A Treatise on
the Family
Gary S. Becker (1981)
"If participants in marriage markets have complete information about all
prospects, divorce would be a fully anticipated response to a demand
for variety in mates or to life-cycle changes in traits. Most divorces
would then occur after many years of marriage, because traits change
gradually. The facts, however, suggest the opposite: about 40 percent of all divorces (and annulments) occur prior to the fifth year of marriage,
and separation usually precedes divorce by a year or more (U.S.
Department of Health, Education, and Welfare, 1979).
If, however, participants had highly imperfect information, most divorces
would occur early in marriage by virtue of the fact that information
about traits increases rapidly after marriage. Several years of marriage
is usually a far more effective source of information on love and
many other traits than all the proxies available prior to marriage. I
suggest that marriages fail early primarily because of imperfect information
in marriage markets and the accumulation of better information
during marriage. This suggestion is supported by the fact that unexpected
changes in earnings and health do raise the probability of divorce
(BLM,5 1977).
Women who divorced early in their marriage report that "difficult" spouses and value conflicts were major sources of their discontent,
presumably because these traits are much better assessed after a few
years of marriage. Personality conflict, sexual incompatibility, and similar
traits should be less important sources of later than of earlier divorces;
little additional information about these traits is acquired after
a few years of marriage. On the other hand, some information, including
information about other women and about earnings potential,
is acquired more slowly and should be more important in later divorces.
Indeed, another woman and/or financial conflict are frequently
cited by women divorcing after ten years of marriage (Goode, 1956, pp.
128-129).
The major sources of discontent and divorce are not necessarily the
major determinants of marital well-being. Education, age, physical
appearance, and other easily assessed traits are not major sources of
discontent because not much more is learned about them after marriage.
Just as the emphasis on easily assessed traits in marriage
markets does not imply that these traits contribute more to marital
well-being than other traits, neither does the opposite emphasis on
difficult-to-assess traits in "divorce markets" imply that those contribute
more.
The more rapid accumulation of information during the first few
years of marriage implies that divorce is more likely early in marriage
than later. Divorce rates are highest during the first few years of marriage
and decline steeply after four or five years, although the explanation
is partly that those most prone to divorce tend to drop out early
from the cohort of married persons (see Heckman, 1981, on the effects
of heterogeneity).
Divorce is less likely later in the marriage for the additional reason
that capital accumulates and becomes more valuable if a marriage stays
intact ("marital-specific" capital). Children are the prime example, especially
young children, although learning about the idiosyncrasies of
one's spouse is also important (Heimer and Stinchcombe, 1979). Divorce
is much less likely when there are children, especially young
children-not only in the United States and other rich countries
(Goode, 1963, pp. 85, 364; BLM, 1977), but also in primitive societies
(Saunders and Thomson, 1979).
The accumulation of marital-specific capital is, in turn, discouraged
by the prospect of divorce because, by definition, such capital is less
valuable after a divorce. Presumably, trial or consensual marriages
produce fewer children than legal marriages at least partly because the
former are less durable (see the evidence in Kogut, 1972, on consensual
and legal marriages in Brazil). Persons who marry outside their
race or religion are far more likely to divorce than are others with similar
measurable characteristics. Therefore, we can readily understand
why marriages between persons of different races or religions have significantly
fewer children even when intact marriages are compared
(see the evidence for the United States in BLM, 1977), and why marriages
between Indians of different castes have fewer children than
marriages within a caste (Das, 1978).
Expectations about divorce are partly self-fulfilling because a higher
expected probability of divorce reduces investments in specific capital
and thereby raises the actual probability. For example, consensual and trial marriages are less stable than legal marriages, and marriages
between persons of different religions or races are less stable than
those within a religion or race, partly because mixed marriages have
fewer children. At the same time, as indicated, mixed marriages have
fewer children partly because they are expected to be less stable.
Specific investment and imperfect information can explain why
homosexual unions are much less stable than heterosexual marriages
(Saghir and Robins, 1973, pp. 56-58,226-227). Homosexual unions do
not result in children, and generally they have a less extensive division
of'labor and less marital-specific capital than heterosexual marriages.
Moreover, the opprobrium attached to homosexuality has raised the
cost of search to homosexuals and thereby has reduced the information
available to them. Furthermore, homosexual unions, like trial marriages,
can dissolve without legal adversary proceedings, alimony, or
child support payments.
Women have usually married earlier than men partly because the
maturation and independence of men has been delayed by greater investments
in their human capital. Since investments in men and
women have become more equal over time as the demand for children
has decreased (see Chapter 3), men and women now marry for the first
time at rather similar ages. For example, the difference in the United
States between the median age at first marriage of men and women declined
from four years in 1900 to about two and a half years in 1970
(U.S. Bureau of the Census, 1971c).
Yet divorced women have remarried more slowly than divorced
men even when divorced at young ages. They almost always receive
custody of children, a factor that discourages remarriage. For the same
reason, women with illegitimate children marry for the first time more
slowly than women without children (Berkov and Sklar, 1976).
Young children raise the cost of searching for another mate and significantly
reduce the net resources of divorced women (Weitzman and
Dixon, 1979). Possibly for these reasons they raise the probability that remarriage will fail, even though children born during the remarriage
lower this probability (BLM, 1977). It is noteworthy that illegitimate
children and other pregnancies prior to first marriage also raise the
probability of marital failure (Christensen and Meissner, 1953; Berkov
and Sklar, 1976) Divorced women might well remarry earlier than divorced men, just
as single women without children marry earlier than single men, if
divorced women did not receive custody of children. Indeed,
perhaps 45 percent of divorced women would have remarried within
the first two years of their divorce if they did not have custody, which
is double their actual percentage (22) and considerably higher than the
percentage for men (31). This estimate assumes that women without
custody marry as rapidly as women without children. It is based on a
regression equation that relates whether a woman remarries within a
specified period of time to several variables, including number of children
(BLM, 1977)."
