixed-effects regression to test whether the model
coefficients differ between individuals who remain married and those who divorce during the
sampleperiod, andIfoundessentiallythesameresultsaswithhierarchicallinearmodeling(Table
2.4).
20
Table 2.4: Fixed effects regression of life satisfaction on cohabitation, marriage, divorce, and
specified time-variant control variables, with interaction terms for divorce subgroup (Div)
Standard
Model term Coefficient error t-ratio p-value
Intercept 0.056 0.032 1.77 0.077
Cohabitation 0.165 0.035 4.69 0.000
Cohabitation x Div -0.202 0.114 -1.78 0.076
Marriage reaction period
a
0.364 0.037 9.91 0.000
Marriage reaction period x Div -0.118 0.122 -0.96 0.335
Marriage adaptation period
b
0.152 0.043 3.54 0.000
Marriage adaptation period x Div -0.288 0.137 -2.10 0.036
Divorce period
c
-0.508 0.177 -2.86 0.004
Time-variant covariates
Employed 0.168 0.026 6.53 0.000
Employed x div 0.257 0.080 3.23 0.001
Age -0.013 0.003 -4.13 0.000
Age x Div 0.013 0.012 1.08 0.279
Household income 5 x 10
− 6
8 x 10
− 7
6.42 0.000
Household income x Div -2 x 10
− 6
3 x 10
− 6
-0.56 0.574
n of observations 23,632
n of groups 1,582
R
2
within 0.0142
R
2
between 0.0512
R
2
overall 0.0252
a. Year of marriage and following year.
b. Second year after marriage and thereafter.
c. Third year after marriage and thereafter.
2.3 Findings
In this section I concentrate on the findings from the hierarchical linear model. The average
course of life satisfaction for individuals in intact marriages, who account for over 90 percent of
all first marriages in my sample, is shown schematically in figure 2.2. The effects found here are
comparable in magnitude to those of cross-sectional studies. The baseline value (0.10) of those in
first marriage is not significantly different from the value for the German panel population as a
whole. In the absence of controls for special socioeconomic characteristics, this group would rank
significantly higher on life satisfaction than the general population, but socioeconomic controls
21
eliminate this disparity. If it were possible to control for the greater health of the first marriage
group, then the disparity would doubtless be reduced even further.
4
Figure 2.2: Added life satisfaction before and after marriage for persons in first marriages
*significant at 0.001 level or better.
Note: The slope coefficients are usually added to the coefficient of the intercept. The value for the intercept
(baseline) is not statistically different from 0 here. The figure thus does not show slope coefficients that are added
to 0.102 (baseline), but added to 0.
This result for the baseline value runs counter to the idea that those who marry are distinctive
with regard to those personality traits that make one happier and are likely to attract a marriage
partner (the traits chiefly suggested in the literature are high extroversion and low neuroticism;
see Diener and Lucas 1999). If the first marriage group were selective in regard to such traits,
then their baseline life satisfaction value after socioeconomic controls would remain significantly
higher than that for the population as a whole, reflecting the favorable impact of these traits
on life satisfaction. But controls for the distinctive socioeconomic characteristics of those in first
marriages put them in essentially the same baseline situation as the general German population
andleavenoroomfortheinferencethatthefirstmarriagegrouphasdistinctivepersonalitytraits.
4
Stutzer and Frey (2006: Figure 1) compare singles who will marry with those who will not and find selection
effects for those who marry at a young age and those marrying late in life. Their comparison, however, does not
control for cohabitation, which may account for the higher life satisfaction of those who will marry.
22
As in most previous studies, the formation of cohabiting unions before marriage raises life
satisfaction significantly, in this case above the baseline value by 0.183 (Figure 2.2). In the year of
marriage and the year following, a significant boost in life satisfaction occurs for cohabitors and
noncohabitors alike, to a value 0.369 above the baseline, a significantly higher value than that for
premarital cohabitation. Thereafter, life satisfaction drops back to a value of 0.173, but is still
significantly above the baseline. This “marriage adaptation” value is not significantly different
from the value for cohabitation, a result not entirely surprising because I am comparing the
life satisfaction effects of cohabiting and marital unions for essentially the same partners. (The
effect estimated here for cohabitation excludes, of course, cohabitors who did not marry during
the survey period.) Thus, I find a “honeymoon period” effect on life satisfaction, followed by a
decline, presumably attributable to habituation. However, individuals in marital unions are still
happier, on average, than they were in their baseline period.
5
The results of the fixed-effects
analysis are very similar.
I find that the formation of successful unions, whether cohabiting or marital, has a positive
impact on well-being, but that there is no significant difference in the life satisfaction effect of
the two types of union in both the HLM and fixed-effects analyses. The implication appears
to be that the crucial element of life satisfaction is finding a compatible partner, whereas the
formalization of a union via marriage adds nothing to life satisfaction in terms of long-term well-
being. ThequalificationtothisconclusionisthatIdonotknowwhatthecourseoflifesatisfaction
would have been in the absence of marriage. I also point out that these results are averages. The
significantvariabilityintherandomslopecoefficientsmeansthatsomepeoplemightadaptfullyto
marriage while others might remain at or above the level of their honeymoon period. The divorce
subgroup differs from the first marriage group as a whole in two ways. As noted earlier, it is a
lower socioeconomic status group; in the absence of controls it has a baseline value significantly
5
Stutzer and Frey (2006: Table A2), using the German Socio-Economic Panel 1984-2000 and a fixed-effect
methodology, compare life satisfaction four or more years after marriage with that four or more years before, and
reach a similar conclusion.
23
less than for the first marriage group. The divorce subgroup also appears to be selective with
regard to personality traits conducive to lower life satisfaction. With controls for socioeconomic
characteristics, the baseline value of the divorce subgroup remains significantly negative (Table
2.2; see also Stutzer and Frey 2006; Lucas 2005).
My model gives no hint in the life satisfaction trajectory of the divorce subgroup before and
during marriage of impending marital dissolution. Although this group starts from a significantly
lower baseline value, it experiences effects from premarital cohabitation andfrom marriage – both
in the reaction and adaptation periods – that are not significantly different from those in intact
marriages (Table 2.2). Thus, the indications of prospective marriage breakup appear to lie in
the selective features of the divorce subgroup, and not in a different premarriage-to-postmarriage
trajectory. This result has to be taken with a grain of salt because the sample size of the divorce
subgroup is rather small. The coefficients for the divorce subgroup appear relatively large and
are likely not statistically significant because the small sample size leads to large standard errors.
A longer panel study in which the percentage of eventual divorces approaches the actual German
divorce rate might reject my result for this group.
I find no significant effect of children on life satisfaction, either for individuals who remain
married or those who do not. (For this reason, I omit children from the final regression results
given in table 2.2.) Studies of the life satisfaction effect of children are few and their results
mixed (Stutzer and Frey., 2006; Kohler et al., 2005). The effects of children on life satisfaction
is possibly exerted via two channels. On the one hand, children increase satisfaction with family
life; on the other hand, the added financial burden of children reduces satisfaction with one’s
economic situation. The disparate effects of children on the two domains tend to offset each
other, leaving overall life satisfaction unchanged. In my model, age has a significant negative
effect on life satisfaction. The implication here is that life circumstances other than the formation
of unions, such as circumstances related to health or working conditions, on average reduce life
satisfaction. If controls (such as age) for circumstances other than the formation of unions are
24
not included in the regression, the estimate of life satisfaction two or more years after marriage is
lowered and eliminates the lasting effect of the formation of a union itself. The omission of such
time-variant covariates that have been shown to influence life satisfaction can therefore lead to
erroneous conclusions about the effect of forming a union.
6
Although my model owes much to that of Lucas and colleagues in its baseline-reaction-
adaptation conception, there are important differences in the findings. Most importantly, I find
that individuals who remain married two or more years do not revert to their baseline value be-
fore marriage. On the contrary, I find life satisfaction of those who are married to be significantly
higher than their baseline value, at a level corresponding to that found for cohabitation preceding
marriage. ThedifferencebetweenmyresultsandthoseofLucasandcolleaguesdoesnotarisefrom
my larger sample. If I run my model on their sample, the same difference is found as reported
here. The difference arises from their failure to treat age as varying with time, and thus to control
for life circumstances that affect life satisfaction negatively. This peculiar treatment of age occurs
again in a recent defense of their conclusion that adaptation to marriage is rapid and complete
(Lucas and Clark, 2006).
I also differ from Lucas and colleagues in our baseline value for the first marriage sample.
They find that baseline life satisfaction of those who marry is significantly greater than for the
German Socio-Economic Panel population generally, and they posit a selection into marriage of
individuals with personality characteristics that attract marriage partners. My results, however,
indicate that the marriage sample is selected on socioeconomic characteristics and that these
characteristics suffice to explain the higher baseline value of life satisfaction. Once these charac-
teristics are controlled for, there is no room for a personality-based explanation of the baseline
6
Lucas and Clark (2006) note that happiness levels decline somewhat from ages 18 to 29 in the German Socio-
EconomicPanelandothernationalsamples. Toaccountforthiseffect, whichmightbecausedbylifecircumstances
that are highly correlated with age (e.g., declines in health), it is necessary to include age as a level-1 variable in
the hierarchical linear modeling analysis (within-subject). The inclusion of age as a level-2 moderator (between-
subjects) would only account for the effect of an individual’s time-invariant age (e.g., mean age or age at marriage)
on the slopes (see Raudenbush and Bryk (2002) for an explanation of time-variant vs. time-invariant covariates).
25
life satisfaction of the marriage sample. But my results do suggest that individuals who eventu-
ally divorce, a group not included in the analysis by Lucas and colleagues, may be selected on
personality characteristics that predispose this group to significantly lower baseline satisfaction
than those whose marriages remain intact. My results agree with Lucas et al.’s 2003 study in
two substantive respects. I, too, find a “honeymoon period” effect – a significant increase in life
satisfaction around the time of marriage. We both also find that life satisfaction drops two years
after marriage. But whereas they report that it falls to the premarriage baseline level – to the
setpoint value – I find it remains significantly above the baseline and at the same value as that
found for cohabitation.
2.4 Discussion
My study of data from the German Socio-Economic Panel covering the years 1984-2004 supports
the conclusions of previous cross-sectional studies on the effects of cohabitation, marriage, and
divorce on life satisfaction. I find that the formation of unions has a significant positive effect
on life satisfaction, while the dissolution of unions through separation or divorce has a significant
negative effect. These results are consistent with the “social support” interpretation commonly
offered for the association between marriage and life satisfaction. In contrast to other studies
in economics on marriage, I concentrate on the non-pecuniary benefits of a marital union by
controlling for household income. The formation of a union provides lasting gains in well-being
beyondpossiblebenefitsthataresolelybasedonincreasesinconsumptionthroughpooledincomes.
I find no evidence that children affect life satisfaction, either for those who remain married or
forthosewhodivorce. Intheyearofmarriageandthefollowingyear, Iseeasignificantadditional
boostinlifesatisfaction,a“honeymoonperiod”effect. Althoughthelifesatisfactionofindividuals
in intact marriages drops two or more years after marriage, presumably reflecting habituation, it
remains significantly higher than it was before marriage. The contrary conclusion of the study
26
by Lucas and colleagues (2003) – that life satisfaction two or more years after marriage reverts
to its level two or more years before marriage – arises from their failure to control for other life
circumstances that negatively affect life satisfaction. My findings thus run counter to the setpoint
model of psychology, whereby rapid adaptation to life transitions and events is pervasive and
happiness centers around a setpoint determined by genetics and personality. Instead, I find that
the formation of unions has an enduring positive effect on life satisfaction.
I find no significant difference between life satisfaction two or more years after marriage and
life satisfaction in cohabiting unions prior to marriage. Although I cannot identify the partners in
cohabitingunions, theyaremostlythesameonesasinsubsequentmaritalunions, becausealmost
70 percent of individuals in first marriages were cohabiting in the year preceding marriage. The
similarity in the life satisfaction estimates before and after marriage of those in unions suggests
that the formalization of unions by marriage has no significant impact on life satisfaction. What
is important is finding the right partner, not the nature of the union itself. This inference must
be qualified, however, by recognition that I do not know what course life satisfaction would have
followed had the partners in the unions I studied not married.
I find also, and not surprisingly, that compared with the German panel population generally,
the marriage sample is selective with regard to a number of socioeconomic characteristics: they
are younger, better educated, healthier, more likely to be employed, and more religious. Once
allowance is made for these characteristics, I find no evidence that persons who marry also have
personality traits that would make them more attractive as marriage partners. There is evidence,
however, thatthosewhosemarriages breakupdohavepersonalitytraits differentfrom the overall
populationthatmightadverselyaffectthelikelihoodofanenduringunion. Moreover,this“divorce
subgroup” is also distinctive in its lower socioeconomic status. I do not find a marriage trajectory
for this divorce subgroup – cohabitation-marriage reaction-marriage adaptation – any different
from that of individuals in first marriages that remain intact, but this result might change with
a larger sample because of currently lacking statistical power. The implication I draw is that the
27
roots of prospective dissolution lie in the distinctive socioeconomic and personality traits of those
destined for separation and divorce, and not in a disparate course of life satisfaction in the first
years of marriage. The results for the divorce group have to be taken with caution because of the
small size of the sample.
Though my analysis is based on German panel data I would expect to find the same results
for the U.S. and other countries. Cross-sectional studies have repeatedly shown that married
individuals report higher levels of life satisfaction than their unmarried counterparts. Easterlin
(2006) finds that satisfaction with family life in the U.S. increases until midlife and then declines,
indicating that people benefit from marriage and experience declines after midlife due to marital
dissolution.
Myanalysisfocusedonthequestionwhetheradaptationtomarriageoccursonlytwoyearsafter
marriage as has been found by Lucas et al. (2003). Although I do not find complete adaptation
only two years after marriage, it is of course possible that full adaptation occurs later on due
to habituation and issues related to marriage duration. An interesting extension of the present
chapter could include models which allow for a longer reaction period, maybe covering five or ten
years. Extending the reaction period to a considerably larger interval would require more years
of observations for each individual and likely reduce the sample considerably. It would also be
interestingto analyze whether the durationofcohabitationhas aneffectonsubsequent well-being
during marriage. It is possible that individuals who cohabit for a long period before marriage
and possibly with varying partners experience less benefits from formalizing the union than their
counterparts who are less experienced in cohabitation. Another possible extension of the present
analysis could consider the benefits of re-marriage by selecting a sample of previously married
individuals and assessing the benefits in terms of well-being that they experience in a new union.
28
Chapter 3
Financial Satisfaction over the Life Cycle: The Influence of
Assets and Liabilities
3.1 Introduction
Thischapterdiscussesfinancialsatisfactionanditsdeterminantsbylookingatchangesoverthelife
cycle of both financial satisfaction and various explanatory variables. There have been numerous
studies on the determinants of global subjective well-being, but research on domain specific well-
being, such as financial satisfaction, is rather limited. Studies of global well-being often assess the
effect of objective factors – such as income, family status, and employment status – on subjective
well-being and thus do not explicitly take into account the possibility that people’s perception of
these domains of life might differ from the objective circumstances represented by these variables.
The domain satisfaction approach pioneered by psychologist Angus Campbell and collaborators
(Campbell et al., 1976; Campbell, 1981) asserts that global well-being depends on the satisfaction
experienced in various domains of life.
1
In a recent study, Easterlin (2006) shows that life cycle
changes in overall subjective well-being can be explained by changes in satisfaction in various
domains. The regression coefficients in these studies allow one to assess the importance of each
1
The theory that several domains of life determine overall happiness is often referred to as a “bottom-up”
approach.
29
domain on overall life satisfaction, and financial satisfaction – the topic of the current chapter –
consistentlyranksamongthemostimportantdomains(seealsovanPraagandFerrer-i-Carbonell,
2004; Layard, 2005).
Research on financial satisfaction is still sparse and the subject deserves further attention.
Van Praag (2005) shows that there are two substantially different ways to measure financial
satisfaction, which both lead to similar results. One approach – which is also the approach
followed in the present study – is to ask people about their satisfaction with their financial
situation. Another approach, pioneered by the Leyden school in the 1970s, is to employ income
evaluation questions which asses what level of income the respondent perceives to be adequate
(e.g. van Praag, 1971; van Praag and Kapteyn, 1973).
The aim of the present study is to analyze the determinants of financial satisfaction and thus
contribute to our understanding of one of the most important domains influencing overall life
satisfaction. An important contribution of the current analysis is the consideration of assets and
liabilities in addition to income. One of the most discussed topics in the economics literature
on subjective well-being (SWB) is income, and not surprisingly the first paper on SWB by an
economist discussed the influence of income on well-being (Easterlin, 1974). Economists often
assumethatindividualwell-beingisstronglyandpositivelyinfluencedbyincome,andmostpeople
when asked about what is important for their well-being, support this assumption. My results
indicate that this is so, but other factors play an important role too.
3.2 Life cycle financial satisfaction
Evidence from the National Survey of Families and Households (NSFH) indicates that life cycle
financial satisfaction steadily increases with age, whereas life cycle income shows an inverted
U-pattern with a peak at midlife (see figures 3.1 and 3.2). This observation does not support
30
the common assumption that income is the main factor determining one’s perception of financial
well-being.
Figure 3.1: Lowess estimation: Financial satisfaction
Figure 3.2: Lowess estimation: Log household income
Further evidence in the NSFH suggests that people’s perceptions of their financial situation
changewithage. Whenaskedwhethertheythinkthattheirstandardoflivingwillgetmuchworse
when they retire, respondents in wave 3 of the NSFH seem to be substantially less worried about
retirement than in wave 2. Only the responses of people who answered this question in both wave
31
2 and 3 are listed in table 3.1, yielding a sample size of 2,416 in each survey year. This result
is interesting because most of the respondents did probably not experience significant changes in
their employment or their household situation over the 10-year period between the two surveys.
Individuals who remain at the same place of employment should be able to estimate reasonably
well the level of income they can expect when they retire.
Table 3.1: Answers of respondents who answered the question “My standard of living will get
much worse when I retire”, NSFH, wave 2 and 3
Response Wave 2 Wave 3
Strongly agree 4.9 4.1
Agree 19.0 22.4
Neither agree nor disagree 38.7 10.9
Disagree 30.1 55.0
Strongly disagree 7.3 7.7
This pattern of increasing financial satisfaction with age is not idiosyncratic. In the U.S.,
one can observe the same pattern in the General Social Surveys (Davis et al., 2005), a repeated
cross-sectional survey with observations from 1972 to 2004 (Figure 3.3). This pattern has also
been observed by other researchers. Numerous studies in financial gerontology have shown that
financial satisfaction is surprisingly high at old age despite low levels of income after retirement
(see George 1992 for an overview). Most of the studies reviewed by George (1992) use – like
the present analysis – U.S. data and one might argue that this is a peculiarity of Americans’
sense of financial well-being. A recent study using Norwegian data, however, points to the same
seemingly paradoxical observation (Hansen, 2005). The Norwegian study employs income reports
from public registries and thus avoids problems with possible underreporting of financial means
at older ages. Nor does the life cycle pattern depend on the wording of the question. My own
analysis of German panel data shows a similar pattern for satisfaction with household income and
one’s standard of living. Although these questions measure somewhat different concepts one can
still observe the same upward-sloping pattern of life cycle financial satisfaction at old ages after
an initial decline (Figure 3.4). Though the authors do not specifically point out the high levels
32
of financial satisfaction at old ages, evidence of this pattern can also be found in studies from
Spain (Vera-Toscano et al., 2006) and Ireland (Delaney et al., 2006) in the form of a significant
and positive coefficient of age in regressions of financial satisfaction on a group of explanatory
variables.
Figure 3.3: Financial Satisfaction in the U.S., General Social Surveys, 1972-2004, ages 18-89
Figure 3.4: Satisfaction with household income and standard of living in West Germany, German
Socio Economic Panel, 1991-2003, ages 18-90
33
Someofthepreviouslymentionedstudiesemploycross-sectionaldata,i.e. financialsatisfaction
is measured at one point in time; hence it is impossible to infer whether seemingly age-related
differences in financial satisfaction are actually associated with age or in fact reflect differences
between birth cohorts. The present analysis employs longitudinal data and confirms the financial
satisfaction-age relationship that can be observed in cross-sectional studies.
My analysis suggests that there are other factors, economic and non-economic, besides income
which lead to rising financial satisfaction in old age. In what follows I will look at changes in
several measures of assets and liabilities over the life cycle. The present study further takes into
account changes in financial obligations represented by the dependency burden and medical costs
in the form of self-rated health.
To my knowledge, apart from a cross-sectional analysis by Hansen (2005), the current analysis
is the only study which specifically considers the influence of assets and liabilities on financial
satisfaction. Anotherdistinctfeatureofmystudyistheanalysisofindividualfinancialsatisfaction
and important explanatory variables over the life cycle in order to assess whether these variables
might explain the seemingly paradoxical observation of increasing financial satisfaction despite
decreasing incomes.
InwhatfollowsIdescribethedataandmethodology,followedbyapresentationofthelifecycle
profiles of various explanatory variables. I then postulate a model of financial satisfaction, and
present my findings from regression analyses which assess the relative weight of each explanatory
variable on financial satisfactiob. The last section concludes the analysis.
34
3.3 Data and methods
3.3.1 Data
The data are from the National Survey of Families and Households (NSFH).
2
Interviews for the
NSFH were conducted in three waves in 1987-1988, 1992-1994, and 2001-2003. For the first wave
in 1987-88 (Sweet et al., 1988), one adult per household was randomly selected from 13,007 U.S.
households. Households with blacks, Puerto Ricans, Mexican Americans as well as single-parent
families, families with step-children, cohabiting couples and recently married persons were over-
sampled. A large portion of the interviews with the primary respondent were self-administered to
allow the respondent more privacy. Shorter questionnaires were given to the spouse or cohabiting
partner of the primary respondent. The second wave was conducted as a five year follow-up study
from1992-1994(SweetandBumpass,1996). Thisfollow-upincluded10,007oftheoriginalwave1
primary respondents, as well as their current spouses or cohabiting partners and, if relationships
had ended between waves, also interviews with the NSFH1 spouses and partners. In addition,
telephone interviews were conducted with some of the household’s children. Data for the third
wave from 2001-2002 (Sweet and Bumpass, 2002) were collected through telephone interviews
with only primary respondents who were either above age 45 or had a child who was interviewed
in wave 2, as well as with their spouses and their previously interviewed children. The sample
size of the third wave survey is thus considerably lower than for the first two waves (n = 7,277).
The present study uses two different samples of the NSFH data. The first sample consists of
primaryrespondentsinwave2whoarehouseholdersandhavenon-missingvaluesforthefinancial
satisfactionmeasure. Iselectonlyhouseholdersforthesamplebecausenon-householderswerenot
2
The first wave of the National Survey of Families and Households was funded by a grant (HD21009) from
the Center for Population Research of the National Institute of Child Health and Human Development; and the
second and third waves were funded jointly by this grant and a grant (AG10266) from the National Institute on
Aging. The survey was designed and carried out at the Center for Demography and Ecology at the University
of Wisconsin-Madison under the direction of Larry Bumpass and James Sweet. The field work for the first two
waves was done by the Institute for Survey Research at Temple University, and the third wave by the University
of Wisconsin Survey Center.
35
asked about the income of other household members; hence the household income measure for
these respondents reflects only their personal incomes. 90.67 % of the respondents in the original
data set are considered to be householders, i.e. they or their spouses rent or own the place where
they reside. Table 3.2 includes descriptive statistics for this sample. Financial satisfaction is
measured by a question which asks the respondents to rate on a scale from 1 to 7 how satisfied
they are, overall, with their financial situation, where 1 denotes “very dissatisfied” and 7 denotes
“very satisfied”. The distribution of the responses to this financial satisfaction question is skewed
towards higher valuations (Figure 3.5).
Figure 3.5: Financial satisfaction, NSFH, wave 2 and wave 3
Income is measured in the form of household income because the debt measures are also
collected at the household level. A few respondents report zero household income and only the
cases in which the respondent owns substantial financial or tangible assets are kept.
36
Table 3.2: Descriptive statistics, full sample, NSFH, wave 2 (income, liabilities and assets in
$1993)
Variable n Mean Std. dev. Min Max
Financial satisfaction 8,855 4.708 1.646 1 7
Household Income
a
9,160 50,052 47,592 0 999,995
Assets
Financial assets 8,458 33,651 49,454 0 205,994
Tangible assets: Value of home 9,161 81,643 108,889 0 1,029,971
Homeownership in % 9,088 0.744 0.437 0 1
Absolute debt
Credit card debt 8,881 1,204 3,052 0 68,253
Debt on home 8,764 28,693 52,139 0 746,730
Loans on purchases 8,998 202 1,146 0 70,000
Educational loans 9,057 569 4,134 0 99,999
Bank loans 9,029 662 4,659 0 102,996
Loans from friends 9,077 146 1,849 0 99,999
Loans for home improvement 9,073 211 2,017 0 72,098
Bills 9,001 244 2,470 0 99,999
Other debt categories 9,027 202 2,626 0 99,999
Monthly debt payments on...
Mortgage payments 9,027 345 544 0 8254
Loans on purchases 8,975 17 182 0 8,000
Educational loans 9,052 11 111 0 8,000
Bank loans 9,010 25 271 0 10,299
Loans from friends 9,032 4 65 0 4,000
Loans for home improvement 9,069 4 32 0 600
Other debt categories 9,102 7 149 0 9,749
Other controls
Self-rated health (1 = low) 8,921 3.962 0.840 1 5
Children in household 9,161 0.829 1.190 0 7
Other adults in household 9,161 1.032 0.767 0 7
Married 9,158 0.698 0.459 0 1
Separated 9,158 0.027 0.163 0 1
Divorced 9,158 0.094 0.292 0 1
Widowed 9,158 0.088 0.283 0 1
Never married 9,158 0.092 0.289 0 1
Unemployed 9,161 0 .014 0.116 0 1
Retired 9,161 0.170 0.375 0 1
Black 9,149 0.096 0.295 0 1
Male 9,161 0.467 0.499 0 1
Education above high school 9,134 0.466 0.499 0 1
Age 9,156 48.730 16.097 23 97
NSFH2 weights provided by the survey institute are used.
a
The few people reporting zero household incomes own substantial assets (see Appendix B.1).
37
Table 3.3: Descriptive statistics, waves 2 and 3, balanced panel (income, liabilities and assets in
$1993)
Wave 2 Wave 3
Variable n Mean Std. dev. n Mean Std. dev.
Financial satisfaction 3,751 4.667 1.620 3,751 5.224 1.511
Household Income
a
3,751 55,965 53,818. 3,668 61,503 70,210
Assets
Financial assets 3,502 37,264 51,280 3,356 80,326 108,989
Tangible assets: Value of home 3,751 88,334 105,664 3,751 110,720 123,052
Homeownership in % 3,734 0.805 0.396 3,602 0.863 0.344
Absolute debt
Credit card debt 3,647 1,405 3,291 3,304 746 1,533
Debt on home 3,580 32,512 52,190 2,939 39,304 58,488
Loans on purchases 3,694 245 1,177 3,682 247 1,756
Educational loans 3,719 425 3,113 3,704 752 4,599
Bank loans 3,706 781 5,081 3,680 769 5,016
Loans from friends 3,729 184 2,432 3,705 62 853
Loans for home improvement 3,726 300 2,499 3,706 286 2,220
Bills 3,698 308 3,080 3,652 273 2,342
Other debt categories 3,706 233 2,764 3,660 286 2,942
Monthly debt payments on...
Mortgage payments 3,680 373 496 3,488 384 535
Loans on purchases 3,683 20 184 3,654 11 65
Educational loans 3,719 8 94 3,688 10 61
Bank loans 3,696 27 283 3,654 19 135
Loans from friends 3,715 4 75 3,693 1 22
Loans for home improvement 3,724 6 39 3,693 6 46
Other debt categories 3,725 7 167 3,624 6 52
Other controls
Self-rated health (1 = low) 3,709 3.994 0.801 3,751 3.964 0.917
Children in household 3,751 0.881 1.202 3,751 0.398 0.829
Other adults in household 3,751 1.026 0.771 3,751 1.007 0.796
Married 3,751 0.677 0.468 3,751 0.625 0.484
Separated 3,751 0.027 0.162 3,751 0.026 0.160
Divorced 3,751 0.164 0.370 3,751 0.178 0.383
Widowed 3,751 0.071 0.257 3,751 0.117 0.321
Never married 3,751 0.062 0.240 3,751 0.054 0.226
Unemployed 3,751 0.015 0.120 3,751 0.016 0.125
Retired 3,751 0.101 0.302 3,751 0.146 0.353
Black 3,747 0.134 0.341 3,747 0.134 0.341
Male 3,751 0.360 0.480 3,751 0.359 0.480
Education above high school 3,748 0.518 0.500 n/a n/a n/a
Age 3,751 48.25 10.26 3,751 56.95 10.12
a
The few people reporting zero household incomes own substantial assets (see Appendix B.1)
38
The second sample, a balanced panel used to derive life cycle profiles, consists of primary
respondent data from the second wave from 1992-94 and the third wave from 2001-2002. I do
not use data from the first wave from 1987-88 because the measure for financial satisfaction was
not included in that survey. This data set only includes individuals who are householders and
answered the question on financial satisfaction in both waves 2 and 3, thus creating a balanced
panel. I also exclude individuals who are younger than 30 years or older than 80 years due to very
small sample sizes at these ages. The follow-up interviews in the third wave were only conducted
with individuals above age 45 and individuals who had focal children in wave 2. Thus, there are
few individuals who were under age 30 in wave 2 and who were re-interviewed in wave 3. A few
observations which seemed to include misreported household income measures were also deleted
(see Appendix B.2.1). This yields a final sample of 3,751 individuals with observations in both
waves (see table 3.3 for descriptive statistics).
The balanced panel differs from the weighted wave 2 panel in several distinct ways because
of the limited number of observations in wave 3. The respondents in the balanced panel are, on
average, more likely to be divorced, be black and have an education beyond highschool than the
average individual in the weighted wave 2 sample. At the same time, they are less likely to never
have married, be retired and be male. This shift in sociodemographic characteristics probably
mostly occurred because the balanced panel does not use population weights due to the lack of
weights in the third wave sample. The wave 2 respondents who remained in the balanced panel
are also financially better off than the respondents in the whole wave 2 sample, which is reflected
by, on average, higher household income, more financial and tangible assets as well as higher
liabilities. The results could be affected by the selective nature of the balanced panel because
several key characteristics, such as income, are overestimated. The higher average household
income value stems from the overrepresentation of highly educated individuals in the sample, not
from the overrepresentation of divorced and black individuals (Table 3.4). The undersampling
of never married and retired individuals also contributes to this pattern. In the following I will
39
therefore compare the results from calculations using the weighted wave 2 sample to those using
the balanced panel.
Table 3.4: Household income by sample characteristics, NSFH wave 2 balanced sample (income
in $1993)
Characteristic n Mean Std. dev. Min Max
All respondents in
weighted wave 2 sample 9,160 50,052 47,592 0 999,995
All respondents in
balanced wave 2 sample 3,751 55,965 53,818 0 999,995
Divorced 614 38,099 50,558 0 999,995
Black 503 38,616 32,955 0 230,000
Education above high school 1,942 70,477 65,635 0 999,995
Never married 231 32,795 28,318 0 203,000
Retired 380 36,820 35,895 0 310,000
Male 1,347 61,802 51,611 0 732,516
3.3.2 Methods
The advantage of the panel data is that it is possible to follow individuals over time and thus
derive life cycle profiles for the variables of interest. I first estimate nonparametrically the life
cycle profiles of variables other than income that one could reasonably consider to determine
financial satisfaction. In particular, I look at changes over the life cycle of various measures of
assets, liabilities and financial obligations.
A nonparametric approach has the advantage that it neither prescribes the functional form
of the regression curve, nor the error distribution.
3
One might consider the following example to
illustrate the advantage of a nonparametric approach for this analysis in which we assume that
thetruelifecycleprofileoffinancialsatisfactionisasteadyincreasewithageuntilage60followed
by a constant level of satisfaction. If I fit a quadratic curve to life cycle financial satisfaction, this
imposed functional form will probably yield a curve which shows an initial increase of financial
3
See H¨ ardle (1990) for an overview of nonparametric regressions.
40
satisfaction followed by a subsequent decrease. The quadratic functional form does not allow the
curve to flatten out after a certain age. A cubic specification might approximate the true curve
reasonably well, but if I start with a quadratic specification I might easily overlook that the cubic
specification is more appropriate. A nonparametric approach does not impose a functional form
and thus allows the fitted curve to take any shape.
To obtain nonparametric life cycle profiles, I first take the residuals from an individual fixed
effects regression of the variable of interest. I include wave dummy variables to account for
period effects. The residuals thus neither include individual fixed effects nor period effects. The
residuals are then used as the dependent variable in a kernel regression. I use locally weighted
scatter plot smoothing (lowess), proposed by Cleveland (1979), to estimate life cycle curves when
the residuals are plotted against age. This procedure is robust against outliers which might
otherwise dominate the estimated statistics (see H¨ ardle, 1990). The life cycle profiles that are
obtained nonparametrically could be taken as the basis for postulating a parametric model for
each variable of interest, but for my purposes the nonparametric profiles are sufficient. For the
life cycle figures in this paper, I added the sample mean of the variable of interest to the resulting
residual of the lowess estimation.
ThetwoNSFHsurveyyearsarespacedtenyearsapartandthereforedonotallowmetofollow
one single cohort for fifty years from age 30 to 80. The life cycle profiles here have to be regarded
as an approximation of the life cycle profile of a single birth cohort because observations at young
agesaremostlysuppliedbyrespondentsofrecentbirthcohorts. Similarly,thepartofthelifecycle
profiles at old ages is mostly determined by observations of respondents of older birth cohorts.
Although these life cycle curves can therefore only be regarded as approximations to the actual
experience of an individual, comparisons with the actual experience of single birth cohorts in the
Current Population Surveys (see Appendix B.2.2) suggest that the present analysis provides a
reasonably close fit. Longitudinal studies on the life cycle income experience of individuals within
41
cohorts also indicate that incomes usually increase early in the life cycle and decline after midlife
(Duncan et al., 1987).
I then assess the relative influence of each explanatory variable on financial satisfaction in
severalregressions. Thecross-sectionalsampleofthesecondNSFHwaveisthemostrepresentative
sample of the overall population, but it does not allow me to assess whether people change their
valuation of their financial situation when their economic circumstances change or as they age.
The panel structure of the second sample allows the use of individual fixed effects analysis to
accountforunobserved,time-invariantindividualcharacteristics. Thefixedeffectsregression,also
referred to as within-estimation, uses the within-individual average of financial satisfaction as a
dependentvariableandregressesitonthewithin-individualaveragesoftheexplanatoryvariables.
A regression which controls for time-invariant individual attributes by using individual-specific
dummy variables (α i
) would yield the same results.
(FS
it
− FS
i
)=(x
it
− x
i
)β +(
it
−
i
) (3.1)
where FS = financial satisfaction, x = a vector of explanatory variables, = residual, y
i
=
P
t
y
it
/T
i
, x
i
=
P
t
x
it
/T
i
, and
i
=
P
t
it
/T
i
.
Giventheordinalnatureofthemeasureoffinancialsatisfaction,anorderedprobitspecification
with random effects would also be appropriate. Ferrer-i-Carbonell and Frijters (2004) show that
ordinal and cardinal estimations, such as the fixed effects estimation used here, usually yield very
similar results. I also use an ordered probit specification to ensure that the results are robust to
methodology. Similarly,Ireplicatetheweightedleastsquaresregressionsofthecross-sectionusing
an ordered logit specification. The results of both the fixed-effects estimation and the ordered
probitestimationwithrandomeffectsindicatetherelativeimportanceofeachexplanatoryvariable
for an individual’s sense of financial satisfaction. I also include measures of the unemployment
rate and inflation rate to account for macroeconomic period effect.
42
3.4 A model of financial satisfaction
The domain satisfaction approach postulates that satisfaction in each domain depends on the
extent to which objective circumstances fulfill one’s aspirations. Campbell and his collaborators
(1976) note that aspirations are often formed on the basis of comparisons to relative standards.
Satisfaction declines when the gap between aspirations and the individual’s perception of his
own situation increase. This approach can be described as a relative standards model, in which
peopleevaluatetheirstandingbasedonstandardswhicharedeterminedbycomparisonstoothers,
their own past and their desires. Similarly, Michalos’ multiple discrepancy theory (Michalos,
1985, 1991) describes that satisfaction is determined by the discrepancy between an individual’s
attainmentsandmultiplestandards. Inhisempiricalanalysisofthemodel,thestrongestpredictor
of satisfaction was the discrepancy between what one wants and what one has.
Solberg et al. (2002) test the relative standards model in three experimental settings in which,
among other things, they find that satisfaction with income depends to a large extent on an
individual’s ability to purchase desired items with that income. They also note that individuals
seem to adjust their levels of desires and thus regulate satisfaction levels. Similarly, Campbell
et al. (1976) point out that aspirations seem to be lower among the old than among the young
and could thus explain high levels of satisfaction in old age.
Financial satisfaction (FS) can therefore be described to be a function of financial means and
financial aspirations.
FS
it
= f(financial means
it
,financial aspirations
it
) (3.2)
where the subscript i denotes an individual and t indicates time.
Financial means can easily be measured by asking individuals about their income and assets,
but how can one assess financial perceived needs? Indeed, only a few studies have attempted to
assess aspirations directly. Two notable examples in economics are a study by Stutzer (2004),
43
and the next chapter of this dissertation in which I analyze changes in aspirations over the life
cycle. The data set which I employ in the current analysis, the National Survey of Families and
Households, does not include direct measures of aspirations, unlike the Roper survey which I
use in the following chapter. Nevertheless, a strong indicator for aspirations exceeding financial
means can be found in the presence of debt. Individuals who wish to fulfill their perceived
needs accumulate debt if their current financial means do not match their aspirations (Figure 3.6
illustratesthispoint). Moreover, debtmightcauseemotionalstrainandthusloweranindividual’s
financial satisfaction considerably. Financial needs can also be measured by objective indicators,
such as the dependency burden and health status, which contribute substantially to a household’s
necessary expenses.
Figure 3.6: Debt as an indicator of the discrepancy between financial aspirations and financial
means
Others have studied the effect of social comparison on financial satisfaction (Hsieh, 2000;
Burchardt, 2005) and they find that social comparison significantly influences satisfaction with
one’s financial situation. I do not consider the effects of social comparison (what relevant other’s
have; also referred to as relative deprivation) and hedonic adaptation (what the individual had
44
in the past) separately because these two psychological mechanisms are the underlying process in
the formation of aspirations and therefore are indirectly reflected in the level of debt.
But how might the discrepancy between perceived needs and financial means explain the
seeminglyparadoxicalpatternoflifecyclefinancialsatisfaction? Ifinalifecyclemodel,aspirations
are particularly high at young ages and cannot be satisfied with current income then individuals
will incur debt especially at young ages. Decreasing incomes after midlife likely have a negative
impact on financial satisfaction but the upward-sloping pattern of financial satisfaction might be
caused by decreases in debt and thus less emotional strain. Declining levels of debt allow for two
explanations. On the one hand, it is possible that with rising incomes individuals do not have to
incur debt in order to afford the things they want, say a new car. On the other hand, as people
age they might lower their material aspirations and thus do not feel the need to spend more than
they can afford. Moreover, the accumulation of financial and tangible assets with age provides
security and possibly lowers emotional strain from debt because tangible assets, for instance a
house, could be sold in times of financial hardship. The model of financial satisfaction that I
propose includes measures of assets and liabilities. It also includes control variables which reflect
costs and thus economic needs, such as health status and household size.
A model of financial satisfaction ( FS) in this study is thus of the following form:
FS
it
= f( Y
it
,A
it
| {z }
fin. means
, L
it
,O
it
| {z }
fin. aspirations, needs
) (3.3)
where Y = household income, A = a vector of assets, including financial and tangible assets,
L = liabilities, including credit card debt, mortgage debt and other monthly debt payments, and
O = a vector of other control variables which reflect financial needs, such as household size and
health. As these data are panel data the subscript i denotes an individual and t indicates time.
Thepresentstudyfocusesontheeffectsofincome,assetsandliabilitiesonfinancialsatisfaction
and their changes over the life course. In particular, I hypothesize that income and assets both
45
exert a positive impact on financial satisfaction while debt causes emotional strain and thus
reduces satisfaction. Moreover, debt indicates that there is a discrepancy between aspirations and
attainments. I first analyze assets and liabilities separately instead of constructing a composite
net wealth measure because the same amount of net wealth can be derived from substantially
different compositions of the individual components. For instance, someone with $100,000 in
financial assets, $0 debt and no tangible assets has a net worth of $100,000. On the other hand,
a person who has $50,000 in financial assets and owns a home which is valued at $300,000, but
on which he owes $250,000, has the same level of net worth. One can expect that the level of
financial satisfaction differs between these two people. The person who owns the house has to
make monthly mortgage payments and thus feels the pressure of paying his debt on time. On
the other hand, he also benefits from owning this house because it could be sold in times of great
financial need and thus provides security. I will later also consider an aggregate measure which
indicates net worth.
In studies of overall satisfaction with life, income usually shows a significant positive effect.
It is often assumed that the positive effect of income on general life satisfaction is caused by its
impact on financial satisfaction, which in turn is one of the domains that influence overall well-
being (Campbell et al., 1976; Campbell, 1981). Thus, if two households have the same level of
financial assets and liabilities, the household reporting a higher level of income will most likely
report a higher level of financial satisfaction.
Given the level of income and debt, I hypothesize that measures of wealth, i.e. financial and
tangible assets, have a positive impact on financial satisfaction. Higher rates of current savings
increase net wealth and provide a financial source when incomes decline at old age. Individuals
who own tangible assets such as a home benefit from lower living costs once the mortgage is paid
off.
The analytical relationship between liabilities and financial satisfaction is more complicated.
I distinguish between three different measures of debt instead of aggregating all into one measure
46
because some types of debt can be considered to be better than others. Credit card debt is a
measure of the credit card balance that is not paid off at the end of the month. Drentea observes
apositiverelationbetweenanxietyandtheratioofcreditcarddebttoincome(Drentea,2000). In
contrast to mortgage debt, credit card debt is usually not associated with a big-ticket consumer
item which might provide security in times of hardship. A second variable indicates the total
amount that the respondent still owes on his house. A third debt measure aggregates other forms
of debt such as loans on purchases, loans from friends and outstanding bills.
Therelationshipbetweendebtandfinancialsatisfactionissomewhatmorecomplicatedbecause
debtcouldhaveapositiveimpactonsatisfactionthroughthegoodsandservicesthatareacquired
withit. Nevertheless,ananalysiswhichholdsincomeandfinancialassetsconstantandinaddition
accountsforsomeofthetangibleassetsthatarefinancedthroughdebt, shouldrevealasignificant
negative impact of debt. Debt can be assessed in two different ways; either by its total amount or
instead by the monthly payments an individual has to make to pay off this debt. I will consider
both forms of debt in the following.
I further include various explanatory variables that account for financial stressors, such as
large expenditures in the previous year – e.g. due to illness – which might have a negative effect
on financial satisfaction (Joo and Grable, 2004). A measure of self-rated health proxies for the
cost of health care and indicators of household size account for differences in expenditures. This
list of measures of financial needs is certainly far from complete and is restricted due to data
limitations.
3.5 Findings
Incomeisoftentakenasaproxymeasureofwealthduetodatarestrictions,butfigure3.2indicates
that this treatment of income is not sufficient for an analysis of financial satisfaction. Financial
satisfactionincreasessteadilywithagedespitedecreasesinhouseholdincomeaftermidlife(Figure
47
3.1). It is impossible to reconcile these two life cycle profiles with the assumption that income is
the primary determinant of financial satisfaction.
The NSFH data provide detailed information on several components of wealth. Maybe assets
are more important for an individual’s satisfaction with his finances because the accumulation
of assets indicate that one has enough income left to invest in assets. The value of the financial
assets that an individual holds increases steeply until about age 50 and then levels off with a
gradual decline at old age (Figure 3.7). Similarly, the average value of tangible assets in the form
of homes increases until midlife and then remains mostly constant with a slight decline at old age
(Figure 3.8).
Figure 3.7: Lowess estimation: Log financial assets
This increase in the value of homes indicates that individuals do not remain satisfied with the
first homes they purchase, but instead “upgrade” their homes after a while. This can be seen as
evidence of increasing material aspirations and “conspicuous consumption” (Veblen, 1899) – i.e.
individuals do not purchase homes solely for their practical value, but also see them as a means
to position themselves in society. I will discuss the idea of increasing material aspirations and
positional goods more in detail in the next chapter.
48
Figure 3.8: Lowess estimation: Log home value
Liabilities can be seen as an indicator of high material aspirations that exceed financial means
and therefore likely have a negative impact on financial satisfaction. Total credit card debt
increases in the early 30s and then continuously declines with age. This decline in debt might
contribute to higher levels of financial satisfaction at old age. In wave 2 about 44% of the
respondents report having a credit card balance which they do not pay off at the end of the
month. This percentage decreases to 30.96% in the 10-year follow up study (Table 3.5). More
than 50% of all respondents report owing money on their homes in both waves of the survey.
The life cycle profile of mortgage debt shows that debt on homes increases until age 47 and
then steadily declines (Figure 3.10). Some of this decline can be attributed to slight declines
in homeownership, but most of it is probably due to the fact that respondents start to pay off
their mortgages completely. Mortgage payments take up a large part of disposable income, but
homeownership is an important source of wealth for most households (Mishel et al., 2005). The
level of other debts also steadily declines with age (Figure 3.11).
It has to be considered that the acquisition of debt depends to a large extent on the supply
side of debt – namely credit card companies, mortgage companies and other lenders – and their
49
Table 3.5: Percentage of respondents who have certain types of debt
Wave 2 Wave 3
Type of debt Percentage with debt Percentage with debt
Any type of debt 81.35 75.57
Debt on homes 56.52 52.98
Credit card debt 44.04 30.96
Loan on purchases 12.32 8.72
Bills 10.49 9.53
Bank loans 10.79 7.83
Educational loans 6.94 6.83
Other debt 4.18 4.43
Loan for home improvement 3.22 3.43
Loan from friends, family 2.84 1.35
Figure 3.9: Lowess estimation: Log credit card debt
willingness to provide the needed loan. Many individuals who would like to acquire a mortgage
loan might not receive one due to a low credit rating. It is difficult to assess in how far liabilities
in the present study are limited because of restrictions from the supply side. In the cross-section,
one can observe that higher income households also have higher liabilities, but over the life cycle
increases in income are generally associated with declining levels of debt. The present analysis
is mostly concerned with age related changes in assets and liabilities. Possible restrictions from
the supply side of debt do not hinder the study considerably because even though the average
50
respondent experiences increasing income with age, liabilities decrease over the life cycle despite
likely improvements in the access to loans.
Figure 3.10: Lowess estimation: Log debt on homes
Figure 3.11: Lowess estimation: Log other debt
I argued above that a detailed analysis of the components of net wealth probably allows a
better explanation of financial satisfaction. In fact, it would only be appropriate to aggregate the
various measures of assets and liabilities into a composite net wealth measure if the effect of all
single components were the same on financial satisfaction. The regression analysis in the next
51
section will provide this information, but does a descriptive analysis of the life cycle pattern of
net wealth suggest that it might sufficiently explain financial satisfaction? Net wealth increases
somewhat until midlife and then declines slightly at old age (Figure 3.12). Net wealth and income
alone could barely explain the continuously increasing pattern of satisfaction in the financial
domain.
Figure 3.12: Lowess estimation: Log net wealth
Financialaspirationsofcoursedependtoalargeextentonactualfinancialneeds. Thepresence
of other household members indicates higher living costs. As children leave the household or
become older and are therefore no longer classified as children, the total number of children in the
household steeply declines with age (Figure 3.13). Similarly, the number of adults in a household
declines after midlife (Figure 3.14). These declines in the dependency burden possibly partly
explain increases in financial satisfaction.
52
Figure 3.13: Lowess estimation: Children in household
Figure 3.14: Lowess estimation: Other adults in household
I expect that self-rated health will steadily decline with age leading to increased medical costs
for the household and lower financial satisfaction. The analysis confirms this assumption (Figure
3.15). Self-rated health remains rather constant until later in life and then starts to decline.
Together with income, self-rated health is the only variable that would lead one to expect lower
financial satisfaction at old age.
53
Figure 3.15: Lowess estimation: Self-rated health
But do assets and liabilities, which in some way represent the balance of financial aspirations
andmeans,haveindeedalargerimpactonfinancialsatisfactionthanincome? Regressionanalysis
revealstherelativeweightsofeachexplanatoryvariable. Ifirstpresenttheresultsofweightedleast
squares regressions and then compare these results to other specifications, which do not assume
cardinality of the dependent variable, i.e. ordered logit regressions.
4
The initial regressions use
only wave 2 data because this is the most representative sample of the overall population.
Oneofthegoalsofthisstudyistoanalyzewhetherincomeisthemaindeterminantoffinancial
satisfaction, as is often assumed. Not surprisingly income does indeed have a significant positive
impact on an individual’s sense of financial well-being, but financial assets and the measure of
otherformsofdebtalsodisplayalargeimpact(Table3.6,column1). Ownershipoffinancialassets
increases financial satisfaction more than owning tangible assets in the form of a house, which is
reflectedbyacoefficientabouttwiceaslargeastheoneforthelattervariable. Asexpected, assets
generally display a positive impact on financial satisfaction while liabilities display a negative
impact. It does not seem to matter whether one considers the absolute amount of debt that is
4
Ordered probit regressions yield quite similar results and are thus not reported here.
54
Table 3.6: Weighted least squares regression on financial satisfaction, wave 2 (dollar amounts in
$1993)
Full wave 2 sample Restricted sample
(1) (2) (3) (4) (5) (6)
Coeff. Coeff. Coeff. Coeff. Coeff. Coeff.
Variable (t-stat) (t-stat) (t-stat) (t-stat) (t-stat) (t-stat)
Log HH income 0.261 0.257 0.052 0.325 0.320 0.109
(12.26) (12.12) (4.39) (9.77) (9.63) (5.32)
Log financial assets 0.068 0.070 0.079 0.079
(13.67) (14.23) (10.18) (10.30)
Log home value 0.025 0.024 0.019 0.015
(5.27) (5.21) (2.59) (2.15)
Log credit card debt -0.034 -0.034 -0.042 -0.042
(7.09) (7.11) (6.11) (6.22)
Log home debt -0.004 -0.014
(0.83) (2.00)
Log other debt -0.052 -0.039
(11.22) (5.85)
Log mortgage payments -0.007 -0.015
(0.91) (1.33)
Log other debt payments -0.051 -0.039
(11.13) (5.98)
Log net wealth 0.048 0.052
(8.19) (5.20)
Self-rated health 0.351 0.354 0.383 0.358 0.352 0.394
(17.05) (17.38) (18.10) (11.33) (11.27) (12.22)
Children in HH -0.078 -0.078 -0.096 -0.067 -0.062 -0.089
(4.76) (4.84) (5.51) (2.63) (2.49) (3.35)
Other adults in HH -0.109 -0.113 -0.087 -0.071 -0.067 -0.067
(4.17) (4.35) (3.33) (1.79) (1.70) (1.66)
Marital status variables yes yes yes yes yes yes
Unemployed -0.956 -0.972 -1.161 -0.748 -0.781 -1.011
(6.82) (6.99) (7.63) (3.61) (3.81) (4.65)
Retired 0.102 0.100 0.073 0.169 0.178 0.124
(1.63) (1.60) (1.22) (1.57) (1.67) (1.19)
Black 0.000 -0.009 -0.309 -0.025 -0.008 -0.147
(0.01) (0.15) (4.80) (0.32) (0.11) (1.79)
Male -0.102 -0.109 -0.064 -0.050 -0.068 -0.062
(2.97) (3.21) (1.84) (0.95) (1.30) (1.16)
More than HS education -0.198 -0.197 -0.012 -0.213 -0.212 -0.037
(5.37) (5.36) (0.33) (3.90) (3.92) (0.69)
Age (centered) 0.009 0.009 0.014 0.007 0.008 0.012
(5.35) (5.51) (8.22) (1.90) (2.06) (3.06)
Age centered, squared 0.000 0.000 0.000 0.001 0.001 0.001
(5.41) (5.23) (3.82) (1.90) (2.03) (2.43)
Constant 0.419 0.448 2.607 -0.356 -0.258 1.845
(1.90) (2.03) (17.31) (1.02) (0.74) (7.31)
Observations 7,517 7,648 7,378 3,295 3,360 3,272
Adjusted R
2
0.2484 0.2475 0.1492 0.2559 0.2497 0.1512
55
owed or the monthly payments that have to be made to service that debt. The coefficients for
both types of measures are fairly similar (Table 3.6, columns 1 and 2). Credit card debt, which
can be regarded as a “bad debt” compared to mortgage debt (Mishel et al., 2005), has a stronger
negativeimpactonanindividual’sevaluationofhisfinancialsituationthandebtonhomes. Other
forms of debt also display a strong negative impact.
An aggregate measure of net worth displays a large impact on the dependent variable (Table
3.6, column 3), but not surprisingly also reduces the percentage of the variance of the dependent
variable that is explained by the model. A composite measure of net wealth is only appropriate
if it does not lead to a large loss of information. This would be true if the coefficients of the
separate variables that are aggregated in the composite measure were not statistically different
from each other. A F-test reveals that the coefficients are all statistically different from each
other; hence aggregating the asset and debt variables into one single measure of net wealth (or
separate aggregate measures of assets and debts) leads to a considerable loss of information.
5
The coefficients of the remaining explanatory variables are all quite similar in all three model
specifications. Self-rated health, which can be seen as a proxy for the costs of health care, not
surprisingly has a large positive impact on financial satisfaction, whereas the presence of addi-
tional household members indicates increased costs and thus displays a negative coefficient. The
possibility that other household members might earn income is already reflected in the household
income measure; hence more household members indicate greater expenditures. People who are
unemployed have lower levels of financial satisfaction, all other things equal, than those who are
employed or are not in the labor force. Taking into account that the regressions hold economic
variables constant, this results suggests that unemployed individuals are less satisfied with their
current financial situation because their income aspirations are higher than those of people who
engage in paid employment and have the same level of income. These individuals likely know
5
For instance, a F-test testing the hypothesis that financial assets and tangible assets have equal coefficients
and could thus be combined into one measure is rejected with F(1, 7496) = 32.37.
56
that their level of income will be higher once they find a new job and their income aspirations are
formed relative to their potential income instead of their current income. Race only has a signifi-
cant impact in the last specification. Male respondents and those with an education beyond high
school seem to have higher income aspirations; hence they are less satisfied with their financial
situation, ceteris paribus. As reflected in the life cycle profiles, financial satisfaction continuously
increases with age. The life cycle profiles only included controls for individual fixed effects and
period effects, but the regression analysis including measures of assets and liabilities still shows a
significant positive effect of age. I will attempt an explanation for this result later in the chapter.
The results suggest that individuals do not assess their financial situation objectively and solely
basedoneconomiccircumstances. Iftheydid,thereshouldbenogenderoreducationaldifferences
in the evaluation of one’s financial situation. The results are very similar if I use an ordered logit
specification in order to account for the ordinality of the dependent variable (Table 3.7), and are
thus robust to methodology.
The wave 2 NSFH sample is most representative of the overall population and unlike the wave
3 sample includes sample weights. But do the results differ significantly if the same weighted
least square and ordered logit regressions are run on the subset of observations in wave 2 that
is included in the balanced panel described above? The results are actually remarkably similar
(Tables 3.6 and 3.7, cols. 4-6). Due to the considerably smaller sample size and therefore larger
standard errors, the t-values are smaller in the second set of regressions employing the restricted
sample, but the signs of the coefficients are the same.
The cross-sectional analysis shows interesting results, but as all cross-sectional analyses it
suffers from a considerable shortcoming. The age differences in financial satisfaction that one
observesinthecross-sectionmightinfactbebirthcohortdifferences. Itispossiblethatdifferences
in the experience of birth cohorts lead individuals to assess their financial situations differently.
Forinstance,someonewhoexperiencedperiodsofeconomicdepressionmightassessthesamelevel
of income, assets and debts more positively than someone who grew up in a period of relative
57
Table 3.7: Ordered logit regression on financial satisfaction, wave 2 (dollar amounts in $1993)
Full wave 2 sample Restricted sample
(1) (2) (3) (4) (5) (6)
Coeff. Coeff. Coeff. Coeff. Coeff. Coeff.
Variable (z-stat) (z-stat) (z-stat) (z-stat) (z-stat) (z-stat)
Log HH income 0.419 0.410 0.069 0.548 0.531 0.145
(12.53) (12.38) (4.43) (10.14) (9.95) (5.10)
Log financial assets 0.079 0.082 0.094 0.094
(12.05) (12.65) (8.99) (9.16)
Log home value 0.032 0.031 0.023 0.019
(5.19) (5.21) (2.41) (2.11)
Log credit card debt -0.047 -0.047 -0.058 -0.058
(7.97) (7.94) (6.60) (6.70)
Log home debt -0.009 -0.019
(1.58) (2.23)
Log other debt -0.062 -0.050
(10.84) (5.99)
Log mortgage payments -0.017 -0.023
(1.76) (1.68)
Log other debt payments -0.061 -0.050
(10.76) (6.03)
Log net wealth 0.060 0.069
(8.05) (5.39)
Self-rated health 0.442 0.447 0.478 0.455 0.448 0.494
(16.52) (16.85) (17.50) (10.99) (10.95) (11.86)
Children in HH -0.096 -0.095 -0.115 -0.087 -0.079 -0.108
(4.69) (4.73) (5.42) (2.70) (2.50) (3.27)
Other adults in HH -0.163 -0.167 -0.125 -0.103 -0.098 -0.094
(4.95) (5.15) (3.89) (2.04) (1.96) (1.89)
Marital status variables yes yes yes yes yes yes
Unemployed -1.103 -1.129 -1.304 -0.871 -0.929 -1.242
(6.38) (6.58) (7.19) (3.08) (3.34) (4.34)
Retired 0.170 0.165 0.113 0.291 0.301 0.228
(2.12) (2.08) (1.50) (2.10) (2.19) (1.73)
Black 0.019 0.012 -0.349 0.004 0.027 -0.145
(0.26) (0.16) (4.35) (0.04) (0.28) (1.39)
Male -0.152 -0.162 -0.103 -0.100 -0.122 -0.104
(3.56) (3.84) (2.41) (1.51) (1.86) (1.58)
More than HS education -0.277 -0.271 -0.030 -0.296 -0.286 -0.047
(5.99) (5.91) (0.68) (4.25) (4.15) (0.71)
Age (centered) 0.013 0.013 0.019 0.010 0.011 0.017
(5.75) (5.95) (8.71) (2.09) (2.29) (3.52)
Age centered, squared 0.001 0.001 0.000 0.001 0.001 0.001
(6.14) (5.96) (4.72) (2.24) (2.32) (2.24)
Observations 7,518 7,648 7,378 3,295 3,360 3,272
Pseudo R
2
0.0794 0.0791 0.0458 0.0840 0.0816 0.0472
Chi
2
2,181 2,209 1,204 1,008 997 550
Log likelihood -12,635 -12,859 -12,533 -5,496 -5,612 -5,554
58
affluence because his relative standards are considerably different. The cross-section results do
not reveal whether the individuals who we observe at young ages now will eventually have higher
levelsoffinancialsatisfactionastheybecomeolder. Similarly, theolderrespondentsinthesample
might already have had high levels of financial satisfaction when they were younger.
I therefore now turn to an analysis of the second sample described above which is a balanced
panel sample with 3,751 observations in each of the two survey waves (waves 2 and 3 of the
NSFH). I use a balanced panel instead of an unbalanced panel including all available observations
in both waves to avoid selection bias. If, for instance, people who are generally more satisfied
live longer than other people one will most likely observe an increase in satisfaction in the data
because the first wave included some individuals who are generally not satisfied and passed away
before the second wave. I therefore create a balanced panel and look at changes that occur within
an individual. The panel structure of the data allows me to do a fixed-effects regression in which
individual time-invariant characteristics such as birth cohort, gender and race are controlled for.
Thefixed-effectsanalysessuggestthatwithinanindividualhigherincome, financialassetsand
tangible assets lead to increases in financial satisfaction while credit card debt, mortgage debt
and other types of debt have a negative impact. As in the cross-section, a model with separate
measures for assets and liabilities works better than one with an aggregate measure for net worth
(Table 3.8), but the results for models with measures of absolute debt or debt payments are quite
similar. The coefficients for the remaining variables are similar to the coefficients in the cross-
section. Age does not seem to have a significant impact on financial satisfaction, but as described
above, the coefficient of age might be significant in a larger sample.
Fixed-effectsregressionsarebasedontheassumptionthatthemeasureoffinancialsatisfaction
is cardinal – i.e. a satisfaction level of 4 is considered to be twice as good as a satisfaction level of
2. An ordered probit specification with random effects as suggested by Ferrer-i-Carbonell (2005)
does not rely on this strong assumption. In this specification, the financial satisfaction measure
is assumed to be ordinal – i.e. an evaluation of 4 is better than an evaluation of 2, but no
59
Table 3.8: Fixed-effects and ordered probit regressions on financial satisfaction
ordered probit
fixed effects with random effects
Variable (t-stat) (z-stat)
(1) (2) (3) (4) (5) (6)
Log HH income 0.158 0.188 0.032 0.234 0.236 0.264
(3.89) (5.29) (1.70) (9.94) (9.91) (10.98)
Log financial assets 0.071 0.065 0.064 0.064
(6.63) (6.98) (11.16) (11.06)
Log home value 0.025 0.011 0.025 0.025
(2.67) (1.33) (4.52) (4.42)
Log credit card debt -0.028 -0.026 -0.037 -0.038
(3.45) (3.39) (7.43) (7.45)
Log home debt -0.013 -0.021
(1.90) (4.80)
Log other debt -0.024 -0.036
(3.45) (7.94)
Log mortgage payments -0.013 -0.034
(1.11) (4.63)
Log other debt payments -0.018 -0.037
(2.70) (7.91)
Log net wealth 0.018 0.065
(1.80) (6.84)
Self-rated health 0.115 0.155 0.137 0.279 0.281 0.303
(3.00) (4.44) (4.14) (12.47) (12.45) (12.85)
Children in HH 0.007 -0.006 -0.015 -0.078 -0.078 -0.106
(0.20) (0.17) (0.47) (3.98) (3.94) (4.98)
Other adults in HH -0.043 -0.062 -0.026 -0.065 -0.070 -0.121
(1.05) (1.60) (0.75) (2.43) (2.58) (4.31)
Marital status variables yes yes yes yes yes yes
Unemployed -0.729 -0.533 -0.728 -0.685 -0.664 -0.706
(3.52) (2.78) (4.01) (4.88) (4.64) (4.61)
Retired 0.135 0.065 0.043 0.160 0.164 0.218
(1.51) (0.77) (0.58) (2.53) (2.56) (3.37)
Age (centered) 0.027 0.013 0.029 0.006 0.006 0.010
(1.23) (0.66) (1.60) (1.89) (2.00) (3.05)
Age centered, squared 0.000 0.000 0.000 0.000 0.000 0.001
(0.48) (1.20) (0.68) (3.29) (3.30) (4.22)
Unemployment -0.064 -0.141 -0.102 -0.153 -0.158 -0.162
(0.89) (2.15) (1.65) (7.29) (7.42) (7.50)
Inflation -0.145 -0.181 -0.110 -0.127 -0.131 -0.115
(1.49) (2.03) (1.30) (3.64) (3.72) (3.22)
Constant 2.967 3.239 4.825 0.075 0.042 0.655
(3.37) (4.08) (7.02) (0.26) (0.15) (2.19)
within-R
2
0.1966 0.1857 0.1191
between-R
2
0.2726 0.2794 0.1224
overall-R
2
0.2554 0.2556 0.1213
LR χ 2
1,750 1,752 1,208
Log likelihood -8,951 -8,860 -8,205
Number of obs 5,683 6,186 6,708 5,683 5,632 5,191
Number of groups 3,538 3,598 3,623
60
assumptions are being made about the magnitude of the difference between these two levels. The
results of the ordered probit regressions are similar to the results of the fixed effects specification,
with only a few exceptions. The results are thus robust to methodology.
All variables other than income and self-rated health help explain increases in satisfaction at
olderages. Thevariableswhichdisplayanegativeeffectonfinancialsatisfaction,suchasdebtand
household size all decrease with age. Except for income and health, the variables which have a
positiveimpact,namelyassets,increaseoverthelifecycle. Itisthuspossiblethatchangesinassets
and liabilities as well as family circumstances account for the increase in financial satisfaction,
butwhenwelookatthe large sizeofthecoefficientofincome inthe fixed-effects regression(Table
3.8) it seems unlikely that these variables are sufficient to fully explain the life cycle profile of
financialsatisfaction. Likely,thereareotherfactors,suchasdecreasesinaspirations,whicharenot
captured in this model and are important determinants of financial satisfaction. The remaining
positive,significanteffectofageinsomeofthemodelspecificationsindicatesthattherearefactors
– other than the economic factors included in this model – that influence financial satisfaction
substantially and change with age. It is possible that financial aspirations decrease with age.
One of the two theories put forward by George (1993) to explain high financial satisfaction at
old age suggests that older people perceive their compensation to be more equitable than younger
people. Inthisview,thepsychologicalmechanismsthatinfluenceone’sevaluationofone’sfinancial
situation work differently at old age (Liang et al., 1980). For the non-elderly relative economic
statusisanimportantdeterminantoffinancialsatisfactionwhereastheelderlyaremoreconcerned
about income adequacy and distributive justice. The influence of social comparison seems to fade
withageanddiminishesfinancialaspirations. Itisalsolikelythatfinancialneedsthatarestrongly
tied in with family formation and dissolution are not adequately captured by the model.
61
3.6 Discussion
The purpose of the current analysis is to analyze the determinants of financial satisfaction and
assess to what extent changes over the life cycle might be driven by factors other than income. If
indeed other factors prove to be more important than income, this would offer an explanation for
the seemingly paradoxical observation that people at old age have higher satisfaction levels than
younger people despite lower household incomes. Traditional cross-section analysis usually shows
the positive impact of age on financial satisfaction, but the cross-section results do not allow us to
evaluate whether high levels of satisfaction at old age are only due to differences in the experience
of birth cohorts or if indeed the same individuals experience higher levels of satisfaction as they
age. The analysis of a 10-year panel shows that individuals indeed enjoy higher levels of financial
satisfaction at old age. The life cycle pattern on income, with declining income after a peak
in midlife, taken alone would suggest downward-sloping financial satisfaction with age, as would
changes in self-rated health.
On the other hand, the life cycle profiles of assets and debts as well as the presence of other
household members suggest an upward-sloping pattern of financial satisfaction, which one can
indeed observe. I hypothesized that debt can be seen as an indicator of aspirations that exceed
financial means. Although material aspirations seem to continue to increase with age for some
assets, as indicated by the increasing values of tangible assets, financial means seem to increase at
a steeper rate; hence the discrepancy between material aspirations and financial means decreases
and the individual does not have to incur more debt. The acquisition of debt is of course also
regulated by the willingness of the supply side to grant a loan or issue a credit card, which is
highly dependent on the individual’s overall financial situation. Declining levels of debt could
also suggest a downward adjustment of aspirations as observed by George (1992), but the life
cycle analysis does not provide enough evidence for such an assessment, though the significant
62
coefficient on age suggests that there are age-related economic or psychological changes that are
not captured by my model.
The current analysis suggests that increases in assets and decreases in debt contribute sub-
stantially to the life cycle pattern of financial satisfaction. Decreases in debt suggest a decreasing
discrepancy between financial aspirations and financial means which could either be caused by a
downward adjustment of aspirations at old age or a steeper increase in financial means. Decreases
in the dependency burden also suggest fewer financial obligations, whereas decreases in health
status indicate increased health care costs. The distinction between financial aspirations and fi-
nancial needs is not always clear because people’s needs might change objectively – e.g. through
the birth of a child or the enrollment of a child in school – as well as subjectively by ascending to
a higher income class which comes with a new set of perceived needs, such as enrolling a child in
a private school or paying for piano lessons. A more detailed analysis could explore the presence
of threshold effects related to income classes. Although it could be argued that, for instance, the
middle class’ needs objectively increase, these increased needs are in fact often a result of social
comparison. A clear distinction between actual needs and aspirations is difficult because needs
are partly determined by life cycle processes, such as family formation and dissolution, but also
by social processes which increase aspirations. In many cases, differentiating between actual and
perceived needs is rather arbitrary.
A measure of net wealth proved to be inappropriate for the analysis for two reasons. First,
the influence of the separate measures of assets and liabilities differs substantially; hence an
aggregation of these measures into one measure would lead to a significant loss of information.
Second, the same amount of net wealth can represent substantially different compositions of
assets and liabilities. For instance, an individual with high debt and many assets probably has
a significantly different level of financial aspirations than an individual with low debt and few
assets. Thisinformationwouldbeignoredwhenusinganetwealthmeasure. Thepresentanalysis
emphasizes the need to employ measures of wealth in an analysis of financial satisfaction, instead
63
of relying on income as a proxy for wealth. The results further suggest that the direct impact of
income on financial satisfaction is mediated by financial aspirations. Changes in aspirations can
be caused by psychological processes, such as social comparison and hedonic adaptation.
Thenextchapterfurtherinvestigateschangesinaspirationsinseveraldomainsoflife,including
the financial domain.
64
Chapter 4
Chasing the Good Life: Life Cycle Aspirations and
Attainments
4.1 Introduction
Most people have a fairly good idea of what they consider to be the good life – i.e. the life
they wish to have. One’s concept of the good life might change over the life cycle but people
usually steadily pursue their goals, changing or not, and they assess their happiness according to
how well their attainments match their goals. Thus, attainments may not lead to increases in
well-being if they are accompanied by increasing aspirations. This chapter examines aspirations
and attainments for goods, family, work and health over the life cycle. A distinctive feature of the
presentanalysisisnewevidenceonaspirationsinnonpecuniaryaswellaspecuniarydomains, and
how both pecuniary and nonpecuniary aspirations change over the life cycle. Very few researchers
haveattemptedtomeasureaspirationsdirectly, withthenotableexceptionofStutzer(2004), who
estimates income aspirations in Switzerland, but not aspirations in nonpecuniary domains. The
contribution of the current study is mostly descriptive. To my knowledge, this is the first fairly
comprehensive study to show how aspirations change over the life cycle, and how far individuals
realize their aspirations in different domains of life.
65
If,assomeassert(Easterlin,2003;Frank,1997),pecuniaryaspirationsaremoreinfluencedthan
nonpecuniary by hedonic adaptation and social comparison, then one should find that aspirations
forgoodsincreasemoreoverthelifecoursethandoaspirationsinthenonpecuniarydomains. The
present findings support this hypothesis. Goods aspirations rise noticeably over the life course,
but aspirations for a happy marriage decline somewhat, as do job aspirations. Health aspirations
increase only slightly.
People’sperceptionsofthegoodlifeencompassawidevarietyofthings,anddifferconsiderably
among individuals. However, to judge from Hadley Cantril’s (1965) surveys in 14 countries, rich
and poor, on the “best” and “worst” of all possible worlds, the four domains covered here are the
most important of people’s concerns (cf. also Crimmins and Easterlin, 2000).
Conceptually,the“goodlife”encompassesthevarietyofcircumstancesthatentertheeconomist’s
utilityfunction. Thesefactorsarenormallyassumedtoencompassmaterialgoodsofvarioustypes,
family circumstances, work, health, and so on, and an individual’s well-being is taken to depend
on the extent to which attainments in these various domains match one’s conception of the good
life. Empirical evidence on the good life, however, is notoriously lacking. The purpose of the
present analysis is to present evidence on the nature of the good life and how this conception
typically changes over the life cycle. The shortfall between aspirations and attainments in each
domain allows a tentative assessment of the domain’s contribution to general well-being over the
life cycle.
Ifanindividual’sconceptionofthegoodlifewerefixed,thenwell-beingwoulddependsolelyon
attainments. Therearestrandsofeconomictheorygoingbackover50years,however,thatsuggest
that attainments have a feedback effect on what people feel is important in life. One emphasizes
theimportanceofpositionalconcernsorwhatpsychologiststermsocialcomparison(Duesenberry,
1949; Pollak, 1970; Hirsch, 1976; Frank, 1985); the other, the impact of habit formation, or what
psychologistscallhabituation(Pollak,1976;Modigliani,1949). Iftheseinfluencesoperatedequally
across all domains of life and constituents of these domains, then they would reduce a person’s
66
sense of well-being without altering the allocation of time and money (cf. Solnick and Hemenway,
2005). There is ample reason to believe, however, that social comparison and habituation vary
across and within domains. Positional concerns, for example, are more likely to apply to material
goods – especially those that are highly visible – more than to family life and health where the
real nature of one’s life is less available to public scrutiny. While it is certainly possible that
people compare their family life to that of others, one’s group of reference is undoubtedly larger
when evaluating one’s home, car and other material goods relative to the goods of others. Most
individuals have only a small circle of friends and family members whose family life, work and
health they could assess adequately. One can therefore assume that social comparison is more
prevalent in the pecuniary domain, which is more exposed to the scrutiny of others. Indeed,
luxury items are more “visible” than other goods, such as insurances (Heffetz, 2006)
Similarly, habituation may be more important in regard to goods than, say, to health and
physical condition. People get accustomed to their new cars, but not, it seems, to the new shapes
and faces made possible by modern cosmetic “technology” (Frederick and Loewenstein, 1999).
Similarly,socialcomparisonandhabituationdonotoperateequallywithregardtoallconstituents
of a given domain. With regard to the goods domain, Scitovsky (1976) has argued that cultural
goods, such as music, literature, and art, are less subject to hedonic adaptation than “comfort”
goods like homes and cars. Similarly, the distinction drawn between positional and nonpositional
goods by Frank (1985); Hirsch (1976); Ng (1978), and others is an example of a classification of
goods based on whether or not their utility is affected by social comparison. Clearly, if the forces
of social comparison and habituation vary across and within domains, individuals’ allocation of
time and money may not be optimal. For informed individuals to be able to make decisions,
they need to be aware of social comparison and habituation, and the role these forces play in the
various domains of life. This chapter represents a step in this direction, presenting evidence on
how aspirations in regard to material goods, family life, work, and health change over the life
cycle.
67
Thepresentstudyfindsthatinthematerialgoodsdomainaspirationsforluxuryitemsincrease
more substantially than aspirations for goods that one might classify as necessities. The goods
which are categorized as luxuries, such as a vacation home and a swimming pool, could also be
classified as positional goods, i.e. goods that do not only provide utility through their intrinsic
value, but also through the relative position these goods provide the consumer in society. The
acquisition of positional goods is sometimes denoted as conspicuous consumption (Veblen, 1899).
As incomes increase, conspicuous consumption forms a larger share of the consumer’s budget. In
terms of income elasticities, luxury goods are defined as goods with an income elasticity greater
than one – i.e. the demand for the good rises more than proportionate to the increase in income.
Generally, the income elasticity measures in how far the quantity demanded of a good changes
relative to a change in income. Income elasticities are calculated using the following equation:
Income elasticity=
% change in quantity demanded
% change in income
A good with an income elasticity between 0 and 1 is denoted as a necessity because demand rises
less than proportionally to income. The present analysis blends well into this line of thinking.
People usually experience an increase in income with age. Thus, aspirations for luxury items rise
considerably with age. Aspirations for necessities, on the other hand, increase at a smaller rate,
which might be due to the fact that individuals acquire necessities before they acquire luxuries.
InwhatfollowsIdescribethedataandmethodology, andreportmyfindingsfortheindividual
domains. The last section concludes the analysis.
4.2 Data and methods
ThedataarefromnineRopersurveysconductedbetween1978and2003(Roper-StarchOrganiza-
tion,1978-1995). Roper-Starchisaprivateorganizationthatconductssurveysonamonthlybasis
withvaryingquestionsonalargenumberoftopics. Thegoodlifequestionsusedhereareincluded
68
in a survey about every three years.
1
Every survey contains about 2,000 observations with a total
of 17,821 for all nine surveys. The data are not panel data but nationally representative surveys
conducted at successive dates.
The surveys include questions in which the respondents are given a list and asked to identify
items that they consider to be part of the good life (PGL), i.e. the life they personally wish to
have. These items include material goods, such as a lot of money, a home, a car, a swimming
pool, and a vacation home, as well as items from the family, job and health domain, such as a
happy marriage and an interesting job (see Appendix C.1). The wording of the good life question
is the same in every survey year. The respondents are further asked to indicate which items they
already own. It is of course possible that a respondent indicates that he owns an item, which he
doesnotconsidertobepartofthegoodlife. InthepresentanalysisIamparticularlyinterestedin
the shortfall between an individual’s aspirations and attainments. Someone who does not aspire
to a certain item cannot experience a shortfall between his aspirations and attainments – whether
he owns the item or not. I therefore set the binary attainment variable (denoted as “have” in the
tables) to zero if the respondent indicates that the item in question is not part of his conception
of the good life. The items used here are included in all nine survey years with five exceptions –
swimming pool, which is omitted in 1996 and 1999, TV, 2 TVs and children, which are omitted
in 2003, and good health, which is only included in the most recent survey in 2003. Table 4.1
shows the percentage of people that aspire to and own certain items at all ages.
Life cycle patterns are obtained by regressing the dependent variable (here aspirations and
attainments for a particular item) on age with controls for year of birth (birth cohort), race,
education, survey year dummies, gender, as well as the squared value of age and birth cohort
(Table 4.2 gives descriptive statistics for these variables). For the regressions, the values for age
1
The corresponding Roper Reports are published in January of the year following the survey and are therefore
named Roper Report 1-79 etc. As the surveys were conducted in the winter months of the year preceding pub-
lication, I use the year in which the surveys were conducted to classify them. I did not include the 1975 survey
because some of the survey questions were slightly different in that year. The 2003 survey is an exception, it was
conducted in March of that year.
69
Table 4.1: Descriptive Statistics: Mean percentage of people stating that an item is part of the
good life and mean percentage stating that they have the item, 1978− 2003
a
Considers item to be
part of the good life Has the item
Goods:
A home you own 87.52 52.90
Yard 61.27 40.58
TV 55.70 48.43
Car 77.41 65.98
2 cars 40.51 23.71
2 TVs 25.80 14.77
Nice clothes 41.13 15.13
Travel abroad 40.02 9.40
Swimming pool 27.69 2.41
Vacation Home 36.32 2.94
A lot of money 54.85 2.51
Family:
Happy Marriage 78.77 47.47
Children 68.91 41.82
Children’s college education 57.90 12.43
Job:
Interesting job 61.71 26.79
Job that pays well 56.35 13.65
Job that contributes to welfare of society 34.57 11.37
Health:
Good health 87.0 64.81
a
number of observations = 17,825, except for TV, 2 TVs and Children n = 15,821
Swimming pool n = 13,837 and for Good health n = 2,004.
and cohort are centered around the sample mean. Many studies have shown that happiness or
domain satisfaction varies by gender, race and education (e.g. Frey and Stutzer, 2002). Because
the demographic composition varies across ages and cohorts in the sample, I control for these
variables throughout the analysis. I use a logit specification because the aspiration variables are
binary. The same regressions were run using ordinary least squares with quite similar results,
suggesting that the findings are robust with regard to methodology.
Life cycle patterns for aspirations and attainments in each domain are obtained by computing
from the regression the value of each good life item at each age from 19 to 78, using the mean
70
Table 4.2: Descriptive Statistics, 1978− 2003
Number of Standard
Variable observations Mean deviation Min. Max.
Age 17,820 44.01 17.44 19 78
Birth year 17,820 1946.48 18.82 1905 1984
Male = 1 17,824 0.47 0.50 0 1
Black = 1 17,698 0.12 0.32 0 1
Education more than 12 years = 1 17,765 0.44 0.50 0 1
values for all independent variables other than age. The life cycle pattern for the health variable
does not include controls for birth cohort, because the item is only included in the 2003 survey.
Thelifecyclepatternspresentedherearebestviewedasroughapproximations,becausetheage
responses in the survey are categorical, typically starting with 18-21 and 22-24, then progressing
by 5 year groups through 60-64, and ending with 65+. I have assigned each individual the
midpoint value of the individual age category; hence the number of age responses in a survey is
typically 11 or 12 (The cohort observations, derived as the difference between survey year and
age, are therefore similarly limited in number). For the open-ended 65+ category I have used the
mean value (73) of persons aged 65 and over reported in the General Social Surveys (Davis et al.,
2005). In the surveys of 1991, 1994, 1996, 1999 and 2003 the open-ended category is 70+, and I
have used a mean value of 78, similarly obtained. This severe averaging of observations for older
persons means that the trajectory of aspirations in old age cannot really be accurately identified
in these data.
4.3 Findings
I expect that aspirations in the material goods domain continue to increase over the life cycle,
whereas nonpecuniary aspirations should remain fairly constant or decline. Within the material
goods domain, I expect that aspirations for goods that could be classified as luxuries rise more
71
with age than aspirations for “essentials”.
2
In the following I present life cycle curves showing
the aspirations and attainments of a subsample of the items described above.
4.3.1 Material Goods
4.3.1.1 Necessities
Aspirations for necessities, such as a home and a car, start out at a high level and increase
somewhat over the life cycle (Figures 4.1 and 4.2 ). Even at young ages more than 80% of the
respondents consider a home to be part of the good life. Similarly, almost 70% of 20-year olds
consider a car to be part of the good life. Attainments for both goods increase steadily over the
life cycle with a slight decline at old age. A visual comparison of the aspirations and attainment
curvessuggeststhatfornecessitiestheshortfallbetweenthetwodecreasesoverthelifecycle. This
result is quite intuitive because individuals likely invest first in goods that are of more importance
in daily life than luxury items. Increasing attainments are somewhat offset by rising aspirations,
butthedecreasingdiscrepancybetweenthetwosuggeststhatacquiringthesegoodsmightimprove
well-being somewhat.
4.3.1.2 Luxuries
In contrast, aspirations for luxury items, such as a swimming pool, travel abroad and a vacation
homestartatratherlowlevels, butincreasemoresubstantiallyoverthelifecycle(Figures4.3, 4.4
and 4.5). This finding is especially interesting, because only a few people actually acquire these
luxury items at some point during their life – the percentage that actually has a swimming pool
or vacation home averages about 4 percent. Thus, although aspirations for these luxury items
increaseconsiderablyoverthelifecycle, toaboutone-halfofthepopulation, theseaspirationswill
2
When I refer to necessities or luxuries, I do not strictly adhere to the classic definition in economics which is
based on income elasticities. I rather classify goods as necessities that are more or less essential and as luxuries
goods that seem to be more positional and convey an individual’s status in society. These goods most likely fit the
income elasticities definition, but I have not formally tested whether this is true.
72
Figure 4.1: Aspirations and attainments: A home you own, ages 19-78
Figure 4.2: Aspirations and attainments: A car, ages 19-78
remain unfulfilled for virtually all of them. The shortfall between aspirations and attainments
increases substantially over the life cycle, suggesting that higher aspirations for luxury items
undermine the pursuit of happiness in the material goods domain.
73
Figure 4.3: Aspirations and attainments: A swimming pool, ages 19-78
Figure 4.4: Aspirations and attainments: Travel abroad, ages 19-78
4.3.1.3 Lots of money
Aspirationsforhavingalotofmoneyincreaseconsiderablyoverthelifecyclewhereasattainments
only increase somewhat at older ages (Figure 4.6). Money can be considered as a representative
good for all material goods items because people typically acquire money in order to purchase
goods. The life cycle pattern of money aspirations resembles the pattern one can observe for
74
Figure 4.5: Aspirations and attainments: A vacation home, ages 19-78
Figure 4.6: Aspirations and attainments: A lot of money, ages 19-78
luxury items. Even though incomes are rising during the working ages people’s aspirations for
money do not diminish, but increase.
Goods aspirations, it would seem, are not satisfied by the accumulation of consumer wealth
made possible by rising income. Rather, aspirations for goods rise along with the growth of con-
sumer wealth. Whereas the shortfall between aspirations and attainments declines for necessities
75
with age, it steadily increases for luxuries. But in the family, work, and health domains, to which
I now turn, the picture is different.
4.3.2 Family
4.3.2.1 A happy marriage
Aspirations for having a happy marriage start at a high level – well over 80% for both genders
– and decrease substantially with age (Figure 4.7). Marriage aspirations are initially very high,
muchlikethoseforahomeandacar, andthethreetogetherwithhealthrankatthetopofthelist
(Table 4.1). The substantial decline in aspirations for a happy marriage over the life cycle could
possibly be explained by the increasing percentage of persons divorced and widowed, which is
reflected in the declining attainment curves. Individuals who do not get married before the age of
45 experience the largest decline in aspirations for a happy marriage, but even for those persons,
who have been single their entire lives, the percentage who still consider a happy marriage as part
of the good life as they are personally concerned, is surprisingly high (Table 4.3).
It could be argued that aspirations in the nonpecuniary domain are not comparable to pecu-
niary aspirations because there is a natural limit to how many happy marriages a person can have
atapointintime. Moreover,aspirationsforahappymarriagehaveahighlynormativecomponent.
Although a marriage is less exposed to public scrutiny than, say, a car, social comparison could
still play an important role for marital aspirations because people feel the pressure of societal ex-
pectations. Even if social comparison has an impact in the nonpecuniary domain, an individual’s
relevant group of reference in the pecuniary domain is doubtless broader and comparisons in such
a wide network are more compelling. The next item in the family domain, children, illustrates
better the difference in aspirations between the pecuniary and the non-pecuniary domain because
the demand for children is often compared to the demand for consumer durables.
76
Figure 4.7: Aspirations and attainments: A happy marriage, ages 19-78
Table 4.3: Mean percentage of people stating that a happy marriage is part of the good life by
marital status, 1978− 2003
ages ages
Marital status all ages 19-42 43 and over
married 86.1 86.1 86.2
divorced or separated 65.8 69.0 62.0
widowed 63.6 67.1 63.4
never-married 69.3 72.1 47.8
4.3.2.2 Children
Economists sometimes draw an analogy between the demand for children and the demand for
consumerdurables. Aspirationsformaterialgoodsallincreaseoverthelifecycleaspeopleacquire
more goods with rising incomes, but aspirations for having one or more children decline slightly
(Figure 4.8). As can be expected, the shortfall between wanting and having children continuously
decreases with age as people start to form families. This domain should therefore be a source of
increasing satisfaction.
Aspirationsconsideringachild’scollegeeducationshowconcernsaboutthequalityofchildren.
The desire to provide one’s children with a college education increases with age (Figure 4.9). Not
77
surprisingly attainment of this item is strongly related to age because individuals in their 20s
or 30s will rarely have children that are old enough to attend college. The shortfall between
aspirations and attainments also decreases here.
Hence,bythesemeasuresaspirationsforchildrenarenotcomparabletoaspirationsformaterial
goods. Although actual family size usually changes in the early life cycle, aspirations for the
“quantity” or “quality” children do not increase commensurately.
Figure 4.8: Aspirations and attainments: One or more children, ages 19-78
The gap between aspirations and attainments for a happy marriage declines considerably in
the 40s and subsequently increases somewhat. Similarly, attainments for having one or more
children approach aspirations closely in the 50s, followed by a widening gap between aspirations
and attainments. These patterns indicate that satisfaction with one’s family life increases in
midlife and declines thereafter. Easterlin’s research on satisfaction in various domains over the
life cycle confirms this pattern for satisfaction in the family domain (Easterlin, 2006).
78
Figure 4.9: Aspirations and attainments: Children’s college education, ages 19-78
Figure 4.10: Aspirations and attainments: An interesting job, ages 19-52
4.3.3 Work
In the work domain respondents were asked whether they considered an interesting job to be part
of the good life. Aspirations for an interesting job start at a high level but decline through the
working ages (Figure 4.10). I tried to exclude retired persons from the analysis by restricting the
sample to individuals between age 19 and 52. It is possible that people regard their jobs as being
79
less interesting as they are becoming more accustomed to them in the course of the life span. The
shortfall between aspirations and attainments is fairly constant.
4.3.4 Health
Aspirations in the health domain start at a high level and increase somewhat with age (Figure
4.11). Even at young ages, when most individuals do not experience health problems, more than
80% of respondents consider good health to be part of the good life. Life cycle aspirations and
attainments for good health are derived from only 2,004 observations instead of the usual 17,821,
becausethisitemisonlyincludedinthemostrecentsurvey,conductedin2003. Thus,thelifecycle
patternshownhereforagemaypartlyreflecttheeffectofbirthcohort. Theincreasingdiscrepancy
between aspirations and attainments suggests decreasing satisfaction with one’s health over the
life cycle. Indeed, Easterlin finds that health satisfaction declines with age (Easterlin, 2006).
Figure 4.11: Aspirations and attainments: Health, ages 19-78, survey year 2003
Note: The life cycle pattern for the health variable only includes controls for age and age
2
, but not for
birth cohort, because the item is only included in the 2003 survey.
80
4.3.5 Relation of aspirations and attainments to demographic variables
other than age
The regressions on which the previous curves are based include controls for birth cohort, survey
year dummies, race and education in order to reflect the pure effect of age as well as possible.
It is also interesting to see whether aspirations vary across cohorts and time periods, controlling
for age. Are younger birth cohorts indeed more materialistic as is often claimed? Do we live
in more materialistic times than our ancestors? These claims can be easily tested by looking at
the coefficients in the regressions that underlie the life cycle curves. The signs of the regression
coefficients on birth cohort and time in a pooled regression indicate that members of more recent
cohorts do indeed have higher aspirations for material goods and work-related items (Table 4.4).
Younger cohorts also report higher attainments in the pecuniary domain and a higher shortfall
Table 4.4: Regression coefficients on birth cohort and time
Have PGL Shortfall
Coh Coh
2
Time Coh Coh
2
Time Coh Coh
2
Time
Home + + +
Yard +/– +/– +/–
TV + + +/– + + +/– – – +/–
Car + – + + +/– +
2 TVs + + +/– + +/– + +/–
2 Cars + – +/– + – +/– + +
Nice clothes + – + – +/–
Travel abroad +/– + +/– + +/–
Swimming pool + – +/– + – +/–
Vacation home + – + – +/– + – +/–
Lots of money + + +/– + +/–
Happy marriage – – +/– – – +/– +
Children – +/–
Edu. children – – + – – + – –
Interesting job – – – – + – +
Job pays well – + +/– + +/–
Welfare job – + – + –
Note: Only statistically significant coefficients are reported.
81
between aspirations and attainments especially in the non-pecuniary domains. The results for the
survey period are mixed.
Thesameregressionsalsoallowmetoassessdifferencesinaspirationsandattainmentsbetween
genders, race and education levels (Table 4.5). On average, males have higher aspirations in some
of the material domains and in the work domain, accompanied by higher attainments in these
domains. Women report higher shortfalls in the family and work domain than males. In contrast
to the life cycle curves, these regression coefficients reflect average differences across all ages.
With regard to race, blacks report significantly higher discrepancies between aspirations and
attainments than other individuals in virtually all domains, due to higher aspirations and lower
attainments. Not surprisingly, better educated individuals – who probably earn higher incomes
and have different reference groups than less educated individuals – report higher aspirations in
all domains, as well as higher attainments and smaller shortfalls in some of them.
Table 4.5: Regression coefficients on gender, race and education
Have PGL Shortfall
Male Black Edu. Male Black Edu. Male Black Edu.
Home – – + + + + –
Yard – – + – + + + –
TV + + + + –
Car + – + + – + –
2 TVs + – + + + + + –
2 Cars + – + + – + + + –
Nice clothes – + + – + – – + –
Travel abroad + – + + – + +
Swimming pool – + + + + +
Vacation home + – + + + + + + +
Lots of money + + + + – + + –
Happy marriage + – + – + – +
Children – – – – + + +
Edu. children – – + – + – + +
Interesting job + – + + – + – + –
Job pays well + – + + + – + –
Welfare job + + + – + +
Health + – + + + –
Note: Only statistically significant coefficients are reported.
82
4.4 Discussion
The present empirical analysis supports the hypothesis that there is a difference in the way that
pecuniary and nonpecuniary aspirations change over the life cycle. Material goods aspirations
typically rise throughout the life cycle in contrast to family, work and health aspirations which
tend, on average, to be fairly stable or declining.
What do these results imply for whether individuals currently achieve the optimal allocation
of time and money among domains? The answer is that it appears that people may allocate a
disproportionate amount of time to the pursuit of pecuniary rather than nonpecuniary objectives,
as well as to “comfort” and positional goods, and to shortchange goals that will have a more
lasting effect on well-being (cf. also Frank, 1997). This misallocation occurs because in making
decisions about how to use their time, individuals tend to take their aspirations as fixed at their
present levels, and fail to recognize that aspirations may change because of hedonic adaptation
and social comparison. In particular, people make decisions assuming that more income, comfort,
and positional goods will make them happier, failing to recognize that hedonic adaptation and
social comparison will come into play, raise their aspirations and leave them feeling not much
happier than before. As a result, most individuals spend a disproportionate amount of time
working, and sacrifice family life and health, domains in which aspirations remain fairly constant
as actual circumstances change, and where attainment of one’s goals has a more lasting impact
on happiness. Hence a reallocation of time and money in favor of family life and health would,
on average, increase individual happiness.
The results of the present analysis confirm some of the results found in previous chapters.
Aspirations with regards to marriage do not increase; a finding which supports the results of the
firstessaythatmarriagehasalastingeffectonwell-being. Theresultsinthischaptersuggestcon-
tinuously increasing aspirations for material goods, whereas the analysis of financial satisfaction
(Chapter 3) indicates declining financial aspirations. Various studies in gerontology show that
83
older individuals are surprisingly satisfied with their financial circumstances despite low incomes
(George, 1992), suggesting lower material aspirations at old age. The Roper data provide limited
information for older individuals and the findings for respondents over 65 are subject to consider-
able reservation. If the data were better for older individuals one would likely observe declining
material aspirations at old age, which would confirm the results of the previous chapter.
A research topic of considerable interest would be an analysis of gender differences in aspira-
tions. Such differences in aspirations could possibly explain the disparate patterns in life cycle
domain satisfaction for men and women found by Marcelli and Easterlin (2007).
84
Chapter 5
Epilogue: Policy Implications
Research on subjective well-being provides policy makers with valuable information and also
allows individuals to evaluate their decisions and goals more accurately. Diener (2006) suggests
that indicators of subjective well-being can be used both to assess the needs for policies and to
evaluate the outcome of policy interventions. Therefore, some researchers suggest that national
economic measures of well-being like the GDP should at the very least be complemented by
subjective measures (Veenhoven, 2002; Shah and Marks, 2004; Donovan and Halpern, 2002).
Subjective measures of well-being thus allow a more complex assessment of welfare, but public
policy and personal decisions would be futile if subjective well-being was mainly determined by
geneticsandpersonalityassuggestedbyseptpointtheoryinpsychology(Kammann,1983;Lykken
and Tellegen, 1996; Myers, 1992, 2000; Lucas et al., 2003). The results in the previous chapters
indicatethatsubjectivewell-being–measuredhereintheformoflifesatisfactionordomainsatis-
faction – can indeed be improved by changes in life circumstances; especially in the nonpecuniary
domains in which aspirations do not increase in step with attainments (Chapter 4). Individuals
mightnotallocatetheirtimeandmoneyefficientlyamongpecuniaryandnonpecuniaryobjectives
if they base their decisions on current aspiration levels and do not anticipate a subsequent change
in aspirations (see also Easterlin, 2001a). The pursuit of material goals not only often fails to
make one happier due to hedonic adaptation and social comparison, but also imposes high costs
85
onotherpeoplebecausetheconsumptionchoicesofoneindividualinfluencetherelativestandards
of others. Consumption, on average, thus causes negative positional externalities and can lead to
expenditure arms races (Frank, 1997, 2005). A reallocation of time and money in favor of family
life and health would increase individual happiness and diminish the “rat race” because attain-
ments in the nonpecuniary domains are less visible to other people and usually less subject to
social comparison than pecuniary gains. The results in the previous chapters can help individuals
evaluate their goals more accurately by illustrating some of the consequences of individual choices
in terms of well-being. If individuals are aware of the influences of consumption on aspirations
they might be able to improve their forecasts of the gains in well-being from their expenditures.
My results indicate that married individuals experience lasting improvements in well-being
(Chapter 2). But should public policy aim at encouraging marriage in order to let more people
benefit from its positive effects? Policies that encourage marriage – i.e. tax benefits and other
preferentialtreatmentsofmarriedcouples–mightcausebad“matches”betweenpeoplewhowould
have not considered marriage without these extra perks. I therefore argue that policies should
not directly encourage marriage, but instead aim at stabilizing marital unions.
1
Of course, any
policy that has the stabilization of marital unions as a goal also indirectly encourages marriage.
Such policies could be aimed at reducing stressors in unions, for instance by providing access to
couple counseling services. Policies could encourage the participation in education or counseling
programs prior to marriage, which teach couples effective communication and conflict resolution
skills (Brotherson and Duncan, 2004). During marriage, programs could help couples anticipate
and effectively handle common stressors – e.g. the transition to parenthood, job loss, health
problems, etc. (Bradbury and Karney, 2004). Policies to strengthen marriages can either help
couples prevent and overcome marital difficulties by providing counseling services or directly
attempt to eliminate the source of marital stressors. For instance, a significant source of marital
1
Other forms of long-term relationships, such as cohabiting relationships should probably receive the same
policy benefits, but my research focused on unions that lead to marriage.
86
stress can stem from child care and policies could therefore be aimed at providing subsidized child
care centers. This would also allow more women to participate in the labor market and alleviate
financial stress in a marriage.
My research on financial satisfaction indicates that liabilities cause a significant decline in
financial satisfaction (Chapter 3). Credit card debt and other forms of debt proved to have a
larger negative impact on satisfaction than mortgage debt. In fact, it has been observed that
credit card debt causes more anxiety than other forms of debt (Drentea, 2000). Public policy
could be aimed at influencing the demand side as well as the supply side of debt. For instance,
programs could be offered to educate people about financial management, which would inform
peopleabouttheimpactofloansontheircreditrating,howmuchtheywilleventuallypayfortheir
debt if they only make minimum credit card payments, and other related issues. Well-informed
consumers handle their financial situation better and probably avoid personal bankruptcy. In
fact, basic financial management strategies could be taught in the school curriculum, for instance
by teaching how to calculate interest payments in math classes. Policies targeting the supply
side of debt – i.e. credit card companies, banks, mortgage lenders, etc. – would probably be
met by resistance from lenders, but could have more significant effects. Such policies could be
aimed at changing rules about the disclosure of interest payments or restrict access to loans.
The main goal would be to prevent lenders from giving loans to people who will have considerable
difficultytopayofftheseloans. Atthetimeofthewritingofthisdissertationnumerousnewspaper
articles reported of the perils of subprime mortgage loans, which required no down payment or
interest-only payments during the recent housing boom (e.g. Creswell and Bajaj, March 5, 2007).
Many borrowers of these subprime loans did not realize that their interest rates could increase
to levels that were beyond their financial means, leading to a considerable number of foreclosures
subsequently. Public policy could prevent such personal financial disasters by imposing stricter
regulations on lenders for granting loans. At the very least, lenders should be required to explain
the repayment structure of loans carefully.
87
Asmentionedabove, lastinggainsinsatisfactioncanbeachievedinthenonpecuniarydomains
in which the positive influence of attainments on well-being is not undermined by increasing as-
pirations (Chapter 4). This finding indicates that public policies that are aimed at improvements
in the nonpecuniary domain will lead to more gains in well-being than policies that improve peo-
ple’s financial situation.
2
One domain in which public policy could lead to increases in subjective
well-being is the health domain because of only slightly increasing aspirations for good health
over the life cycle. Policies addressing preventive medical care or better access to health care for
individuals without adequate health insurance coverage could therefore improve well-being in a
society. Indeed, Kirkcaldy et al. (2005) find higher levels of satisfaction in countries with efficient
health care systems – i.e. health care systems of high quality and equity, which do not necessarily
require high expenditures.
Policies could also be aimed at shifting the focus away from consumption and thus prevent
the negative effects of a consumption “rat race”. Frank (1997) suggests the implementation of
a consumption tax, which would imply that only savings are tax free and people are therefore
encouragedtorefrainfromconsumption. Itisquestionablewhethersuchaconsumptiontaxcould
indeed prevent excessive spending. After all, people save money in order to spend it in the future.
Such a tax would likely have to be quite high in order to discourage consumption. A progressive
income tax might be more efficient because it would discourage individuals from working exces-
sively if any additional income is taxed at a high rate (see also Layard, 2005). On the other hand,
it could also be argued that a progressive income tax might diminish motivation and productiv-
ity in an economy. Nevertheless, the tax proceeds could be used to invest in improvements in
other domains of life which have been shown to lead to lasting gains in well-being, such as the
health domain. Another approach to prevent excessive consumption could involve a change in the
school curriculum. Scitovsky (1976) suggests that a focus on liberal arts education in school will
2
Though it should be noted that financial gains lead to large increases in subjective well-being for people
with low income. Transfer payments that assist low-income individuals should certainly not be diminished, but
individuals above a certain income threshold will, on average, not benefit much from pecuniary gains.
88
lead people to appreciate cultural goods instead of “comfort” goods, which only bring short-term
improvements in happiness.
Some of the policies described above restrict people’s choices, for instance, by limiting access
to loans that they might not be able to repay. Other policies – e.g. health care and marriage
counseling programs – could be expensive and maybe require new taxes. Proponents of consumer
sovereignty would likely dismiss many policy recommendations and argue that individuals should
maketheirownchoices,butpeopleoftendonotestimategainsinutilitycorrectlywhentheymake
decisions (Kahneman, 2000). Another point of discussion could be the size of the welfare state.
A large welfare state apparatus requires increased taxation, but it is often assumed that citizens
in nations with considerable welfare state benefits are happier. Veenhoven (2000) actually does
not find a link between the level of well-being and the size of the welfare state – measured in the
form of social insurances. This results is surprising but reasonable if welfare state expenditures
are not applied adequately – i.e. funds are allocated to projects that do not increase well-being.
My research and that of others can therefore help identify the domains of life in which public
policy would have a positive impact on subjective well-being.
89
Bibliography
Michael Argyle. Causes and correlates of happiness, chapter in Daniel Kahneman, Ed Diener and
Norbert Schwarz (eds.), Well-Being: The Foundations of Hedonic Psychology, pages 353–373.
Russell Sage, New York, 1999.
Gary S. Becker. A theory of marriage: Part I. The Journal of Political Economy, 81(4):813–846,
1973.
Gary S. Becker. A theory of marriage: Part II. The Journal of Political Economy, 82(2):S11–S26,
1974.
Eva Bernhardt and Ewa Fratczak. Family status and subjective well-being: Comparing Poland
and Sweden. Presented at the International Union for the Scientific Study of Population Con-
ference, Tours, France, 2005.
David G. Blanchflower and Andrew Oswald. Money, sex and happiness: An empirical study.
Scandinavian Journal of Economics, 106:393–415, 2004a.
DavidG.BlanchflowerandAndrewOswald. Well-beingovertimeinBritainandtheUSA. Journal
of Public Economics, 88:1359–1386, 2004b.
Thomas N. Bradbury and Benjamin R. Karney. Understanding and altering the longitudinal
course of marriage. Journal of Marriage and Family, 66(4):862–879, 2004.
Sean E. Brotherson and William C. Duncan. Rebinding the ties that bind: Government efforts
to preserve and promote marriage. Family Relations, 53(3):459–468, 2004.
Susan L. Brown. Moving from cohabitation to marriage: Effects on relationship quality. Social
Science Research, 33:1–19, 2004.
Susan L. Brown and Alan Booth. Cohabitation versus marriage: A comparison of relationship
quality. Journal of Marriage and the Family, 58(3):668–678, 1996.
Bundeszentrale f¨ ur politische Bildung. Die soziale Situation in Deutschland. Bundeszentrale f¨ ur
politische Bildung, Bonn, 2004.
Tania Burchardt. Are one man’s rags another man’s riches? Identifying adaptive expectations
using panel data. Social Indicators Research, 74:57–102, 2005.
Angus Campbell. The Sense of Well-Being in America. McGraw-Hill, New York, 1981.
Angus Campbell, Philip E. Converse, and Willard L. Rodgers. The Quality of American Life:
Perceptions, Evaluations, and Satisfactions. Russell Sage, New York, 1976.
Hadley Cantril. The Pattern of Human Concerns. Rutgers University Press, NJ, 1965.
William S. Cleveland. Robust locally weighted regression and smoothing scatter plots. Journal
of the American Statistical Association, 74:829–836, 1979.
90
RobertH.Coombs. Maritalstatusandpersonalwell-being: Aliteraturereview. Family Relations,
40(1):97–102, 1991.
Julie Creswell and Vikas Bajaj. Mortgage crisis spirals, and casualties mount. The New York
Times, March 5, 2007.
Eileen M. Crimmins and Richard A. Easterlin. What goals motivate individual behavior?, chapter
inK.WarnerSchaieandJohnHendricks(eds.),TheEvolutionoftheAgingSelf,pages159–168.
Springer Publishing Company, New York, 2000.
Richard Davidson and Nathan Fox. Asymmetrical brain activity discriminates between positive
and negative affective stimuli in human infants. Science, 218:1235–1237, 1982.
James A. Davis, Tom W. Smith, and Peter V. Marsden. General Social Surveys, 1972-2004
[CUMULATIVE FILE] [Computer file]. ICPSR04295-v2. Chicago, IL: National Opinion Re-
search Center [producer]. Storrs, CT: Roper Center for Public Opinion Research, University
of Connecticut/Ann Arbor, MI: Inter-university Consortium for Political and Social Research
[distributors], 2005.
Liam Delaney, Carol Newman, and Brian Nolan. Reference dependent financial satisfaction
over the course of the Celtic Tiger: A panel analysis utilising the Living in Ireland Sur-
vey 1994-2001. Unpublished manuscript, 2006. Retrieved on November 20, 2006 from
http://www.tcd.ie/Economics/TEP/2006papers/TEP11.pdf.
Alfred DeMaris and Gerald R. Leslie. Cohabitation with the future spouse: Its influence upon
marital satisfaction and communication. Journal of Marriage and the Family, 46(1):77–84,
1984.
Ed Diener. Guidelines for national indicators of subjective well-being and ill-being. Journal of
Happiness Studies, 7(4):397–404, 2006.
Ed Diener and Richard E. Lucas. Personality and subjective well-being, chapter in Daniel Kah-
neman, Ed Diener and Norbert Schwarz (eds.), Well-Being: The Foundations of Hedonic Psy-
chology. Russell Sage, New York, 1999.
Ed Diener and Martin E. P. Seligman. Beyond money: Toward an economy of well-being. Psy-
chological Science in the Public Interest, 5(1):1–31, 2004.
Rafael DiTella and Robert MacCulloch. Some uses of happiness data in economics. Journal of
Economic Perspectives, 20(1):25–46, 2006.
Nick Donovan and David Halpern. Life Satisfaction: The State of Knowledge and Implications
for Government. Cabinet Office analytical paper; REVIEW, London, 2002.
Patricia Drentea. Age, debt and anxiety. Journal of Health and Social Behavior, 41(4):437–450,
2000.
James S. Duesenberry. Income, Savings, and the Theory of Consumer Behavior. Harvard Uni-
versity Press, Cambridge, MA, 1949.
GregJ.Duncan, MarthaS.Hill, andWillardRodgers. The changing economic status of the young
and the old, chapter in Demographic Change and the Well-Being of Children and the Elderly:
Proceedings of a Workshop. 54-106 (Committee on Population, National Research Council).
National Academy Press, Washington, DC, 1987.
91
Richard A. Easterlin. Does economic growth improve the human lot?, chapter in Paul A. David
and Melvin W. Reder (eds.), Nations and Households in Economic Growth: Essays in Honor
of Moses Abramovitz. Academic Press, Inc., New York, 1974.
Richard A. Easterlin. Income and happiness: Towards a unified theory. The Economic Journal,
111(473):465–484, July 2001a.
Richard A. Easterlin. Life cycle welfare: Evidence and conjecture. The Journal of Socio-
Economics, 30(1):31–61, 2001b.
Richard A. Easterlin. Explaining happiness. Proceedings of the National Academy of Sciences,
100(19):11176–11183, May 2003.
RichardA. Easterlin. Life cycle happiness and its sources: Intersections of psychology, economics,
and demography. Journal of Economic Psychology, 27(4):463–482, 2006.
Richard A. Easterlin and Anke C. Zimmermann. Life satisfaction and economic outcomes in
Germany pre- and post-unification. unpublished manuscript, 2007.
PaulEkman,WallaceFriesen,andMaureenO’Sullivan. Smileswhenlying. Journal of Personality
and Social Psychology, 54(3):414–420, 1988.
Ada Ferrer-i-Carbonell. Income and well-being: An empirical analysis of the comparison income
effect. Journal of Public Economics, 89:997–1019, 2005.
Ada Ferrer-i-Carbonell and Paul Frijters. How important is methodology for the estimates of the
determinants of happiness? The Economic Journal, 114(497):641–659, 2004.
Robert H. Frank. Choosing the Right Pond: Human Behavior and the Quest for Status. Oxford
University Press, New York, 1985.
Robert H Frank. The frame of reference as a public good. The Economic Journal, 100(19):
1832–1847, 1997.
Robert H. Frank. Positional externalities cause large and preventable welfare losses. American
Economic Review, 95(2):137–141, 2005.
Shane Frederick and George Loewenstein. Hedonic adaptation, chapter in Daniel Kahneman and
Ed Diener (eds.), Well-Being: The Foundations of Hedonic Psychology, pages 302–329. Russel
Sage Foundation, New York, 1999.
Bruno S. Frey and Alois Stutzer. Happiness and Economics. Princeton University Press, Prince-
ton, N.J., 2002.
Linda K. George. Economic status and subjective well-being: A review of the literature and an
agenda for future research, chapter in Neal E. Cutler, Davis W. Gregg and M. Powell Lawton
(eds.), Aging, Money, and Life Satisfaction. Springer Publishing Company, New York, 1992.
Linda K. George. Financial Security in Later Life: The Subjective Side. Boettner Institute of
Financial Gerontology, Philadelphia, PA, 1993.
JohnP.Haisken-DeNewandJoachimR.Frick.DesktopcompaniontotheGermanSocio-Economic
Panel (SOEP). Version 8.0. Berlin: DIW (German Institute for Economic Research), 2005.
Thomas Hansen. Financial satisfaction in old age – a satisfaction paradox or a result of accu-
mulated wealth? Paper presented at The Gerontological Society of America’s 58th Annual
Scientific Meeting, Orlando , 2005.
92
Wolfgang H¨ ardle. Applied Nonparametric Regression. Cambridge University Press, Cambridge,
1990.
Ori Heffetz. Conspicuous consumption and the visibility of consumer expendi-
tures. Unpublished manuscript, 2006. Retrieved on March 20, 2007 from
http://forum.johnson.cornell.edu/faculty/heffetz/papers/conspicuous.pdf.
Fred Hirsch. Social Limits to Growth. Harvard University Press, Cambridge, MA, 1976.
AllanV.Horwitz,HeleneRaskinWhite,andSandraHowell-White. Becomingmarriedandmental
health: Alongitudinalanalysisofacohortofyoungadults. Journal of Marriage and the Family,
58:895–907, 1996.
Chang-Ming Hsieh. Trends in financial satisfaction among middle-age and old-age Americans,
1972-1996. International Journal of Aging and Human Development, 51(2):105–113, 2000.
David R. Johnson and Jian Wu. An empirical test of crisis, social selection, and role explanations
of the relationship between marital disruption and psychological distress; A pooled time series
analysis of four-wave panel data. Journal of Marriage and the Family, 64:211–224, 2002.
So-Hyun Joo and John E. Grable. An exploratory framework of the determinants of financial
satisfaction. Journal of Family and Economic Issues, 25(1):25–50, 2004.
DanielKahneman. Experiencedutilityandobjectivehappiness: Amoment-basedapproach,chapter
in Amos Tversky and Daniel Kahneman (eds.), Choices, Values, and Frames. Cambridge
University Press, Cambridge, 2000.
Daniel Kahneman and Alan B. Krueger. Developments in the measurement of subjective well-
being. Journal of Economic Perspectives, 20(1):3–24, 2006.
Richard Kammann. Objective circumstances, life satisfactions, and sense of well-being: Consis-
tencies across time and place. New Zealand Journal of Psychology, 12:14–22, 1983.
Bruce Kirkcaldy, Adrian Furnham, and Ruut Veenhoven. Health care and subjective well-being
in nations, chapter in Alexander-Stamatios G. Antoniou and Cary L. Cooper (eds.), Research
Companion to Organizational Health Psychology. Edward Elgar Publishers, Cheltenham, UK,
2005.
Hans-Peter Kohler, Jere R. Behrman, and Axel Skytthe. Partner + children = happiness? The
effects of partnerships and fertility on well-being. Population and Development Review, 31(1):
407–445, 2005.
Edward O. Laumann, John H. Gagnon, Robert T. Michael, and Stuart Michaels. The Social
Organization of Sexuality: Sexual Practices in the United States. University of Chicago Press,
Chicago, 1994.
Richard Layard. Happiness: Lessons from a New Science. The Penguin Press, New York, 2005.
JerseyLiang,EvaKahana,andEdmundDoherty. Financialwell-beingamongtheaged: Afurther
elaboration. Journal of Gerontology, 35:409–420, 1980.
RichardE.Lucas. Timedoesnothealallwounds.Alongitudinalstudyofreactionandadaptation
to divorce. Psychological Science, 16(12):945–950, 2005.
RichardE.LucasandAndrewE.Clark. Dopeoplereallyadapttomarriage? Journal of Happiness
Studies, 7(4):405–426, 2006.
93
Richard E. Lucas, Andrew E. Clark, Yannis Georgellis, and Ed Diener. Reexamining adaptation
and the set point model of happiness: Reactions to changes in marital status. Journal of
Personality and Social Psychology, 84(3):527–539, 2003.
Richard E. Lucas, Andrew E. Clark, Yannis Georgellis, and Ed Diener. Unemployment alters the
set point for life satisfaction. Psychological Science, 15(1):8–13, 2004.
Richard E. Lucas, Ed Diener, and Eunkook M. Suh. Discriminant validity of well-being measures.
Journal of Personality and Social Psychology, 71(3):616–628, 1996.
Douglas A. Luke. Multilevel Modeling. Series: Quantitative Applications in the Social Sciences;
143. Sage Publications, Thousand Oaks, CA, 2004.
Erzo F.P. Luttmer. Neighbors as negatives: Relative earnings and well-being. The Quarterly
Journal of Economics, pages 963–1002, 2005.
David Lykken and Auke Tellegen. Happiness is a stochastic phenomenon. Psychological Science,
7:186–189, 1996.
Enrico A. Marcelli and Richard A. Easterlin. The X-relation: Life cycle happiness of American
women and men. Unpublished manuscript, 2007.
Nadine F. Marks and James D. Lambert. Marital status continuity and change among young and
midlife adults: Longitudinal effects on psychological well-being. Journal of Family Issues, 19:
652–686, 1998.
Arne Mastekaasa. Marital dissolution and subjective distress: Panel evidence. European Socio-
logical Review, 11(2):173–185, 1995.
Alex C. Michalos. Multiple discrepancy theory (MDT). Social Indicators Research, 16:347–413,
1985.
Alex C. Michalos. Global Report on Student Well-Being. Springer-Verlag, New York, 1991.
Lawrence Mishel, Jared Bernstein, and Sylvia Allegretto. The state of working America
2004/2005. An Economic Policy Institute Book. ILR Press, Ithaca, NY, 2005.
Franco Modigliani. Fluctuations in the savings-income ratio: A problem in economic forecasting,
chapter In Conference on Research in Income and Wealth, Studies in Income and Wealth, XI.,
pages 371–443. National Bureau of Economic Research, New York, 1949.
David G. Myers. The Pursuit of Happiness. Avon, New York, 1992.
David G. Myers. The funds, friends, and faith of happy people. American Psychologist, 55(1):
56–67, 2000.
Yew-KwangNg. Economicgrowthandsocialwelfare: Theneedforacompletestudyofhappiness.
Kyklos, 31(4):575–587, 1978.
Steven L. Nock. A comparison of marriages and cohabiting relationships. Journal of Family
Issues, 16(1):53–76, 1995.
RobertA.Pollak. Habitformationanddynamicdemandfunctions. Journal of Political Economy,
78(4):745–763, 1970.
Robert A. Pollak. Interdependent preferences. American Economic Review, 66(3):309–320, 1976.
94
Nattavudh Powdthavee. A direct test for utility interdependence in marriage. Unpublished
manuscript, 2005.
Stephen Raudenbush and Anthony Bryk. Hierarchical Linear Models: Applications and Data
Analysis Methods. Sage Publications., Thousand Oaks, CA, 2002.
Stephen Raudenbush, Anthony Bryk, and Richard Congdon. HLM 6. Scientific Software Inter-
national, Lincolnwood, IL, 2000.
Roper-Starch Organization. Roper Reports. The Roper Center, University of Connecticut, Storrs,
1978-1995. Roper Reports 12-97, 12-2000 and 03-2003 were generously made available by Nop-
World (www.nopworld.com).
Tibor Scitovsky. The Joyless Economy. Oxford University Press, Oxford, 1976.
Hetan Shah and Nic Marks. A Well-Being Manifesto for a Flourishing Society. New Economics
Foundation, London, 2004.
Robin W. Simon. Revisiting the relationship among gender, marital status, and mental health.
American Journal of Sociology, 107:1065–1096, 2002.
M. Joseph Sirgy, Alex C. Michalos, Abbott L. Ferriss, Richard A. Easterlin, Donald Patrick, and
William Pavot. The qualityity-of-life (QOL) research movement: Past, present, and future.
Social Indicators Research, 76(3):343–466, 2006.
Emily Crawford Solberg, Ed Diener, Derrick Wirtz, and Richard E. Lucas. Wanting, having, and
satisfaction: Examining the role of desire discrepancies in satisfaction with income. Journal of
Personality and Social Psychology, 83(3):725–734, 2002.
Sara J. Solnick and David Hemenway. Are positional concerns stronger in some domains than in
others? American Economic Review, 95(2):147–151, 2005.
Steven Stack and J. Ross Eshleman. Marital status and happiness: A 17-nation study. Journal
of Marriage and the Family, 60(2):527–536, 1998.
Alois Stutzer. The role of income aspirations in individual happiness. Journal of Economic
Behavior and Organization, 54:89–109, 2004.
Alois Stutzer and Bruno S. Frey. Does marriage make people happy, or do happy people get
married? Journal of Socioeconomics, 35(2):326–347, 2006.
James A. Sweet and Larry L. Bumpass. The National Survey of Families and Households - Waves
1and2: DataDescriptionandDocumentation. Center for Demography and Ecology, University
of Wisconsin-Madison, (http://www.ssc.wisc.edu/nsfh/home.htm), 1996.
James A. Sweet and Larry L. Bumpass. The National Survey of Families and Households -
Waves 1, 2, and 3: Data Description and Documentation. Center for Demography and Ecology,
University of Wisconsin-Madison, (http://www.ssc.wisc.edu/nsfh/home.htm), 2002.
James A. Sweet, Larry L. Bumpass, and Vaughn Call. The Design and Content of The Na-
tional Survey of Families and Households. Center for Demography and Ecology, University of
Wisconsin-Madison, NSFH Working Paper # 1, 1988.
U.S. Bureau of the Census. Current Population Reports, P60-104, Household Money Income
in 1975 and Selected Social and Economic Characteristics of Households. U.S. Government
Printing Office, Washington, DC, 1977.
95
U.S. Bureau of the Census. Current Population Reports, P60-156, Money Income of Households,
Families and Persons in the United States: 1985. U.S.GovernmentPrintingOffice,Washington,
DC, 1987.
U.S. Bureau of the Census. Current Population Reports, P60-193, Money Income in the United
States: 1995(WithSeparateDataonValuationofNoncashBenefits) .U.S.GovernmentPrinting
Office, Washington, DC, 1996.
U.S. Bureau of the Census. Current Population Reports, P60-231, Income, Poverty, and Health
Insurance Coverage in the United States: 2005. U.S. Government Printing Office, Washington,
DC, 2006.
Rien van der Leeden. Multilevel analysis of repeated measures data. Quality and Quantity, 32:
15–29, 1998.
Bernard M. S. van Praag. The welfare function of income in Belgium: An empirical investigation.
European Economic Review,, 2:337–369, 1971.
BernardM.S.vanPraag. Theconnectionbetweenoldandnewapproachestofinancialsatisfaction ,
chapter in Luigino Bruni and Pier Luigi Porta (eds.), Economics and Happiness: Framing the
Analysis. Oxford University Press, Oxford, 2005.
Bernard M. S. van Praag and Arie Kapteyn. Further evidence on the individual welfare function
of income: An empirical investigation in the Netherlands. European Economic Review, 4:33–62,
1973.
Bernard M.S. van Praag and Ada Ferrer-i-Carbonell. Happiness Quantified. A Satisfaction Cal-
culus Approach. Oxford University Press, Oxford, 2004.
Thorstein Veblen. The Theory of the Leisure Class: An Economic Study of Institutions. MacMil-
lan, New York, 1899.
Ruut Veenhoven. Well-being in the welfare state: Level not higher, distribution not more equi-
table. Journal of Compartive Policy Analysis: Research and Practice, 2:91–125, 2000.
Ruut Veenhoven. Why social policy needs subjective indicators. Social Indicators Research, 58
(1):33–45, 2002.
Esperanza Vera-Toscano, Victoria Ateca-Amestoy, and Rafael Serrano-Del-Rosal. Building finan-
cial satisfaction. Social Indicators Research, 77:211–243, 2006.
Linda J. Waite. Does marriage matter? Demography, 32(4):483–507, 1995.
Linda J. Waite, Dan Browning, William J. Doherty, Maggie Gallagher, Ye Lou, and Scott M.
Stanley. Does Divorce Make People Happy? Findings from a Study of Unhappy Marriages.
Institute for American Values, New York, 2002.
Linda J. Waite and Kara Joyner. Emotional satisfaction and physical pleasure in sexual unions:
Time horizon, sexual behavior and sexual exclusivity. Journal of Marriage and the Family, 63:
247–264, 2001.
Linda J. Waite and Evelyn L. Lehrer. The benefits from marriage and religion in the United
States: A comparative analysis. Population and Development Review, 29(2):255–275, 2003.
LindaJ.WaiteandYeLuo. Maritalhappinessandmaritalstability: Consequencesforpsychologi-
calwell-being. Presented at the 97th meeting of the American Sociological Association, Chicago,
IL, 2003.
96
Appendix A
Chapter 2
A.1 German Socio Economic Panel data
Satisfaction:
In conclusion, we would like to ask you about your satisfaction with your life in general. Please
answer according to the following scale: “0” means completely dissatisfied, “10” means completely
satisfied.
How satisfied are you with your life all things considered?
Marital status:
What is your marital status?
• Married, living together with spouse
• Married, living (permanently) separated from my spouse
• Single
• Divorced
• Widowed
• Spouse living in different country (This marital status category is only asked of foreigners
and in a different section of the survey, but it is included in the marital status variable
generated by the German Institute for Economic Research (DIW).)
Cohabitation:
1984: question missing (but cohabitation can be derived from retrospective question in 1985 sur-
vey)
1985:
• Has your family situation changed since the beginning of [the year that was 2 years before
the current survey]? Please answer whether any of the following applies to you, and if so,
when
- moved in with partner
If answer to (a) is yes, cohab. = 1; otherwise = 0.
97
1986-1990:
• Are you living with someone in a long-term relationship?
• (if yes) since when have you lived together?
• (or) live in separate apartments since
• Has your family situation changed since the beginning of [the year that was 2 years before
the current survey]? Please answer whether any of the following applies to you, and if so,
when
- moved in with partner
If answer to (a) or (d) is yes, cohab. = 1; otherwise = 0.
1991-1997:
• Are you living with someone in a long-term relationship?
• (if yes) Does your partner live in your household?
• Has your family situation changed since the beginning of [the year that was 2 years before
the current survey]? Please answer whether any of the following applies to you, and if so,
when
- moved in with partner
If answer to (b) or (c) is yes, cohab. = 1; otherwise = 0.
1998-2004:
• Are you in a serious/permanent relationship?
• (if yes) Does your partner live in the same household?
• Has your family situation changed after December 31, [year that was 2 years before the
current survey]? Please indicate if any of the following apply to you and if so, when this
change occurred.
- I moved in with my partner (if the respondent marks “yes”, he/she also indicates whether
this event took place in the year of the survey or the year preceding the survey.)
If answer to (b) or (c) is yes, cohab. = 1; otherwise = 0.
Employed:
Are you currently engaged in paid employment? Which of the following applies best to your status?
• [1] Full-time employment
• [2] Regular part-time employment
• [3] Vocational training
• [4] Marginal part-time employment
• [5] Maternity leave (not available 1984-1990, 1999-2004)
• [6] Military, community service
• [7] Not employed
98
• [8] Unemployed (only available in 1984)
• [9] Disabled employment (only available in 1998-2003)
• [10] Near retirement, zero working hours (only 2002-2004)
Creation of a dummy variable “employed”: “Employed” has a value of one in a given survey
year if [1] Full-time employment, [2] Regular part-time employment, [3] Vocational training [6]
Military, community service, or [9] Disabled employed.
Religiosity:
Religiosity is measured by church attendance and the importance of religion.
Church attendance:
1 = daily, 2 = weekly, 3 = monthly, 4 = less frequently, 5 = never
Importance of Religion:
1 = very important, 2 = important, 3 = less important, 4 = very unimportant
Religiosity = 1 if church attendance = 1 or 2 or importance of religion = 1 or 2 at some point
during the years surveyed; otherwise religiosity = 0.
Health status: (available 1992, 1994-2003)
How would you describe your current health?
Very good (5), Good (4), Satisfactory (3), Poor (2), Bad (1) [original coding reversed]
Education more than highschool:
Generated CNEF (cross-national equivalent file) variable.
8
Education with respect to high school. Less than high school (1), completed high school (2),
more than high school (3). The level of education might change during the survey period, but the
analysis takes into account the highest level of education reported during the survey period.
Children:
Generated CNEF (cross-national equivalent file) variable.
8
Number of children in household (under age 18).
The analysis only considers children that are in the household in the year of marriage and after
(to avoid counting siblings or other relatives as children).
Income:
Generated CNEF (cross-national equivalent file) variable.
8
Household post-tax income.
Originally measured in current year Euros and then converted toe1995.
8
Constructed variables are not directly available in the original surveys and derived from several questions in
the survey.
99
A.2 Cohabitation and Divorce
Table A.1: Prevalence of cohabitation and divorce in each sample period
Created years before and Number of % cohabiting % divorced mean
Variable after marriage observations in that year in that year age
baseline -18 3 0 0 19.67
baseline/cohab -17 8 0 0 21.50
baseline/cohab -16 28 3.57 0 19.57
baseline/cohab -15 46 4.35 0 19.76
baseline/cohab -14 75 0 0 20.81
baseline/cohab -13 107 2.80 0 20.96
baseline/cohab -12 158 5.70 0 20.88
baseline/cohab -11 231 6.93 0 20.94
baseline/cohab -10 309 6.47 0 21.26
baseline/cohab -9 390 10.00 0 21.81
baseline/cohab -8 508 12.20 0 22.21
baseline/cohab -7 637 15.54 0 22.71
baseline/cohab -6 784 20.92 0 23.26
baseline/cohab -5 923 24.70 0 23.95
baseline/cohab -4 1,107 31.71 0 24.57
baseline/cohab -3 1,322 42.06 0 25.29
baseline/cohab -2 1,582 54.36 0 26.07
reaction/cohab -1 1,582 67.32 0 27.10
reaction wedding (0) 1,582 0 0 28.17
reaction 1 1,582 0 0 29.19
adaptation 2 1,582 0 0 30.21
adaptation/divorce 3 1,413 0 2.62 31.17
adaptation/divorce 4 1,260 0 4.13 32.02
adaptation/divorce 5 1,100 0 4.91 32.80
adaptation/divorce 6 975 0 6.26 33.71
adaptation/divorce 7 852 0 7.63 34.52
adaptation/divorce 8 742 0 9.43 35.43
adaptation/divorce 9 630 0 9.84 36.37
adaptation/divorce 10 527 0 10.44 37.40
adaptation/divorce 11 442 0 9.95 38.27
adaptation/divorce 12 359 0 10.31 39.22
adaptation/divorce 13 296 0 9.80 40.12
adaptation/divorce 14 244 0 10.25 41.05
adaptation/divorce 15 184 0 11.96 42.09
adaptation/divorce 16 139 0 11.51 42.90
adaptation/divorce 17 84 0 8.33 44.24
adaptation/divorce 18 34 0 8.82 45.62
100
Appendix B
Chapter 3
B.1 National Survey of Families and Households data
Financial satisfaction
On a scale of 1 to 7, where 1 is very dissatisfied and 7 is very satisfied, overall, how satisfied are
you with your financial situation?
Household Income
Aggregate measure consisting of income of all household members from wages, salaries, com-
missions, and tips, self-employment, social security or railroad retirement income, retirement or
pension income, public assistance, income from any other government program, such as veterans’
benefits, unemployment compensation, worker’s compensation, or supplemental security Income,
child support, alimony, or family support, income from interest, dividends, rent, or other invest-
ments, and income from any other source.
The household income variable is a “best measure” income variable based on a comparison of the
main respondents’ reports and their spouses’ reports (for more details, see Appendix B).
Financial Assets, excluding checking accounts
Aggregate measure derived from answers to the following two questions:
1. What is the approximate total value of your (and your husband’s/and your wife’s) savings,
including savings accounts, savings bonds, IRAs, money market funds, and CDs?
2. In addition to these savings, what is the approximate total value of your (and your hus-
band’s/and your wife’s) other investments, including stocks, bonds, shares in mutual funds, or
other investments?
Net worth
Financial and tangible assets minus total debt.
Homeownership
Homeownership is derived from the question:
Do you (and your husband/wife) own your own home or are you renting?
Home value
How much do you think your home would sell for now?
Home debt
How much, if anything, do you (or your wife/or your husband) owe on your home?
101
Credit card debt
Total outstanding credit card balance:
How much, if anything, do you (and your husband/wife) owe on credit cards or charge accounts
that you are paying off gradually? If you almost always pay off your credit card balance each
month, answer “0”.
Mortgage payments
Aggregate measure which includes monthly mortgage payments plus property tax derived from
the following questions:
How much is your monthly payment on your home loan? If you have a second mortgage on your
home, include it in your answer.
Does this amount include property taxes?
What was the total property tax on your home last year?
Loans on purchases
How much, if anything, do you (and your husband/wife) owe on installment loans for major
purchases, such as furniture or appliances, but other than auto loans?
How much are you supposed to pay each month on this debt?
Educational loans
How much, if anything, do you (and your husband/wife) owe on educational loans?
How much are you supposed to pay each month?
Bank loans
How much, if anything, do you (and your husband/wife) owe on personal loans from banks and
other businesses, other than mortgage or auto loans or loans you have already told me about?
How much are you supposed to pay each month?
Loans from friends
How much, if anything, do you (and your husband/wife) owe on personal loans from friends or
relatives, other than those you have already told me about?
How much are you supposed to pay each month?
Loans for home improvement
How much, if anything, do you (and your husband/wife) owe on home improvement loans, other
than those you have already told me about?
How much are you supposed to pay each month?
Bills
How much, if anything, do you (and your husband/wife) owe on other bills you’ve owed for more
than two months?
Other debt payments
How much, if anything, do you (and your husband/wife) owe on any other debts that we have not
mentioned?
How much are you supposed to pay each month?
Self-rated health
Compared with other people your age, how would you describe your health?
Very poor (1), poor (2), fair (3), good (4), excellent (5)
102
Unemployed
Includes respondents who report that they are not currently working and were actively looking
for paid work during the previous four weeks.
Have you looked for work during the last 4 weeks?
Standard of living during retirement
My standard of living will get much worse when I retire.
Strongly agree, agree, neither agree nor disagree, disagree, strongly disagree
B.2 NSFH Income Data
B.2.1 Income measure
The household income variable for wave 2 is a “best measure” variable provided by the survey
institute. For wave 3, I constructed a best measure income variable based on the description
provided by the survey institute about their construction of a best measure income variable (see
Appendix J, NSFH
1
).
Householders and their spouses were asked about their own incomes as well as the incomes
of all other household members. Non-householders were only asked to report their own incomes.
For this reason, the final data only include householders. Respondents were asked about the
total amount of income in various income categories. If the respondent did not know or refused
to indicate the absolute amount of their income, they were asked whether their income lies in
a certain interval. For instance, the first question would be “Was it less than $20,000”, and if
so, the next question would be “Was it less than $15,000”, and so on until the income could be
allocated to a narrow income interval.
Forthebestmeasureincome,thereportsofthemainrespondentandthespousewerecompared
andseveralcriteriawereestablishedtochoosewhichincomereportshouldbeselectedforthebest
income measure. The respondent’s income is the default income (i.e. the primary respondent or
the spouse). The respondent’s spouse’s income report is preferred if the spouse reports a dollar
amount and the primary respondent reports an interval. It is also chosen if the respondent does
not give an answer and the spouse does. For the incomes of household members other than the
primary respondent and the spouse, the decision rules are as follows. The primary respondent’s
report is chosen if he and the spouse report the same dollar value or the same interval. If both
providedifferingreports,theaverageofbothreportsischosen. Ifoneprovidesanintervalresponse
andtheotherreportsadollaramount,theresponsewiththedollaramountispreferred. Similarly,
if an answer is missing, refused or a $0 amount is reported, the other’s report is preferred as well.
B.2.2 Life cycle income
The balanced panel sample which includes waves 2 and 3 is a sample of householders with non-
missing reports of the financial satisfaction question. Respondents were re-interviewed 8-11 years
aftertheinitialinterview. Thelifecycleprofilesinthispaperdothereforenotreflecttheexperience
of only one birth cohort which is followed over time. The data are a 10-year panel which means
that the observations for young ages are based on respondents of recent birth cohorts. Similarly,
observations for older ages in the life cycle profiles are derived from the responses of members
of older birth cohorts. It could therefore be argued that the life cycle profiles do not accurately
reflect the experience of a single birth cohort, but instead reflect distinctive cohort differences
in financial satisfaction. To see if this argument has merit, I compared the life cycle profile for
1
Available on the NSFH website at http://www.ssc.wisc.edu/nsfh/codedata2.htm (retrieved Nov. 30, 2006).
103
household income derived in this paper to the experience of single birth cohorts in the Current
Population Surveys (CPS).
For this comparison, I used the report on mean and median household incomes for various
age groups in four CPS surveys, from 1975 to 2005 (U.S. Bureau of the Census, 1977, 1987, 1996,
2006). Thesedatafollowseveralbirthcohortsovertime. Forinstance, the1975CPSreportshows
householdincomesforthe1931-40birthcohortwhenmembersofthiscohortwerebetween35and
44 years old. The 1985 CPS report shows incomes for the same birth cohort at age 45-54, and
so on. I followed several birth cohort and averaged indexed household incomes. Tables B.1 and
B.2 show that the indexed values are remarkably similar to the values derived from the life cycle
profile of the log of household income in this paper. This comparison with CPS data suggests
that the life cycle profile of household income is similar to the experience of a single birth cohort
and not simply determined by differences in the experiences of older and younger birth cohorts.
Table B.1: Comparison of median household incomes, CPS and NSFH. Household incomes are
converted to index form with age 45-54 = 100.00
CPS birth cohort CPS NSFH wave 2
Age group 1921-30 1931-40 1941-50 1951-60 mean log income
25-34 77.19 72.94 75.07 67.01
35-44 93.32 91.56 89.23 91.37 90.23
45-54 100.00 100.00 100.00 100.00 100.00 100.00
55-64 78.48 80.90 84.86 81.41 79.10
65-74 49.93 52.51 51.22 52.02
75 & over 36.95 36.95 36.54
Table B.2: Comparison of mean household incomes, CPS and NSFH. Household incomes are
converted to index form with age 45-54 = 100.00
CPS birth cohort CPS NSFH wave 2
Age group 1921-30 1931-40 1941-50 1951-60 mean log income
25-34 66.31 62.42 64.37 66.25
35-44 89.21 84.57 84.07 85.95 91.33
45-54 100.00 100.00 100.00 100.00 100.00 100.00
55-64 87.88 92.77 93.11 91.25 79.51
65-74 65.02 71.14 68.08 53.53
75 & over 48.23 48.23 36.59
104
Appendix C
Chapter 4
C.1 Roper Survey data
Respondents were asked the following questions about the “good life”.
We often hear people talk about what they want out of life. Here are a number of different things.
(HAND RESPONDENT CARD) When you think of the good life – the life you’d like to have,
which of the things on this list, if any, are part of that good life as far as you personally are
concerned?
The following list of variables is a subset of the items included in the surveys. The variable name
is the name that I will use for the analysis. The description follows the exact wording found in
the survey.
Goods
Home A home you own.
Yard A yard and lawn.
TV A color TV set.
Car A car.
2 cars A second car.
2 TVs A second color TV set.
Nice clothes Really nice clothes.
Travel abroad Travel abroad.
Swimming pool A swimming pool.
Vacation Home A vacation home.
A lot of money A lot of money.
105
Family
Happy Marriage A happy marriage.
Children One or more children
1
.
Children’s college education A college education for my children.
Job
Interesting job A job that is interesting.
Job that pays well A job that pays much more than average.
Welfare job A job that contributes to the welfare of society.
Health
Good Health Good Health.
1
The surveys ask about having 1, 2, 3 and 4 children. I created a dummy variable which indicates whether the
respondent considers 1 or more children to be part of the good life.
106
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Essays on the economics of subjective well-being in transition countries
PDF
Four essays on how policy, the labor market, and age relate to subjective well-being
PDF
Three essays on social policy: institutional development, and subjective well-being as a cause and consequence of labor market outcomes
PDF
What leads to a happy life? Subjective well-being in Alaska, China, and Australia
PDF
Explaining well-being: essays on the socio-economic factors accompanying life satisfaction
PDF
An economic analysis of cooperative group behavior: Theory and evidence
PDF
Essays on health and well-being
PDF
Essays on behavioral decision making and perceptions: the role of information, aspirations and reference groups on persuasion, risk seeking and life satisfaction
PDF
Life-cycle demographic decisions and the economic well-being of single parents in the United States, 1964-1989
PDF
Three essays on cooperation, social interactions, and religion
PDF
Essays on the new economic history of the Middle East
PDF
Essays on education programs in Costa Rica
PDF
Essays on development and health economics
PDF
The Catholic church in Latin America: an evaluation of the institutional and political impacts of progressive church reforms
PDF
Essays on the economics of non-profit institutions
PDF
Essays in political economics
PDF
Essays on development economics and adolescent behavior
PDF
Three essays in education and urban economics
PDF
Essays in political economy and mechanism design
PDF
Children 's migration and the financial, social, and psychological well-being of older adults in rural China
Asset Metadata
Creator
Zimmermann, Anke C.
(author)
Core Title
Adaptation, assets, and aspiration. Three essays on the economics of subjective well-being
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Economics
Publication Date
04/13/2007
Defense Date
03/20/2007
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
adaptation,aspirations,financial satisfaction,OAI-PMH Harvest,subjective well-being
Language
English
Advisor
Easterlin, Richard A. (
committee chair
), Kuran, Timur (
committee member
), Musick, Kelly (
committee member
), Nugent, Jeffrey B. (
committee member
), Silverstein, Merril (
committee member
), Tan, Guofu (
committee member
)
Creator Email
azimmerm@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-m361
Unique identifier
UC1234770
Identifier
etd-Zimmermann-20070413 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-323212 (legacy record id),usctheses-m361 (legacy record id)
Legacy Identifier
etd-Zimmermann-20070413.pdf
Dmrecord
323212
Document Type
Dissertation
Rights
Zimmermann, Anke C.
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Repository Name
Libraries, University of Southern California
Repository Location
Los Angeles, California
Repository Email
cisadmin@lib.usc.edu
Tags
adaptation
aspirations
financial satisfaction
subjective well-being