Upload
gaed
View
32
Download
1
Embed Size (px)
DESCRIPTION
An original socio-economic analysis of marriage and happiness by Prof. Dr. Donata Bessey, who has reserved all intellectual property rights.
Citation preview
Love Actually?Dissecting the
Marriage-Happiness Relationship
Donata Bessey∗
July 18, 2012
∗Yonsei University, School of Government and Business, Department of Economics, 1Yonseidae-gil, 220-778 Wonju, Republic of Korea, [email protected]. I thank SimonJanssen, Taejeong Lee, Kimiko Osawa and Ines Pelger as well as session participants atthe 2011 annual meeting of the Canadian Economics Association, at the 2012 annualmeeting of the Korean Association for Applied Economics and seminar participants atYonsei University for helpful comments.
1
Abstract
Using theoretical concepts based on identity economics, this pa-per empirically tests the idea that adherence to social norms to getmarried can provide an additional utility gain from marriage. Normsto get married should be stronger among more traditionalist individ-uals, so they should put more emphasis on the mere fact of gettingmarried and put less emphasis on match quality. In line with thosetheoretical predictions, marital satisfaction increases utility for non-traditionalist individuals, while traditionalists seem to be in marriageswith lower match quality. The mere fact of being married is associ-ated with lower happiness when controlling for marital satisfaction fornon-traditionalists, but there seems to be an identity-based happinessgain from marriage for traditionalist individuals. These findings canbe interpreted as evidence for identity-based utility effects from mar-riage.
JEL classification: I31, J12Keywords: Marriage, subjective well-being, identity
2
1 Motivation
One of the most pervasive findings from empirical happiness research is the
positive relationship between marital status and happiness: married people
are found to be happier than singles, divorce(e)s, separated and widowed
individuals in a considerable number of previous studies. See, for example,
Diener et al. (1999) for a review of the psychological literature on research in
subjective well-being, or Frey and Stutzer (2002) for a review of the economic
literature.
Most papers provide an explanation of the benefits from marriage in terms
of specialization of partners, mutual support and companionship, or the pro-
vision of services for which there are no or imperfect markets (Kohler et al.
2005). However, there might also be another source of beneficial effects from
marriage: social norms to marry and the resulting relief once one has com-
plied with the norms and is married. Morgan and Berkowitz King (2001)
discuss the importance of marriage as a social norm in the context of fertility
decisions. From an economic point of view, this relates to recent research
on identity economics (Akerlof and Kranton 2000). In this paper, I test the
hypothesis that there are additional benefits from marriage because of the
existence of social norms using data from the 2006 East Asian Social Survey
(EASS), including data from Japan, the Republic of Korea and the Republic
of China (Taiwan).
Despite important differences between the countries, traditional values in
all East Asian countries are deeply influenced by Confucian thought (Chang
1997), and its principles are still present in everyday life (Tu 1996). The Con-
3
fucianist cultures of East Asia place a special emphasis on the family (Lee
1989), with resulting strong prescriptions to marry and continue the family
line. This emphasis is an implication of filial piety, one of the five human
relations in Confucian thought. According to Mencius, the most important
Confucian philosopher after Confucius himself, it is the worst crime against
filial piety to not leave any posterity (Tang 1995). Therefore, marriage as
the only means of fathering legitimate posterity is given special prominence
in all countries influenced by Confucian thought. Compared to, for example,
European countries, marriage rates are still comparatively high (although
declining, most sharply in Japan) and cohabitation is a relatively unknown
phenomenon (Jones 2004). This suggests that social norms to marry are still
relatively strong in East Asian countries and that there might be the possibil-
ity of identity-based utility gains from marriage, especially for traditionalist
individuals.
In this paper, I empirically test the relationship between traditional val-
ues and different possible happiness effects of marriage. Assuming that so-
cial norms to marry are stronger among more traditionalist individuals, one
would expect an additional positive effect of the mere fact of being married
for them because they have complied with social norms. However, because
traditionalists should put more emphasis on the simple fact of being mar-
ried and probably put less emphasis on the quality of the match with their
spouse, I do not expect a positive effect stemming from marital satisfaction
on happiness for them. In line with the theoretical predictions, I find indeed
that marital satisfaction, but not the mere fact of being married, has a posi-
tive effect on happiness for non-traditionalists, and traditionalists seem to be
4
more likely to be in marriages with worse match quality as measured by lower
marital satisfaction. Finally, I also find that, controlling for marital satis-
faction, the simple fact of being married has a negative effect on happiness
for non-traditionalists, while it has a positive albeit statistically insignificant
effect on happiness for traditionalists. However, the regressor capturing the
fact of being married and the interaction term for being a married tradition-
alist are highly significant together. These findings suggest that there are
indeed identity-based utility gains for traditionalists.
The remainder of this paper is organized as follows. Part 2 presents theoret-
ical considerations, part 3 introduces the data set, describes the regressors
used in the empirical analysis and provides some interesting descriptive statis-
tics, part 4 outlines the estimation strategy and presents estimation results,
and part 5 concludes.
2 Theoretical Considerations
Akerlof and Kranton (2000) introduced the concept of identity into economic
theory, and show how identity, or an individual’s sense of self, matters for
economic outcomes. The model assumes that identity depends on an indi-
vidual’s assigned social category, how the individual’s characteristics match
that category’s ideals, and how the individual’s and other society members’
actions correspond to the set of behaviors for the assigned category.
Following their terminology, I assume that there are two different social cat-
egories Ci: traditionalists (type A), and non-traditionalists (type B). Corre-
spondingly, there are two different prescriptions or sets of behaviors Pi con-
5
sidered to be appropriate for members of the categories, with our behavior
of interest being the decision to get married. Let Pa denote the prescriptions
for the traditionalists, and Pb the prescriptions for the non-traditionalists.
Assume further that type A individuals’ expectations towards others include
to comply with the social norm to get married, while type B individuals do
not have this expectation.
Akerlof and Kranton incorporate identity into the following utility function:
Uj = Uj(aj, a−j, Ij) (1)
In this utility function, aj denotes the vector of one’s own actions , a−j
denotes the vector of other society members’ actions, and Ij denotes an in-
dividual’s identity. Our action aj of interest is the decision to get married.
For the sake of simplicity, assume that marriage increases utility through two
channels: a marriage-derived utility component and an identity-derived util-
ity component. The marriage-derived utility component includes the familiar
explanations for the utility-increasing characteristics of marriage, i.e., spe-
cialization of partners, mutual support and companionship, or the provision
of services for which there are no or imperfect markets (Kohler et al. 2005).
Let aj denote this utility component. The identity-derived component affects
utility through approval or disapproval from individuals of the same social
category (a−j), but also through Ij if an individual acts in accordance with
his or her identity.
All individuals choose whether to get married or not. Because the prescrip-
tion to get married should be much stronger among traditionalists, tradi-
6
tionalists should probably put more emphasis on the mere fact of getting
married and put less emphasis on the quality of the match with their spouse,
compared to non-traditionalists. This should lead to the following two pre-
dictions: firstly, married traditionalists should derive utility from the fact of
being married, and secondly, they should, ceteris paribus, derive less utility
through the channel of marital satisfaction. Hence, I expect ∂Uj/∂aj = 0
and ∂Uj/∂a−j, ∂Uj/∂Ij > 0. On the other hand, non-married traditional-
ists should derive disutility from the fact of not being married, and I expect
∂Uj/∂aj = 0 (because they are not married) and ∂Uj/∂a−j, ∂Uj/∂Ij < 0.
For the second category, the non-traditionalists, I expect the following. Non-
traditionalists’ social norms do not include the prescription to marry, so they
should only marry if they have found a partner with sufficiently good match
quality. Because of better match quality, they should derive utility from their
marriage, but because they do not have the prescription to get married, they
should not derive utility from the fact of being married: ∂Uj/∂aj > 0 and
∂Uj/∂a−j, ∂Uj/∂Ij = 0. Finally, for non-married non-traditionalists, there
should be a difference between singles or divorce(e)s and cohabiting ones.
While the first two categories should not derive any relationship-related util-
ity, the cohabiting ones should. However, as they do not care about the
prescription to marry, they should not derive any disutility from their non-
married status, leading to the following predictions: ∂Uj/∂aj > 0 (if they
are in a relationship) or ∂Uj/∂aj = 0 (if they are single) and ∂Uj/∂a−j,
∂Uj/∂Ij = 0. These theoretical considerations can be summarized in the fol-
lowing table, with the expected signs of the partial derivatives written over
the arguments in the identity-augmented utility function.
7
[Insert Table 1 about here]
The next section introduces the data set and provides some interesting de-
scriptive statistics.
3 Data set and descriptive statistics
In this section, I provide a short introduction of the data set that I use for the
empirical analysis, as well as some descriptive statistics. Complete summary
statistics are provided in Appendix A.
For the empirical analysis of the relationship between marriage, identity and
happiness, I use the 2006 East Asian Social Survey (EASS). Participating
countries include the People’s Republic of China, Japan, the Republic of Ko-
rea and the Republic of China (Taiwan). Because the happiness question
was asked with a different wording in the PR China, I could use only the
observations from the latter three countries.1 After dropping observations
with missing values, the data set consists of n = 5442 individuals.
3.1 Core Independent Variables
The three core independent variables and their interactions are used to test
the theoretical predictions that marital status should have different effects
on happiness for non-traditionalists and for traditionalists, and that non-
traditionalists and traditionalists should be in marriages with different match
8
quality, reflected by different levels of marital satisfaction.
3.1.1 Marital Status
Respondents’ marital status is captured by a set of dummy variables that
take the value of 1 if the respondent is married, divorced/separated (i.e.,
living separated and intending to get a divorce) or widowed, respectively,
with singles as the baseline category.
3.1.2 Traditional Attitudes
The EASS contains eighteen questions on individuals’ atttitudes towards
marriage in general (e.g. ”It is all right for a couple to live together with-
out intending to get married”, gender roles (e.g., ”A husband’s job is to
earn money, a wife’s job is to look after the home and family”), relation-
ships inside the family (e.g., ”The authority of father in a family should be
respected under any circumstances”), and filial piety (e.g., ”Children must
make efforts to do something that would bring honor to their parents”) on
a scale from 1 to 7 (”strongly disagree” to ”strongly agree”). All questions
can be found in Appendix A. Using these questions, I construct an index
to determine which individuals are traditionalist and which individuals are
non-traditionalist. Respondents are considered to be traditionalist if they an-
swered with the most extreme category to more than two of the statements,
which corresponds to approximately 33% of the total sample.2
9
3.1.3 Marital Satisfaction
The EASS contains a question on marital satisfaction with the following
wording: ”Considering all things together, how would you describe your
marriage? Would you say that you are very satisfied or dissatisfied with your
marriage?” Respondents could answer this question on a scale from 1 to 5,
with 1 being ”completely dissatisfied” and 5 being ”completely satisfied”.
I use this question as a measure of marital satisfaction that can provide
marriage-based utility gains, leading to higher levels of reported happiness.
3.2 Control Variables
Finally, the EASS contains a series of standard background questions, e.g.,
on respondents’ employment status, education, living situation, income, and
country of residence. I use several of those variables as controls in the happi-
ness regressions. The selection of these control cariables is based on previous
reported relations in the literature on happiness research.
3.2.1 Gender
Previous research on gender differences in happiness levels has shown in-
conclusive results, but if gender differences are found, women usually report
higher levels of happiness (Diener et al. 1999). In order to control for possi-
ble differences in happiness between the sexes, I included a dummy variable
that is coded as 1 if the respondent is female.
10
3.2.2 Age
Previous research on happiness across the life span has shown inconclusive
results, with some earlier studies suggesting that happiness declines with
age (Wilson 1967), others finding a U-shaped relationship with young and
old individuals being happier than middle-aged ones (Christoph and Noll
2003, Hayo and Seifert 2003), and finally others suggesting that happiness
is constant across the life span (Diener and Suh 1998). In order to control
for the possibility that age affects happiness, I included respondents’ age as
a control variable in the regressions.
3.2.3 Self-reported health status
Previous research suggests that self-reported health status is strongly corre-
lated with happiness, with individuals who report better health status being
happier as well (Diener et al. 1999), so I included information on self-reported
health as a control variable in the regressions. All respondents were asked to
rate their health status on a scale from 1 to 5, with 1 meaning ”very good”
and 5 meaning ”very bad”. I re-coded the variable in order to assign higher
values to a better health status to make its interpretation more intuitive. The
highest value (5) now corresponds to the best self-reported health category,
i.e. ”very good”.
3.2.4 Education
Previous research on the relationship between an individuals’ level of formal
education and happiness has shown inconclusive results (see, for example,
11
Haller and Hadler 2006). In order to control for possible differences in hap-
piness among individuals with different levels of formal education, I included
the natural logarithm of the number of years of education that respondents
have received.
3.2.5 Number of family members
Household size might affect the respondent’s happiness in a positive way
because of the existence of close and supportive relationships with family
members (see, for example, Stutzer 2004). However, one could also imagine
the opposite case where a large family size affects respondent’s happiness in
a negative way because of stressful relationships with family members, more
household chores (especially if the respondent alone is responsible for them)
or because of lack of financial resources in larger households. In order to
control for the possibility that the number of family members living with
the respondent affects his or her happiness, I included information on the
number of household members as a control variable.
3.2.6 Country of residence
As the data set consists of respondents in three different countries, I include
a control variable for their country of residence. The baseline category is
Japan, and I included two dummy variables that take the value of 1 if the
respondent lives in Korea or in Taiwan, respectively.
12
3.2.7 Employment Status
Previous research suggests that paid employment has a positive effect on
individuals’ happiness (see, for instance, Clark and Oswald 1994 or Hayo
and Seifert 2003). In order to control for that fact, I included a control
variable on individuals’ employment status. Employment status is coded as
a dummy variable that takes the value of 1 if the respondent said she or he
was unemployed, and the value of 0 otherwise (including working full-time,
working part-time, being self-employed or a helping family member, being in
school, vocational, or university training, being retired, being permanently
disabled, being a homemaker and not being in the labor force).
3.2.8 Number of children
Previous research suggests that the presence of children has an impact on
individuals’ happiness (see, for example Kohler et al. 2005 or Vanassche
et al. (forthcoming)). In order to control for the possible influence of the
presence of offspring on happiness, I included information on the number of
the respondent’s children as a control variable.
3.2.9 Household income
Previous research suggests that there are small but significant positive corre-
lations between personal income and happiness (Diener et al. 1999). In order
to control for the respondent’s general financial situation, I included a set of
9 dummy variables that take the value of 1 if the respondent belongs to the
second up to the tenth decile of the respective country’s income distribution
13
respectively, with the first (lowest) income decile being the baseline category.
3.3 Descriptive Statistics
In the following, I am going to present some descriptive statistics on the
relationship between marital status and happiness and on the differences
between traditionalists and non-traditionalists.
[Insert Table 2 about here]
The first table displays the marital status of respondents and mean values
for their response to the question ”All things considered, how satisfied are
you with your life as a whole these days?”, with 1 being very dissatisfied and
5 being very satisfied. As in previous research, married individuals report
the highest levels of happiness. In addition, two interesting facts stand out.
Firstly, widowed individuals report higher levels of happiness than divorced
or separated individuals (who plan to get a divorce, but have not obtained
it yet). While both widowed and divorced or separated individuals have lost
their partner, they differ in one key aspect: widowed individuals had complied
with social norms to get married in the past, but separated or divorced
ones have broken social norms. This non-compliance with social norms can
lead to lower utility levels for the divorced individuals. Secondly, cohabiting
individuals report by far the lowest levels of happiness. While they should
also benefit from the existence of a partner much like married individuals do,
they are also potentially breaking social norms.3 For example, in the EASS
2006, only about 25% of individuals ”strongly”, ”fairly” or ”somewhat agree”
that it is okay for a couple to live together without intending to get married.
14
[Insert Table 3 about here]
This table displays differences between traditional and non-traditional indi-
viduals in terms of marital status and marital satisfaction. A higher number
of traditionalist individuals is married, and traditionalist individuals report
slightly higher levels of marital satisfaction than non-traditionalists. Finally,
traditionalists are on average about 4 years older and have about 2 years less
education than non-traditionalists. All differences are statistically significant
at the 1% level.
The next section presents the estimation strategy and estimation results.
4 Estimation strategy and results
Ferrer i Carbonell and Frijters (2004) show that assuming cardinality or or-
dinality matters only little for results in happiness research. Because of the
difficulties associated with the interpretation of interaction terms in nonlin-
ear models (Ai and Norton 2003), I estimated simple ordinary least squares
(OLS) models instead of ordered probit or logits.
In order to test the theoretical predictions about possible identity-based util-
ity gains, I augment a standard OLS happiness regression by several dummy
variables and interaction terms. In addition to dummy variables for marital
status (married, divorced, widowed), I entered a dummy variable for being a
traditionalist, and a variable measuring individuals’ self-rated marital satis-
faction (on a scale from 1 to 5, with 5 being ”very satisfied”). Finally, I also
entered interaction terms between marital status and having traditional val-
ues as well as between marital satisfaction and traditional values in order to
15
test whether the effects of the mere fact of being married and marital satisfac-
tion are indeed different for traditionalists and non-traditionalists. Similarly,
I also entered interaction terms for the marital status of being divorced and
being widowed with traditional values. Marital satisfaction serves as a mea-
sure for marriage-based utility gains (aj in equation (1)), while marital status
serves as a measure for identity-based utility gains and utility gains caused
by other individuals’ actions, such as approval or disapproval from them (Ij
and a−j in equation (1)).
Finally, the control vector Xi contains standard controls in happiness re-
search for an individual’s gender, age, self-rated health status, the log of
education years, household size, number of children, country of residence,
employment status, and the decile in the income distribution of the respec-
tive country which the respondent belongs to. This leads to the following
estimation equation.
happinessi = β1 + δ2maritalstatus i + δ3traditionali + β4maritalsatisfaction
+ δ5maritalstatus i ∗ traditionali
+ δ6maritalsatisfactioni ∗ traditionali
+ β7Xi + εi
The following table displays results from a standard OLS regression. Robust
standard errors are given in parentheses, ***, **, and * denote significance
levels of 1%, 5%, and 10%, respectively. Full regression results can be found
in Appendix B, as well as additional robustness checks using only subsamples
(men vs. women and separate regressions for the three countries included in
16
the sample).
[Insert Table 4 about here]
The first interesting result is that the simple fact of being married decreases
happiness for non-traditionalists, while marital satisfaction increases happi-
ness for them. This provides empirical evidence for the existence of benefits
from marriage in terms of specialization of partners, mutual support and com-
panionship, or the provision of services for which there are no or imperfect
markets (Kohler et al. 2005). It also provides evidence for the hypothesis
that non-traditionalists only marry if they find a partner with whom the
match quality is sufficiently good, but not for the mere fact of being married.
The next interesting finding is that very traditional individuals report higher
levels of happiness. A tentative explanation might be that traditionalists’
values are associated with less questioning about the meaning of life and,
consequently, higher levels of reported satisfaction.
As predicted by the theoretical considerations, the simple fact of being mar-
ried increases utility for traditionalists, but the effect is not statistically sig-
nificant. A Wald test for the joint significance of the two variables married
and married ∗ traditional, however, shows that they are highly significant
together (F (2, 5413) = 125.31, prob > F = 0.0000). This finding suggests
that there is indeed an additional identity-based utility gain from the fact of
being married for traditionalist individuals.
As predicted by the theoretical considerations, more traditional individuals
seem to be less satisfied with their marriages, as shown by the negative effect
of the interaction term traditional∗maritalsatisfaction. The reason for this
17
might be the lower match quality of traditionalists’ marriages, compared to
non-traditionalists.
For widowed and divorced individuals, the theoretical predictions are not
confirmed by my estimations. No statistically significant effects are found
for the effect of widowhood on happiness. For divorced individuals, I find
that non-traditional divorce(e)s are less happy than singles. This suggests
that even though they probably do not to care about social norms, non-
traditionalists suffer from divorce-related stress.
In addition, I carried out robustness checks (see Table 7 in Appendix B)
using various subsamples and running regressions for men, for women and
for each of the three countries separately. The results for marital satis-
faction and the simple fact of being married remain unchanged in all five
estimations. As in the regression for the whole sample, the interaction term
married ∗ traditional is positive albeit insignificant, but again, Wald tests
show that the two variables married and married ∗ traditional are always
highly significant together. This provides further evidence for identity-based
happiness gains from the simple fact of being married for traditionalist in-
dividuals. Finally, the interaction term traditional ∗maritalsatisfaction is
still negative but becomes insignificant in all robustness checks.
The control variables show some interesting results as well. Women are found
to be happier than men in the three East Asian countries that I analyzed.
This goes in line with previous research findings: if gender differences are
found, then usually women are found to be happier than men (see, for ex-
ample, Diener et al. 1999). Next, older individuals are found to be happier,
which is not a standard result in the previous literature. The higher happi-
18
ness levels of older individuals might be explained by the high levels of stress
that most individuals in East Asian countries experience during their school
and work years (see, for example, Lee and Larson 1999 for the case of Ko-
rean students, Park et al. 2008 for the high levels of job stress and depression
among Korean employees, and Spector et al. 2001 for the higher psycholog-
ical strains of Japanese and Taiwanese employees, compared to American
employees) and resulting higher levels of happiness once ”examination hell”
and stressful work life are over.
Third, as reported in the previous literature, individuals in better health sta-
tus are also happier, and unemployed individuals are less happy than those
working full- or part-time, being in school, being a homemaker, or retired. I
do not find any significant relationship between household size or the number
of one’s own children and respondents’ happiness. More educated women are
happier, but there is no such effect for male respondents in the sample.
Fourth, respondents in Korea are found to be significantly less happy than
those in Japan, and respondents in Taiwan are found to be significantly hap-
pier than those in Japan. Ohayon and Hong (2006) present results on the
prevalence of major depressive disorder in Korean adults and find that about
20% of respondents in their sample suffer from depression symptoms. They
also resport that only few respondents seek help for their symptoms, and
very few received appropriate treatment for their condition. This might be
part of the reason why Korean respondents are significantly less happy than
those in the other two countries.
Fifth, I find positive effects of income on happiness starting from the upper
half of the income distribution (i.e., the sixth decile of the income distribution
19
and above). Again, this is in line with previous research reporting positive
correlations between income and happiness (see, for example, Diener et al.
1999).
5 Conclusion
This paper presented an empirical test of the prediction that marriage pro-
vides utility not only through marital satisfaction, but also through identity-
based utility gains for traditionalist individuals in East Asian countries. The
reason are probable stronger prescriptions to get married among more tradi-
tionalist individuals, and hence the possibility of identity-based utility gains
of marriage for traditionalists. At the same time, traditionalists should, on
average, be less satisfied with their marriage because they should put more
emphasis on the fact of being married and less emphasis on the quality of
the match with their spouse.
The empirical findings show indeed that marital satisfaction increases hap-
piness for non-traditionalists, while traditionalists are significantly less satis-
fied with their marriage. This suggests that they are in marriages with worse
match quality, compared to non-traditionalists.
I find a positive effect of the mere fact of being married for traditionalists.
Although it is not statistically significant, the joint significance of the two
variables measuring an individual’s traditional attitudes and an interaction
term between marital status and traditional attitudes could not be rejected
in a Wald test. This suggests that there are indeed identity-based utility
gains from marriage for traditionalist individuals in East Asian countries.
20
A last interesting result is the fact that when controlling for marital sat-
isfaction, the mere fact of being married decreases happiness in a statis-
tically significant way. This suggests that marital satisfaction (or maybe
love, actually), is responsible for the happiness effects of marriage for non-
traditionalists.
However, there are also several limitations to this study. The first one con-
cerns the lack of information in the dataset used for my empirical analysis. It
is, for example, likely that both overall satisfaction with life or happiness and
marital satisfaction are driven by personality traits, but there is no informa-
tion on personality traits in the EASS, as in most other social surveys and
comparable datasets. The positive relationship between marital satisfaction
and happiness should therefore be seen as a correlation and not as a causal
effect.
The second one follows from the fact that the EASS is a cross-sectional
dataset. It could for example be that people who are more satisfied with life
in general are more likely to get married and also more satisfied with their
marriage. Longitudinal datasets would offer an opportunity to overcome this
limitation following from a selection process into marriage.
Finally, a separate analysis of cohabiting couples could provide more insights
into the possible differences between cohabitation and marriage and their
effects on happiness, but the very low number of cohabiting couples in the
sample made an analysis of this question impossible.
21
Notes1In China, respondents were asked ”All things considered, how happy are you with
your life as a whole these days?”, and in the other three countries, ”All things considered,how satisfied are you with your life as a whole these days?”.
2Results do not change if I use the index instead of this dummy, but as interpretationis more intuitive for the dummy variable, I decided to use the dummy.
3However, as the number of cohabiting individuals in the sample is very low, I am notgoing to analyze them in more detail in the estimations.
22
References
Ai, C., & Norton, E. C. (2003). Interaction terms in logit and probit models.Economics Letters , 80 , 123-129.
Akerlof, G. A., & Kranton, R. E. (2000). Economics and identity. QuarterlyJournal of Economics , 115(3), 715-753.
Chang, H.-C. (1997). Language and Words: Communication in the Analectsof Confucius. Journal of Language and Social Psychology , 16 , 107-131.
Christoph, B., & Noll, H.-H. (2003). Subjective well-being in the EuropeanUnion during the 90s. Social Indicators Research, 64 , 521-546.
Clark, A. E., & Oswald, A. J. (1994). Unhappiness and unemployment. TheEconomic Journal , 104 (424), 648-659.
Diener, E., & Suh, E. (1998). Age and subjective well-being: An internationalanalysis. Annual Review of Gerontology and Geriatrics , 17 , 304-324.
Diener, E., E.M. Suh, R.E. Lucas, and H.L. Schmidt. (1999). Subjectivewell-being: Three decades of progress. Psychological Bulletin, 125(2),276-302.
Ferrer-i Carbonell, A., & Frijters, P. (2004). How important is methodologyfor the estimates of the determinants of happiness? The EconomicJournal , 114 , 641-659.
Frey, B. S., & Stutzer, A. (2002). The economics of happiness. WorldEconomics , 3(1), 1-17.
Haller, M., & Hadler, M. (2006). How social relations and structures can pro-duce happiness and unhappiness: An international comparative analy-sis. Social Indicators Research, 75 , 169-216.
Hayo, B., & Seifert, W. (2003). Subjective economic well-being in EasternEurope. Journal of Economic Psychology , 24 , 329-348.
Jones, G. W. (2004). Not ”When to Marry” but ”Whether to Marry”: thechanging context of marriage decisions in East and Southeast Asia. InG. W. Jones & K. Randas (Eds.), (Un)Tying the knot: ideal and realityin Asian marriage (p. 3 - 58). Singapore: Singapore University Press.
Kohler, H.-P., J. R. Behrman, and A. Skytthe. (2005). Partner + children= happiness? The effects of partnerships and fertility on well-being.Population and Development Review , 31(3), 407-445.
Lee, K.-K. (1989). The practice of traditional family ritual in contemporaryurban Korea. Journal of Ritual Studies , 3 , 167-183.
Lee, M., & Larson, R. (2000). The Korean ”Examination Hell”: Long hoursof studying, distress, and depression. Journal of Youth and Adoles-cence, 29 , 249-271.
Morgan, S. P., & King, R. B. (2001). Why have children in the 21st century?Biological predisposition, social coercion, rational choice. European
23
Journal of Population/Revue europeenne de Demographie, 17(1), 3-20.Ohayon, M. M., & Hong, S.-C. (2006). Prevalence of major depressive dis-
order in the general population of South Korea. Journal of PsychiatricResearch, 40 (1), 30 - 36.
Park, S.-G., Min, K.-B., Chang, S.-J., Kim, H.-C., & Min, J.-Y. (2009).Job stress and depressive symptoms among Korean employees: theeffects of culture on work. International Archives of Occupational andEnvironmental Health, 82 , 397-405.
Spector, P. E., Cooper, C. L., Sanchez, J. I., O’Driscoll, M., Sparks, K.,Bernin, P., . . . Yu, S. (2001). Do national levels of individualismand internal locus of control relate to well-being: an ecological levelinternational study. Journal of Organizational Behavior , 22 (8), 815–832.
Stutzer, A. (1994). The role of income aspirations in individual happiness.Journal of Economic Behavior and Organization, 54(1), 89 - 109.
Tang, Z. (1995). Confucianism, Chinese culture, and reproductive behavior.Population and Environment: A Journal of Interdisciplinary Studies ,16 (3), 269-284.
Tu, W.-M. (1996). Confucian Traditions in East Asian Modernity: MoralEducation and Economic Culture in Japan and the Four Mini-Dragons.Cambridge, MA: Harvard University Press.
Vanassche, S., Swicegood, G., & Matthijs, K. (in press). Marriage andchildren as a key to happiness? Cross-national differences in the effectsof marital status and children on well-being. Journal of HappinessStudies .
Wilson, W. (1967). Correlates of avowed happiness. Psychological Bulletin,22 , 520-537.
24
6 Tables
Table 1: Categories and choices
Complier: married/widowed Non-Complier: single/divorced/cohabiting
Ca (Traditionalist) U(
0︷︸︸︷aj ,
+︷︸︸︷a−j ,
+︷︸︸︷Ij ) U(
0︷︸︸︷aj ,
−︷︸︸︷a−j ,
−︷︸︸︷Ij )
Cb (Non-traditionalist) U(
+︷︸︸︷aj ,
0︷︸︸︷a−j ,
0︷︸︸︷Ij ) U(
+,0︷︸︸︷aj ,
0︷︸︸︷a−j ,
0︷︸︸︷Ij )
Table 2: Marital status and happiness levels
Percent Happiness
married, living as married 70.07% 3.5202widowed 8.29% 3.4035divorced 4.17% 3.1165separated, but married 0.39% 3.0476single, never married 17.48% 3.3565cohabiting 0.13% 2.7143
Total 5442
Table 3: Characteristics of traditionalists and non-traditionalists
Traditionalist Non-Traditionalist Two-sample t-test statistics
Married 72.86% 68.74% 3.1542Marital satisfaction 2.87 2.54 5.8304Age 51.62 47.46 8.8765Years of education 10.20 12.29 16.8042
n 1754 3688 df = 5440
25
Table 4: Marriage, traditional values, and happiness: OLS regression results
Happiness
Marital satisfaction 0.399***[0.021]
1 = married -1.353***[0.094]
1 = very traditional 0.169**[0.074]
traditional * married 0.121[0.168]
traditional * marital satisfaction -0.063*[0.037]
1 = widowed -0.027[0.081]
traditional * widowed 0.145[0.118]
1 = divorced -0.218***[0.083]
traditional * divorced -0.015[0.168]
Observations 5442R-squared 0.27
26
7 Appendix A
7.1 Complete summary statistics
Table 5: Complete summary statistics
Variable Obs Mean Std. Dev. Min Max
Happiness 5442 3.4683 0.9534 1 5Marital satisfaction 5442 2.6477 1.9104 0 5Married 5442 0.7007 0.4580 0 1Very traditional 5442 0.3223 0.4674 0 1Widowed 5442 0.0829 0.2757 0 1Divorced 5442 0.0417 0.2000 0 1Female 5442 0.5382 0.4986 0 1Age 5442 48.8014 16.2780 18 92Health status 5442 3.5932 1.0600 1 5Log of education years 5442 2.3487 0.5784 0 3.1355Household size 5442 3.8307 1.7150 1 16Korea 5442 0.2585 0.4379 0 1Taiwan 5442 0.3598 0.4800 0 1Unemployed 5442 0.0447 0.2066 0 1Number of children 5442 1.9131 1.4616 0 12Second income decile 5442 0.1130 0.3166 0 1Third income decile 5442 0.1152 0.3193 0 1Fourth income decile 5442 0.0362 0.1868 0 1Fifth income decile 5442 0.0871 0.2820 0 1Sixth income decile 5442 0.0742 0.2622 0 1Seventh income decile 5442 0.0752 0.2637 0 1Eighth income decile 5442 0.0882 0.2836 0 1Ninth income decile 5442 0.0741 0.2619 0 1Tenth income decile 5442 0.0707 0.2564 0 1
7.2 Questions used for constructing the TraditionalValues Index
• If you were to have only one child, would you prefer a boy or a girl? (boy:traditional)
• Who do you think is most responsible for taking care of old parents? (eldestson: traditional)
To what extent do you agree or disagree with each of the following statements?
• Husband should be older than wife (completely agree: traditional)
• It is not necessary to have children in marriage (completely disagree: tradi-tional)
• Married men are generally happier than unmarried men (completely agree:traditional)
27
• Married women are generally happier than unmarried women (completelyagree: traditional)
• It is all right for a couple to live together without intending to get married(completely disagree: traditional)
• People who want to divorce must wait until children are grown up (com-pletely agree: traditional)
• Divorce is usually the best solution when a couple can’t seem to work outtheir marriage (completely disagree: traditional)
• It is more important for a wife to help her husband’s career than to pursueher own career (completely agree: traditional)
• A husband’s job is to earn money; a wife’s job is to look after the home andfamily (completely agree: traditional)
• Men ought to do a larger share of household work than they do now (com-pletely disagree: traditional)
• During economic recession, it is all right for women to be laid-off prior tomen (completely agree: traditional)
• The authority of father in a family should be respected under any circum-stances (completely agree: traditional)
• Children must make efforts to do something that would bring honor to theirparents (completely agree: traditional)
• The eldest son should inherit a larger share of the property (completelyagree: traditional)
• A child who has taken good care of parents should inherit a larger share ofthe property (completely agree: traditional)
• To continue the family line, one must have at least one son (completelyagree: traditional)
• If husband’s family and wife’s family need help at the same time, a marriedwoman should help husband’s family first (completely agree: traditional)
• One must put familial well-being and interest before one’s own (completelyagree: traditional)
28
8 Appendix B
Table 6: Marriage, traditional values, and happiness: Full results
Happiness
Marital satisfaction 0.399***[0.021]
1 = married -1.353***[0.094]
1 = very traditional 0.169**[0.074]
traditional * married 0.121[0.168]
traditional * marital satisfaction -0.063*[0.037]
1 = widowed -0.027[0.081]
traditional * widowed 0.145[0.118]
1 = divorced -0.218***[0.083]
traditional * divorced -0.015[0.168]
1 = female 0.211***[0.023]
age 0.004***[0.001]
health 0.214***[0.013]
log of education years 0.034[0.028]
number of family members -0.006[0.008]
1 = Korea -0.299***[0.033]
1 = Taiwan 0.166***[0.031]
1 = unemployed -0.152**[0.060]
Number of kids -0.004[0.013]
Second income decile -0.002[0.044]
Third income decile -0.023[0.044]
Fourth income decile 0.049[0.062]
Fifth income decile 0.069[0.045]
Sixth income decile 0.132***[0.046]
Seventh income decile 0.085*[0.046]
Eighth income decile 0.191***[0.043]
Ninth income decile 0.254***[0.047]
Tenth income decile 0.326***[0.047]
Constant 1.922***[0.122]
Observations 5442R-squared 0.27
Table 7: Robustness checks: subsamples
Women Men Japan Korea Taiwan
Marital satisfaction 0.412*** 0.376*** 0.384*** 0.401*** 0.356***[0.027] [0.035] [0.027] [0.040] [0.051]
1 = married -1.393*** -1.267*** -1.198*** -1.351*** -1.246***[0.118] [0.154] [0.124] [0.174] [0.215]
1 = very traditional 0.162 0.176* 0.138 -0.013 0.178*[0.116] [0.098] [0.180] [0.136] [0.104]
traditional 0.005 0.177 0.091 0.408 0.056* married [0.223] [0.264] [0.406] [0.262] [0.314]traditional -0.024 -0.084 -0.011 -0.095 -0.042* marital satisfaction [0.048] [0.059] [0.087] [0.060] [0.072]1 = widowed -0.046 0.134 0.17 0.058 -0.194
[0.096] [0.171] [0.107] [0.216] [0.163]traditional 0.162 0.079 0.128 0.278 0.391*** widowed [0.153] [0.247] [0.273] [0.255] [0.191]1 = divorced -0.158 -0.293** -0.200* -0.252 -0.054
[0.109] [0.123] [0.118] [0.244] [0.131]traditional 0.022 -0.032 0.381 0.041 -0.191* divorced [0.251] [0.228] [0.484] [0.347] [0.242]age 0.006*** 0.003* 0.006*** -0.003 0
[0.001] [0.002] [0.001] [0.003] [0.002]health 0.211*** 0.223*** 0.311*** 0.116*** 0.188***
[0.016] [0.020] [0.020] [0.024] [0.021]log of educ. years 0.069* -0.028 -0.117 0.045 -0.027
[0.036] [0.046] [0.097] [0.059] [0.037]number of 0.003 -0.015 0.018 -0.049** -0.015family members [0.011] [0.012] [0.014] [0.023] [0.012]1 = Korea -0.252*** -0.344***
[0.044] [0.052]1 = Taiwan 0.228*** 0.088*
[0.041] [0.047]1 = unemployed -0.08 -0.247*** -0.429*** -0.161 -0.061
[0.079] [0.092] [0.135] [0.112] [0.080]Number of kids -0.003 0 -0.027 0.041 0.026
[0.016] [0.020] [0.021] [0.028] [0.020]Second income decile 0.023 -0.057 0.006 0.03 0.006
[0.060] [0.065] [0.069] [0.115] [0.071]Third income decile 0.063 -0.141** -0.045 0.057 0.042
[0.057] [0.068] [0.066] [0.109] [0.084]Fourth income decile 0.132 -0.058 0 0.315** 0.04
[0.080] [0.093] [0.000] [0.151] [0.081]Fifth income decile 0.102* 0.013 0.001 0.322*** 0.036
[0.057] [0.072] [0.063] [0.120] [0.085]Sixth income decile 0.188*** 0.041 0.134** 0.216* 0.189**
[0.057] [0.074] [0.059] [0.122] [0.086]Seventh income decile 0.079 0.083 0.037 0.326*** 0.008
[0.060] [0.073] [0.067] [0.116] [0.094]Eighth income decile 0.216*** 0.142** 0.055 0.412*** 0.294***
[0.057] [0.065] [0.057] [0.144] [0.079]Ninth income decile 0.265*** 0.228*** 0.173** 0.458*** 0.253***
[0.060] [0.074] [0.083] [0.115] [0.085]Tenth income decile 0.374*** 0.257*** 0.279*** 0.545*** 0.342***
[0.063] [0.073] [0.077] [0.126] [0.079]Constant 2.080*** 2.463*** 2.128*** 2.545*** 2.843***
[0.149] [0.180] [0.297] [0.250] [0.168]
Observations 2929 2512 2076 1407 1958R-squared 0.31 0.23 0.3 0.22 0.18Wald test F(2,2485) = 41.90 F(2,2902) = 85.17 F(2,2052) = 50.32 F(2,1382) = 35.33 F(2,1933) = 27.36Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000