Gender-Specific Trends in Educational Attainment and Self-Rated Health, 1972–2002

Abstract

Objectives. We tested whether self-rated health has improved over time (1972–2002) for women and men. We also considered the degree to which historical gains in educational attainment help to explain any observed trends.

Methods. Using 21 years of repeated cross-sectional data from the General Social Survey, we estimated a series of ordered logistic regression models predicting self-rated health.

Results. Our results show that women’s health status has steadily improved over the 30-year period under study, and these improvements are largely explained by gains in educational attainment. We also found that the health trend for men is nonlinear, suggesting significant fluctuations in health status over time.

Conclusions. Based on the linear health status trend and strong mediation pattern for women, and the nonlinear health status trend for men, women have benefited more than men, in terms of self-rated health, from increased educational attainment.

The National Center for Health Statistics recently issued a report showing that the overall health of the US population has improved over the past 50 years.1 This report confirms that the health of the nation has improved across a range of outcomes, including health behaviors (e.g., smoking), self-rated health, and all-cause and cause-specific mortality. If the health of the US population has improved over the years, what factors might mediate or explain this trend? Some of the trend may be attributable to health education, public health programs, health research, and health care; however, it is unlikely that these factors alone have driven the trend toward better health. We used 21 years of repeated cross-sectional data to test whether self-rated health status has indeed improved over time and whether historical gains in educational attainment help to explain this trend.

Aggregate gains in educational attainment over the past 30 years have been well documented.2^,3 In 1975, for example, 63% of adults aged 25 years and older had a high school diploma, and 14% had a bachelor’s degree or higher. By 2000, 84% of adults had completed high school, and 26% had a bachelor’s degree or higher. If educational attainment has improved over time, could this trend be responsible for changes in the health of the US population?

Our central proposition that trends in educational attainment could explain trends in health is largely based on research that shows that higher education predicts better health. In general, there are two explanations for the association between socioeconomic status (SES) and health. One explanation is that SES determines health status. The other explanation is that health status determines SES. Although there is some evidence to suggest that poor health might undermine SES, the majority of studies report that SES predicts health.4

Numerous studies show that the well educated tend to be healthier, both mentally and physically, than the less-well educated.5^–11 Education indirectly affects health through its association with economic resources and productive activities. Education increases household income and reduces financial hardship, which negatively impacts health. The well educated tend to work full time and to have jobs that are safe and intrinsically rewarding, which contributes to better health.5

Education also shapes resources that promote health and well-being. For example, education enhances a sense of personal control, which facilitates a healthy lifestyle, characterized by regular exercise, moderate alcohol use, and the avoidance of obesity and smoking. Formal schooling develops skills and abilities that can be used to solve a variety of problems, including those related to productivity and health.5 In general, education serves as a fundamental determinant of health because it influences access to important resources (e.g., money, power, and prestige) that can be used to avoid risks and to minimize the consequences of ill health.12

Although both women and men have experienced gains in educational attainment over the past 30 years, these gains may have benefited the health of women and men differently. Mirowsky and Ross’ theory of resource substitution proposes that individuals who have multiple material and/or psychosocial resources at their disposal may benefit less from the addition of any specific resource.5^,13 Conversely, resource substitution implies that the effect of having any specific resource is greater for individuals who have fewer resource alternatives. Because women have been historically disadvantaged in terms of socioeconomic position, status, and power,14 it is reasonable to expect that gains in educational attainment have had a greater influence on the trend toward better health for women than for men. Preliminary evidence already suggests that the effect of education on depression is stronger for women than for men.13

We examined three important research questions. First, has self-rated health actually improved over the past 30 years for women and men? Second, do historical gains in educational attainment over this period help to mediate or explain these improvements? Finally, does the mediating influence of education differ for women and men?

METHODS

Sample

To chart trends in self-rated health over time, we used General Social Survey (GSS) data from the National Opinion Research Center (NORC). The GSS is a useful data source because of its consistency in question format over the years (1972–2002). Each year of the GSS is based on a national probability sample of noninstitutionalized US residents, 18 years of age and older. Education and self-rated health have been included in the survey since 1972. However, self-rated health was omitted from the survey in 1978, 1983, and 1986; as a result, subsequent analyses do not include these years.

Measures

Self-rated health is the dependent variable. It is a widely used measure of general health status that is strongly correlated with more objective measures of health, including physician diagnoses and mortality.15^–17 According to Idler and Benyamini, self-rated health is a robust predictor of mortality, over and above numerous specific medical, behavioral, and psychosocial risk factors and other relevant covariates.17 Respondents were asked: “Would you say your own health, in general, is excellent, good, fair, or poor?” The response categories were coded poor (1) to excellent (4).

Survey year is the focal predictor variable. It represents our measure of time, and is the basis upon which health trends are assessed. Each year, each respondent was assigned a survey year. For example, all respondents from 1972 have been assigned a value equal to “1972.” The original year variable ranges from 1972 to 2002. We have recoded survey year to range from 1 (1972) to 21 (2002). We have also created quadratic and cubic versions of this variable to test for nonlinear trends in self-rated health.

Education is the focal mediator variable. It measures formal education in years. Numerous studies have shown that education is a significant predictor of self-rated health.5^,7^,8^,10^,11 We modeled education as a continuous variable because formal tests of credential-versus-quantity models show that each additional year of education improves health, whereas categorical degrees have no effect over and above the continuous specification.7

Numerous demographic characteristics have been identified as significant correlates of self-rated health. Poorer self-rated health is associated with being older, female, and black.7^,8^,10^,11^,18^–20 In accordance with previous research, our multivariate analyses included controls for age (in years, limited to adults aged 25 years and older), gender (1 = female), and race (Black, other races, and White—the reference category). Because we are primarily interested in explaining the trend toward better health and not the effect of education on health, other significant correlates of self-rated health are not included in subsequent analyses (e.g., work status, income, and health behaviors).

Statistical Procedures

Because self-rated health is measured as ordinal with 4 categories, we used ordered logistic regression to estimate predictor effects.21 Table 1 ▶ presents unstandardized logistic regression coefficients for the independent variables. These coefficients were estimated to describe how a 1-unit change in an independent variable affects the change in the cumulative odds of intensity in the dependent variable.

TABLE 1—

Unstandardized Coefficients for Ordered Logistic Regression of Subjective Health: General Social Survey, 1972–2002

	Women		Men
	Model 1	Model 2	Model 3	Model 4
Survey year	0.018***	4.52E–4	0.017**	0.004
Survey year2	. . .	. . .	−0.001*	−0.001*
Survey year3	. . .	. . .	−1.93E−4**	−1.85E−4**
Education	. . .	0.175***	. . .	0.144***
Age	−0.033***	−0.024***	−0.034***	−0.025***
Black	−0.731***	−0.546***	−0.510***	−0.305***
Other race	−0.502***	−0.382***	−0.358***	−0.268***
Model χ2	1662.224***	2735.612***	1075.410***	1847.630***
Δ Model χ2	. . .	1073.391***	. . .	772.224***
n	16 658	16 658	12 969	12 969

*P < .05; **P < .01; ***P < .001.

Our multivariate analytic strategy proceeded in 3 steps. We first examined unadjusted trends in education and self-rated health for women and men, respectively. We then regressed self-rated health on survey year to test whether self-rated health has improved over time. A significant positive slope, for example, would indicate that self-rated health has improved over time, and that the cumulative odds of better self-rated health are greater for respondents surveyed more recently. Finally, we controlled for education to test whether the relationship between survey year and self-rated health is mediated by education. A noticeable reduction in the magnitude of the year coefficient would suggest mediation. These steps were followed for women and men, respectively. To account for nonlinear health trends for men, we have included quadratic and cubic year terms in our multivariate analysis. The health trend for women was found to be linear.

RESULTS

Descriptive Statistics

The sample consists mainly of women (57%). In terms of race and ethnicity, the sample consists of Whites (84%), Blacks (13%), and those of other races (3%). On average, respondents were 48 years of age, with approximately 12 years of formal education. Finally, with respect to the dependent variable, the average respondent reported having “good” health. This pattern of self-rated health is consistent with the results of other surveys.11^,25

Unadjusted Trends

Figures 1 ▶ and 2 ▶ present unadjusted trends in mean educational achievement and mean self-rated health by survey year for women and men, respectively. The sex-specific trends indicate that education has improved steadily over the study period for both women and men. Self-rated health has also improved steadily for women. The pattern for men is less straightforward. We observed declines in the health of men over the 1970s and steady improvements over the 1980s. The health of men declined once again over the 1990s, and this pattern represents a departure from steady improvements in educational attainment.

FIGURE 1—

Mean educational level and self-rated health status, by year, for women.

Note. Education and self-rated health have been standardized to account for metric differences.

FIGURE 2—

Mean educational level and self-rated health status, by year, for men.

Note. Education and self-rated health have been standardized to account for metric differences.

Multivariate Trends

Table 1 ▶ presents multivariate estimates of predictor effects on the cumulative log odds of better self-rated health. Model 1 shows significant linear improvements in self-rated health over time for women. The coefficient for survey year (0.018) corresponds to a cumulative odds ratio of 1.018 (e^0.018). This implies that every unit increase in time is associated with about a 1.82% ([e^b − 1] × 100) increase in the cumulative odds of better self-rated health. For example, the cumulative odds of better self-rated health are approximately 38% (1.82× 21) greater for respondents surveyed in 2002 than they are for respondents surveyed in 1972.

With the introduction of education in Model 2, the association between survey year and self-rated health is reduced by approximately 97% ([b₁ − b₂]/b₁). The association between survey year and self-rated health is no longer statistically significant at conventional levels with the introduction of education, which suggests mediation. Education is also positively associated with self-rated health. In fact, every unit increase in years of formal education is associated with about a 19% increase in the cumulative odds of better self-rated health.

Model 3 demonstrates that the health trend for men is nonlinear, with significant linear, quadratic, and cubic components. The linear component represents the average trend, whereas the quadratic and cubic components represent a cyclical pattern around the average trend. The sign of the linear component indicates significant improvements in self-rated health, on average. The signs of the nonlinear components suggest significant declines in health. Once education is controlled in Model 4, the significant linear component is explained away. Thus, gains in education account for the average trend in men’s self-rated health over time. However, the quadratic and cubic components remain statistically significant. This suggests that education does little to explain the nonlinear components of the trend.

To further illustrate the complexities of the health trend for men, Figure 3 ▶ presents adjusted predicted probabilities of reporting excellent self-rated health by survey year for men only. We observed initial declines in the predicted probability of reporting “excellent” health from 1972 to 1982. We also found an increase in the probability of “excellent” health from 1982 to 1992, followed by another decline from 1992 to 2002.

FIGURE 3—

Predicted probability of excellent self-rated health, by year, for men.

In addition to testing whether self-rated health has improved over time, and whether historical gains in educational attainment help to explain this trend, we considered whether the mediating influence of education differed for men and women. It turns out that education explains about 97% of the trend toward better self-rated health for women. The health trend for men is nonlinear, so it is difficult to assess mediation. Taken together, these results suggest that education has had a greater influence on the health trend for women than for men.

Although we are not primarily concerned with the relationships between demographic characteristics and self-rated health, we would like to point out that our results are consistent with previous research. First, younger respondents reported better health than older respondents, and, second, White respondents reported better health than Black respondents and respondents of other races.

DISCUSSION

Although research shows that the health of the US population has improved over the past 50 years, explanations for this trend are not well developed. We tested whether self-rated health had improved over time and whether historical gains in educational attainment could help to explain this trend. Our results show that women’s health has steadily improved over the 30-year period under study, and that these improvements are largely explained by gains in educational attainment. We also found that the health trend for men is nonlinear, suggesting significant fluctuations in health status over time.

Our results for women are generally consistent with those of other studies. For example, the significant trend toward better health is consistent with the results of research conducted by the National Center for Health Statistics.1 Our finding that gains in educational attainment help to explain this trend is the main contribution of our study.

Although our results reveal significant linear improvements in self-rated health for women, the pattern for men is less straightforward. Our results show significant declines in men’s health during the 1970s and 1990s. These patterns are unexpected, given that men also exhibit steady gains in educational attainment during these periods. These findings offer support for the idea that, in terms of self-rated health, women may benefit more than men from increased educational attainment. If women and men were equally sensitive to gains in educational attainment, men’s health would not be expected to depart from these gains, as our results seem to suggest. Future research might consider the patterns for men in greater detail. For example, gains in educational attainment might only benefit men’s health when economic conditions are favorable. This explanation would imply that the benefits of education vary according to economic conditions, and women are less vulnerable than men to shifts in the economy.

The results of this study demonstrate the value of analyzing trends in health separately for women and men, and may have important implications for future trends in the sex mortality gap. Although women in the United States have been historically disadvantaged in terms of socioeconomic position, status, and power, they have generally exhibited substantially lower rates of mortality than their male counterparts. Differences in health behaviors between women and men, including smoking and risk-taking behavior, account for a significant proportion of this difference.23^–25 With more women than men currently earning bachelor’s degrees,3 women’s social and economic disadvantage relative to men is expected to decline even further than it has over the past 30 years. Thus, if women are able to retain their healthy lifestyle advantage, and if men continue to experience declining health despite their own substantial gains in educational attainment, then we should expect to see an increase in the sex mortality gap in coming decades.

While the GSS is a valuable data source for examining trends in self-rated health over time, it does have certain restrictions. The main limitation is our measurement of health. Self-rated health is the only measure of health that is regularly offered by the GSS. We acknowledge that self-rated health is based on self-reports and, as such, is subject to a number of unspecified judgments. Although self-rated health is correlated with other measures of health, and is a robust predictor of mortality, future research should consider whether trends in educational attainment help to explain trends in other, more objective, health measures.

Despite the limitations of the data, our results provide insight into the complex nature of the relationship between education and health. We found that self-rated health has steadily improved over time for women, and gains in educational attainment over the same period help to explain this trend. Taking into account the linear health trend and strong mediation pattern for women, and the nonlinear health trend for men, we conclude that women have benefited more, in terms of self-rated health, from increased educational attainment than their male counterparts.