Who Benefits from Marriage?*
| Published date | 01 February 2009 |
| Author | Dipanwita Sarkar,Esfandiar Maasoumi,Daniel L. Millimet |
| Date | 01 February 2009 |
| DOI | http://doi.org/10.1111/j.1468-0084.2008.00515.x |
1
©Blackwell Publishing Ltd and the Department of Economics, University of Oxford, 2008. Published by Blackwell Publishing Ltd,
9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 71, 1 (2009) 0305-9049
doi: 10.1111/j.1468-0084.2008.00515.x
Who Benefits from Marriage?*
Esfandiar Maasoumi†, Daniel L. Millimet‡ and
Dipanwita Sarkar§
†Departmentof Economics, Emory University, Atlanta, GA 30322-2240 and School of Business
and Economics, Swansea University, Wales, UK (e-mail: esfandiar.maasoumi@emory.edu)
‡Department of Economics, Southern Methodist University, Dallas, TX 75275-0496, USA
(e-mail: millimet@mail.smu.edu)
§Department of Economics, University of Louisiana-Monroe, ADMN 2-31,
700 University Avenue, Monroe, LA 71209, USA (e-mail: sarkar@ulm.edu)
Abstract
The phenomenon that married men earn higher average wages than unmarried men –
the marriage premium – is well known. However, the robustness of the premium
across the wage distribution and the underlying causes of the marriage premium are
unclear. Focusing on the entire wage distribution and employing recently developed
semi-non-parametric tests for quantile treatment effects, our findings cast doubt on the
robustness of the premium. Wefind that the premium is explained by selection above
the median, whereas a positive premium is obtained only at very low wages. The
causal effect at low wages may be attributable to employer discrimination.
I. Introduction
Motivation
Married men, on average, earn more than single men in the labour market, seemingly
even after controlling for some observable attributes, a phenomenon commonly
referred to as the marriage premium (MP). The current notion of the MP is rather
widely held as is exemplified by Loh (1996, p. 566): ‘Virtually all cross-sectional
wage studies find that currently married men typically earn a higher wage rate than
their unmarried counterparts in the labour market.’ Cornwell and Rupert (1997,
p. 285) similarly note: ‘Married men earn more than unmarried men. This fact is
unassailable and is robust across data sets and over time.’
ÅThe authors wish to thank an anonymous referee for many helpful comments.
JEL Classification numbers: C14, C21, C33, D31, J12.
2Bulletin
While the existence of the MP (as defined) may not be controversial, one salient
question has not been asked (to our knowledge) and another has yet to be answered.
First, to what extent does the MP hold across the entire wage distribution, and is it
uniform? Second, what are the underlying sources of the MP? Answering the former
question is vital, for if some men do not benefit, or benefit to different degrees, a
narrow focus on an average measure of the MP creates a false sense of robustness,
and may be inconsistent with a uniform ranking of wage distributions across
reasonable classes of utility functions. With respect to the latter question, Cornwell
and Rupert (1997, p. 285) continue: ‘While there is compelling evidence that
married men earn more than unmarried men, the source of this premium remains
unsettled.’ More recently, Stratton (2002, p. 199) states: ‘Research has failed as yet
to reach a consensus regarding the nature of these differentials.’
In this paper, we hope to shed light on these questions in two important ways.
First, we revisit the claim that the MP is robust by assessing the effects of marriage
across the entire wage distribution utilizing recent tests for quantile treatment effects
(QTE) and stochastic dominance (SD). The former allows one to test for statistically
significant effects of marriage on particular quantiles of the wage distribution, rather
than just the mean. The latter provides a means of aggregating information across
the quantiles in order to make summary statements about how one distribution com-
pares to another. Thus, the distributional approach utilizes all available information,
assesses the potentially heterogeneous returns to marriage across subgroups of men
and makes explicit the welfare framework being employed by different observers
who may place different weights on different subgroups. It is useful to appreciate the
link between quantile analysis and various orders of dominance between outcome dis-
tributions. A joint quantile analysis is equivalent to a test for first-order dominance,
which implies higher orders of dominance as well. If quantile analysis is inconclusive,
as it is in this and many other settings, one has effectively reached the upper limits
of the usefulness of decision-making by quantiles alone. A decision to rank two het-
erogenous outcomes is necessarily based on (supported by) evaluation functions that
put different weights on different quantiles and population groups. This puts them in
the domain of second and higher order stochastic dominance analysis.1Second, we
revisit and distinguish the underlying explanations of the return to marriage within
this broader distributional analysis of the MP.
Our findings are rather striking and support four conclusions. First, the traditional
notion of the MP persists at the distributional level in the CPS data in the 1990s, even
after controlling for a lengthy list of observable attributes. However, there is some
evidence that the return to marriage may not be uniform across the wage distribution;
the gains from marriage are largest for those in the lower tail of the wage distribution.
1The richness of the SD analysis has led to their growing application. For example, Maasoumi and
Millimet (2005) examine changes in US pollution distributions over time and across regions at a point in
time. Maasoumi and Heshmati (2000) analyse changes in the Swedish income distribution over time as well
as across differentpopulation subgroups. Abadie (2002) analyses the impact of veteran status on the distribution
of civilian earnings.
©Blackwell Publishing Ltd and the Department of Economics, University of Oxford 2008
Who benefits from marriage? 3
Second, the MP is no longer statistically significant above roughly the median once
selection on time invariant attributes is admitted, but the MP does persist in the lower
tail of the distribution.
Third, controlling for time invariant and time-varying unobservables correlated
with marital status and labour market performance further amplifies this result; the
MP is at-most a phenomenon associated with the extreme lower tail of the wage
distribution, and the MP may even be negative in the upper tail. This implies that
typical fixed effects methods alone – which are prominent in the MP literature – may
be insufficient to give a complete picture of the returns to marriage. In addition, our
results draw attention to the fact that statistically significant estimates of the average
MP are really capturing effects occurring only at low quantiles of the wage distribu-
tion. Finally, the disappearance of the MP over the majority of the distribution once
unobservables are addressed suggests an important role of selection in explaining the
MP. However, the MP in the lower tail does appear to represent a causal relation-
ship. Moreover, given the strong negative correlation between married women’s
labour supply and husband’s wage, the MP in the lower tail is not likely driven
by household specialization, but rather may reflect employer discrimination in low
wage labour markets (e.g. Blau and Kahn, 2005). In other words, in such labour
markets, the absence of other signals (e.g. schooling) to differentiate among workers
implies a large weight placed by employers on a worker’s marital status.
To reach these conclusions, we begin by using panel data from the Current Pop-
ulation Survey (CPS) for 1992–2001. We offer the following comparisons of uncon-
ditional distributions: (i) single vs. ‘to-be-married’ men (men who marry within the
next year), (ii) single vs. newly married men, and (iii) the distribution of wage changes
for single vs. newly married men. Next, we re-examine the same pairwise compar-
isons after adjusting for a host of observable covariates that may be correlated with
both marital status and/or labour market performance. This two-part strategy enables
us to (i) support (or refute) the existence of a uniform MP, and (ii) comment on the
underlying causes of the MP (assuming it exists). Specifically, if the MP represents
a causal relationship (with marriage affecting wages), then there should be no differ-
ence in the distribution of wages between single and ‘to-be-married’ men. But the
distribution of married men should ‘dominate’ that of single men. On the other hand,
a ‘dominant’ distribution of wages amongst ‘to-be-married’men would suggest a role
for both selection and reverse causation explanations. If there is some validity to all
the proposed explanations, then we might observe a modest disparity in the distribu-
tion of wages favouring ‘to-be-married’men (vs. singles), followed by an even greater
disparity after marriage. Thus, our analysis can help assess the relative role of causal
vs. correlation-based explanations of the MP across the entire wage distribution.
The above strategy separates the causal and selection components of the MP
through comparisons of wage changes before and after marriage, thereby eliminating
time invariant unobservables as in fixed effects methods. An alternative strategy is
to use an instrumental variable (IV) for marital status. The advantage of employing
an IV strategy is that it controls for time-varying unobservables that may be
©Blackwell Publishing Ltd and the Department of Economics, University of Oxford 2008
To continue reading
Request your trial