12 minute read

Marital Quality

Bases Of Marital Quality

Marriage counselors, ministers, various custodians of the folklore, and perhaps even family social scientists and psychologists may possess a great deal of wisdom about how to achieve and maintain happy and satisfying marriages. Many people probably know a great deal about how to achieve a happy marriage, but that knowledge is based hardly at all on systematic research. In spite of the enormous amount of research devoted to the topic, truly scientific evidence on the bases of marital happiness and satisfaction is meager. One reason may be that marital quality is inherently hard to study, but a more certain reason is that most of the research has been seriously deficient.

Consider, for instance, the many cross-sectional studies of samples of married persons in which various demographic and social variables have been related to the respondents' reports of their marital happiness or satisfaction. These studies have amassed a large body of evidence on the correlates of the measures of marital quality and have often inferred cause and effect from the correlational data. Even though some of the studies have used apparently sophisticated causal modeling, virtually none of the research has met the requirements for valid causal inference. An inherent limitation of studies concentrating on currently married persons, especially in a society with very high divorce rates, is that many of the persons among whom negative influences on marital quality have been the strongest have been selected out FIGURE 1 of the sampled population through divorce. Therefore, the effects of the negative influences will be underestimated or not detected at all in analyses of data on married persons, or positive effects may even be attributed to the influences.

There are two aspects of the problem. First, divorce lessens the variance of marital quality, thus attenuating estimates of effects on that variable. Second, if variables that affect marital quality affect the probability of divorce in ways other than via marital quality, the estimates of the effects can be substantially biased in either direction. A hypothetical case will illustrate the point. Suppose that celebrity status typically has negative effects on marital quality and that it increases the probability of divorce (by providing numerous alternatives to the current marriage) at each level of marital quality. If the latter effect is very strong and the former is relatively weak, a causal analysis with cross-sectional data that does not take the selection factor into account will estimate the effect of celebrity status on marital quality to be positive. Suppose that adherence to a particular religious ideology typically exerts positive influence on marital quality but decreases the probability of divorce at each level of marital quality. In this case, the straightforward testing of a simple causal model may indicate a negative effect of religious adherence on marital quality, or at least it will yield an attenuated estimated positive effect.

It is theoretically possible, of course, to incorporate selection processes into causal models to obtain unbiased estimates of effects (Berk 1983; Heckman 1979), but the methods developed to do this have been criticized as inadequate, and in any event, they require information the researcher rarely has. Given the limitations of corrections for "sample selection bias," it is hardly lamentable that they have rarely been used in studies of the bases of marital quality.

A more promising solution to the problems posed by the use of cross-sectional data from samples of married individuals to infer effects on marital quality is to use longitudinal data or cross-sectional data from samples including formerly married as well as currently married individuals. Except in those rare studies in which marital quality has been assessed at frequent intervals before divorces have occurred, such research requires the assumption that marital quality becomes low before divorce occurs, but that assumption seems generally justified. Overall, longitudinal data are, of course, better for the purpose at hand, but cross-sectional data can be used to assess the effects of background variables that can be measured with reasonable accuracy through retrospective reports. Unfortunately, research with these designs has been rare, and such research with data from large and representative national samples is only in its infancy. As it increases, a substantial increment in credible evidence on the effects of several demographic and social variables on marital quality can be expected.

In the meantime, the evidence on the bases of marital happiness and marital satisfaction from "mainstream" quantitative social science consists almost entirely of a body of weak correlations and estimates of weak effects. The lack of strong relationships no doubt results in large measure from some of the most important influences on marital quality not being amenable to measurement on large-scale surveys and thus not being taken into account. Individuals' feelings about their marriages are also subject to considerable short-term fluctuation, and the variance resulting from this fluctuation will not be explained by the major long-term influences on marital quality. There is also the attenuation of measured relationships resulting from the selection of many of the people with the least satisfactory marriages out of the population of married persons.

These limitations of the evidence do not mean that it is worthless, but only that it should be interpreted with caution. Most of the estimates of effects are probably in the right direction, and theory, common sense, and what is known about the phenomena studied from sources other than the data at hand (side information) can provide strong clues about when to suspect that the estimated direction is wrong. When theory and common sense suggest that the direction of the estimated effects is right, it should often be suspected that the real world effects are stronger than the estimated ones. Since correlational data not used to test explicitly specified causal models can suggest effects on marital quality, it is useful to discuss briefly some of the frequently found correlates.

One of the most frequent findings concerning marital quality is that it bears a nonmonotonic relationship to family life stage, being high in the preparental stage, lower in the parental stages, and relatively high in the "empty nest" or postparental stage. Although every few years someone publishes an article challenging the existence of this nonmonotonic relationship, its cross-sectional existence is hardly in doubt; it has been found by numerous studies conducted over more than twenty years, and the data from large national samples reported in Table 1 clearly show it. The reasons for and meaning of this cross-sectional relationship are not clear, however, and it is much less than certain that marriages that survive through all of the stages typically follow the down-up pattern suggested by the cross-sectional data (Vaillant and Vaillant 1993).

Most of the data on family life stage and marital quality confound any effects of presence-absence of children with those of duration of marriage. However, research has shown that each of these variables bears a relationship to measures of marital quality when the other is held constant (e.g., Glenn 1989; McLanahan and Adams 1989; White and Booth 1985; White, Booth, and Edwards 1986). That average marital quality drops during the first few years of marriage is hardly in doubt, especially in view of the fact that a high divorce rate during those years eliminates many of the worst marriages and prevents many marital failures from being reflected in the cross-sectional data. That reported marital quality is typically lower after the birth of the first child than earlier is also not in doubt, but in spite of numerous studies of the transition to parenthood (e.g., Belsky, Spanier, and Rovine 1983; Feldman and Nash 1984; Goldberg, Michaels, and Lamb 1985; McHale and Huston 1985), it still is not certain that the addition of a child to the family typically brings about any permanent reduction in marital quality. At least one study has found evidence that the cross-sectional association of the presence of a child or children with low marital quality results to a large extent, though not totally, from the fact that children tend to prevent or delay divorce and keep unhappily married couples together, at least temporarily (White, Booth, and Edwards 1986). The apparent upturn in marital quality in the postparental stage TABLE 1

Characteristics of married persons who reported very happy marriages
Characteristic Husbands (%) (n) Wives (%) (n)
* Hodge-Siegel-Rossi scores: Low = 9-39; Medium = 40-59; High = 60-82
† Persons in first marriages only
SOURCE: The 1973–1991 General Social Survey Cumulative File (Davis and Smith 1993). Data are weighted on number of persons age 18 or older in the household.
Hours per week wife worked outside home
0–9 68.4 (4,120) 63.7 (4,627)
10–29 65.9 (705) 63.1 (841)
30 or more 66.2 (2,580) 63.1 (2,836)
White 68.6 (6,691) 65.1 (7,482)
African American 54.5 (595) 48.3 (654)
Less than high school 66.4 (2,194) 56.7 (2,135)
High school 67.9 (3,503) 64.8 (4,761)
Junior college 61.4 (233) 62.4 (272)
Bachelor's degree 68.4 (938) 71.6 (844)
Graduate degree 69.1 (548) 67.6 (275)
Husband's cccupational prestige*
Low 64.9 (3,153) 60.2 (3,597)
Medium 69.3 (3,095) 64.9 (3,430)
High 70.8 (745) 74.6 (748)
Age at first marriage†
12–19 65.2 (741) 62.6 (2,690)
20–22 69.6 (2,005) 65.7 (2,284)
23–39 67.5 (3,274) 65.0 (1,912)
Years since first marriage†
0–2 75.0 (323) 78.8 (391)
3–5 63.8 (575) 66.6 (623)
6–8 63.2 (493) 64.4 (543)
9–11 63.7 (393) 61.6 (490)
12–14 68.2 (349) 64.1 (473)
15–19 63.5 (600) 58.5 (681)
20–24 63.6 (584) 59.6 (732)
25–29 69.3 (647) 62.0 (713)
30–39 68.6 (1,040) 67.1 (1,225)
40 or more 74.0 (1,074) 64.0 (1,046)
Family life stage
Preparental 70.8 (994) 74.4 (1,094)
Parental 64.5 (3,771) 59.6 (4,375)
Postparental 70.2 (2,659) 65.1 (2,823)
Spend a social evening with relatives
Never or infrequently 59.9 (484) 51.8 (463)
Several times a year 62.0 (876) 61.8 (880)
About once a month 64.5 (828) 62.8 (785)
Once or twice a week/several times a month 69.3 (2,139) 65.2 (2,534)
Almost daily 72.6 (246) 66.3 (334)
Religious preference
Protestant 68.1 (4,743) 63.9 (5,493)
Catholic 69.5 (1,864) 62.6 (2,172)
Jewish 68.4 (152) 68.6 (188)
None 53.5 (518) 56.8 (337)
Identification with a religion or denomination
Strong 75.1 (2,333) 69.1 (3,436)
Somewhat strong 65.3 (702) 62.5 (755)
Not very strong 64.0 (3,114) 57.5 (2,946)
Attendance of religious services
Never 57.8 (978) 58.9 (855)
Infrequently 62.8 (1,792) 58.7 (1,418)
Several times a year 67.5 (1,006) 61.1 (1,067)
One to three times a month 67.8 (1,117) 59.2 (1,343)
Weekly or almost weekly 73.0 (1,968) 68.0 (2,746)
Several times a week 77.6 (532) 71.2 (836)
is probably real, though small, but the existence of positive duration-of-marriage effects beyond the middle years is not at all certain, since the cross-sectional data confound any effects of marital duration with those of cohort membership and the divorces of unhappily married persons.

Some of the other data in Table 1 suggest influences stronger than those of family stage or duration of marriage. Since reported marital happiness varies directly with frequency of attendance of religious services and strength of identification with a religious group or denomination and is higher for persons who report a religious preference than for those who say they have no religion, religiosity is probably a rather strong positive influence on marital quality, especially since indicators of religiosity are also associated with low divorce rates (Glenn and Supancic 1984). However, more than religiosity may be involved in the relationship between frequency of attendance of religious services and marital quality. Such attendance is an indicator of social participation and social integration as well as religiosity, and there are many suggestions in literature that social integration is conducive to marital success. The data in Table 1 indicating that frequency of association with extended family members is positively related to marital happiness are also consistent with the social integration explanation of marital happiness.

The literature on African-American-white differences in the probability of marital success is extensive, yet there is no definitive evidence as to why reported marital quality is higher among whites (see Table 1) and divorce rates are lower. The measures of socioeconomic status usually included on social surveys do not account for much of the difference and in fact do not relate strongly to either reported marital quality or the probability of divorce (Glenn and Supancic 1984). Of course, this does not mean that past differences in economic conditions or differences in economic security are not important reasons for the racial difference in the probability of marital success, but their influence may be largely through such variables as the lower sex ratio among African Americans (Guttentag and Secord 1983).

The lack of even a moderately strong relationship between reported marital happiness and two of the independent variables included in Table 1 illustrates the danger of concluding that there is no effect on marital quality on the basis of cross-sectional data. The indicated differences in marital happiness by age at first marriage (shown only for persons in their first marriages) are quite small, which might lead to the conclusion that early marriages are almost as successful on average as marriages of more mature persons. However, the divorce rate does vary substantially by age at marriage—an indication that early marriages are unusually likely to have failed, either by ending in divorce or being less than satisfactory (Bumpass and Sweet 1972; Schoen 1975). It appears that persons who find themselves in less than satisfactory marriages within a few years after marriage are more likely to divorce if they were married early, and this results in a very small difference in marital quality between early and late marriers in intact first marriages.

The data in Table 1 show virtually no relationship between the number of hours the wife worked outside the home and the reported marital happiness of either the wife or the husband. There is no definitive evidence that the wife's working outside the home does affect marital quality, but these cross-sectional data do not prove that it does not. Suppose that at each level of marital quality, the probability of divorce varies directly with the number of hours the wife works outside the home—not an unlikely relationship. If so, poor marriages will be more quickly ended among couples in which the wife works more hours outside the home, thus raising the average marital quality in the remaining marriages in that category. If so, marital quality should vary positively with number of hours worked by the wife unless wives' working outside the home has a negative effect on marital success.

Dozens of other variables have been related to measures of marital happiness and satisfaction, but the ones discussed here are among those to which the greatest attention has been devoted. The deficiencies in the evidence of the causal importance of these variables is illustrative of the weakness of the body of "scientific" study concerning the bases of marital quality.

Additional topics

Marriage and Family EncyclopediaRelationshipsMarital Quality - Measurement Issues, Trends In Reported Marital Happiness, Bases Of Marital Quality, Consequences Of Marital Quality