2 Empirical motivation

2.2 Empirical model structure

We use difference-in-differences (DID) estimation to identify a potential causal effect of Title IX on women’s health outcomes at the ages of 42–52 years.Footnote ⁵ We examine this age range because it is the oldest age range that allows us to compare women at the same age in the pre-Title IX and post-Title IX periods. In recognition that Title IX was implemented gradually and federal regulations were not issued until 1975, with compliance not required until 1978, we separate out two periods to allow for a clearer comparator group for the before and after Title IX samples. Specifically, we consider five-year ranges for two groups: the pre-Title IX group born between 1951 and 1955 and the post-Title IX group born between 1966 and 1970.

Equation 1 provides the specific form of our DID equation for the groups of interest. White women provide the reference point, and we use either Black women or Hispanic women as the comparator group using the following equation:

(1) $$ {Health}_i={B}_0+{B}_1 Treate{d}_i+{B}_2 Whit{e}_i+{B}_3 Treate{d}_i\times Whit{e}_i+{B}_4{X}_i+{\gamma}_b+{\delta}_s+{\varepsilon}_i, $$

where $ {Health}_i $ is an indicator variable that takes the value of 1 if individual i had the health outcome being measured; $ Treate{d}_i $ is an indicator variable that takes the value of 1 if individual i was of high school age post-Title IX’s enactment (born from 1966 to 1970); $ {White}_i $ is an indicator variable equal to 1 for white women; $ Treate{d}_i\times {White}_i $ is the interaction term between the treated and white indicator variables for individual i; $ {X}_i $ is a vector of individual characteristics such as age, marital status and education level; $ {\gamma}_b $ accounts for fixed effects for individual i’s year of birth; $ {\delta}_s $ accounts for survey year fixed effects; and $ {\varepsilon}_i $ is the error term. Our main coefficient of interest is $ {B}_3, $ which captures the difference in effects that the implementation of Title IX had on the health outcome for the different groups of interest (white vs. Black females, and white vs. Hispanic females).

Consider the use of the equation to estimate the effect of Title IX on Black women relative to white women. The structure of the model is parallel for analysis of Hispanic women. In the analysis of the impact of Title IX on Black women, the sample consists of non-Hispanic white females and non-Hispanic Black females. The coefficient $ {B}_3 $ captures the difference in effects that the implementation of Title IX had on the health status of white females versus Black females. Suppose the health outcome variable pertains to whether the respondent’s self-assessed health is excellent. If $ {B}_3 $ has a negative value, then Title IX had a comparatively greater beneficial effect on Black women than white women. For health outcome measures that are adverse, such as being a smoker, a positive $ {B}_3 $ coefficient for whites indicates that Black women exhibited a greater drop in smoking rates in the post-Title IX period. Based on evidence regarding declining reported health status of women in more recent time periods, one would expect the coefficient $ {B}_1 $ to be negative for favorable health outcome variables such as being in excellent health, and positive for unfavorable health outcome variables such as being diabetic.

3 NHIS data

We use data from 1993 through 2018 from the NHIS on female respondents who are ages 42 through 52 at the time they were surveyed. The NHIS is a nationally representative survey designed to monitor the health of the US population. In addition to reporting demographic information, respondents report information on a broad range of health topics. We use weights throughout to account for the sample design, which oversamples some populations. We include in our analysis those who are non-Hispanic white, non-Hispanic Black, and Hispanic of any race. Other racial groups are too small for reliable analysis. We report results only for women in this paper because Title IX had almost no effect on sports participation of boys (Kaestner & Xu, Reference Kaestner and Xu2007).

We examine nine health outcomes, with descriptions for all health variables provided in the Appendix. We consider two self-assessed measures of overall health, whether excellent or whether excellent or very good. Smoking status is reported as whether the respondent had smoked at least 100 cigarettes in their lifetime. Obesity is defined as a BMI of 30 and more, and obese or overweight is defined as a BMI of 25 or more. Individuals report whether they had been told that by a doctor that they have diabetes, are borderline diabetic, had hypertension, or had a heart condition or disease. The NHIS asks different health questions of different populations in different years, and the number of observations available in the analyses varied by health outcome examined.

The regressions control for marital status and mutually exclusive categories of less than high school education, high school education, some college education, and a Bachelor’s degree or higher. We also include fixed effects controlling for each age, year born, and year of the survey.

4 Descriptive statistics

Table 1 summarizes the sample characteristics both pre-Title IX and post-Title IX for respondents who are white, Black, or Hispanic. Even with the age restrictions to have parallel samples in the pre-Title IX and post-Title IX periods, the NHIS produce substantial samples for all groups. The characteristics of the demographic groups are similar on dimensions such as age and year born, but differ in the expected ways otherwise. White women have the highest levels of education, followed by education levels for Blacks and Hispanics. White women are most likely to be married, and Black women are least likely to be married. The changes in the sample from the pre-Title IX period to the post-Title IX period are most substantial with respect to education levels. All three groups experienced a substantial increase in receiving a bachelor’s degree or higher. Subsequent regressions control for all the demographics listed in Table 1.

Table 1.

Descriptive statistics for demographic characteristics

Note: Authors’ calculations from National Health Interview Survey (NHIS) data survey years 1993–2018. Sample includes women aged 42–52 years when surveyed. Pre-Title IX observations were born in the years 1951–1955. Post-Title IX observations were born in the years 1966–1970. Standard deviations for continuous variables are given in parentheses. All other variables are percentages. All values are calculated using the NHIS person weight.

The DID regressions focus on the changes in various health outcomes between the pre-Title IX and post-Title IX period. Before turning to the regression results, Table 2 reports the mean values of the health outcome variables for the pre-Title IX period, the post-Title IX period, and the difference for the three sample groups. The mean values of these differences within these demographic groups exhibit changes that preview the regression results. For both measures of overall health status, excellent health and excellent or very good health, there is a significant decline in the self-assessed health for white women, with a percentage point decline in excellent health of 2.3. In contrast, both Black women and Hispanic report a statistically significant improvement in whether their health is excellent, with percentage point increases in excellent health of 3.4 for Black women and 2.6 for Hispanic women. The pattern is the same, but the magnitudes are muted when the self-assessed health status measure is expanded to also include the category of whether the health status is very good.

Table 2.

Pre-Title IX and Post-Title IX health outcomes

Note: Authors’ calculations from National Health Interview Survey (NHIS) data survey years 1993–2018. Sample includes women aged 42–52 years when surveyed. Pre-Title IX observations were born in the years 1951–1955. Post-Title IX observations were born in the years 1966–1970. All values are calculated using the NHIS person weight. Significant differences between pre-Title IX and post-Title IX health outcomes are indicated as follows:

⁺ significant at 10%;

* significant at 5%;

^** significant at 1%.

All three groups report substantial declines in the rate of smoking in the post-Title IX period, but the 15.4 percentage point decline in smoking rates for Black women and the 8.2 percentage point decline in smoking rates for Hispanic women substantially exceed the more modest 3.4 percentage point decline for white women. For both the overall self-assessed health measure and the smoking measure, the more favorable performance of Black women and Hispanic women relative to white women is suggestive of a likely favorable impact of Title IX. Being classified as obese or as either obese or overweight is on the rise for all three groups, with Hispanic women exhibiting the smallest increase. Diabetes rates are up for all three groups, and the patterns for hypertension are more mixed. The reporting of heart conditions is down for all three groups in Table 2, but there is no evident difference in the performance of Black women and Hispanic women compared to that of white women.

The comparator group for our analysis is white women, who indicate a decrease in the percentage reporting that they were in excellent health during the post-Title IX period. Using males in the NHIS data of the same age and born in the same years as the comparator reference group of white women similarly indicates a decline in the percentage reporting that they are in excellent health. In the pre-Title IX period, our calculations find that the percentage of males who reported that they are in excellent health was 31.9%, and the excellent health percentage declined to 27.8% post-Title IX, or a drop of around 4 percentage points (significant at the 0.01 level). Women also experienced a decline in excellent health ratings, but the decline is less for women. Overall, women in our sample had an excellent health percentage of 28.8% pre-Title IX and 27.3% post-Title IX, for a decline of less than 2 percentage points (significant at the 0.01 level). These results for our NHIS sample are consistent with the findings using the Health and Retirement Study in Hudomiet et al. (Reference Hudomiet, Hurd and Rohwedder2022), who also find a decrease over time in both men’s and women’s overall health status.

5 Regression results

The regression results in Table 3 report the key coefficients with respect to the temporal post-Title IX shift in the various health outcomes and the interaction of the post-Title IX variable with whether the respondent was white. The group being compared to white women is non-Hispanic Blacks in the first two columns of estimates and Hispanics in the final two columns.

Table 3.

Difference-in-differences regressions

Note: Sample is comprised of female respondents to the NHIS who were aged 42–52 years when surveyed and were born in the years 1951–1955 or 1966–1970. All regressions include indicators for married and education, as well as year born, survey year, and age fixed effects. All values are calculated using the NHIS person weight. Robust standard errors clustered by year born are reported in parentheses.

⁺ significant at 10%;

* significant at 5%;

^** significant at 1%.

Consider first the overall post-Title IX temporal shift variables, which are mostly consistent with the general deterioration in reported values for women’s health in the post-Title IX period for this age group. A notable exception to such an adverse trend in health status is the reduction in the rate of smoking. The negative coefficient for the variable pertaining to whether the respondent has smoked more than 100 cigarettes in their lifetime mirrors the general downward societal trend in smoking rates. In terms of the temporal trend, this change is the most favorable development in terms of promoting women’s health.

The DID interaction terms for Whites and the post-Title IX period are instructive in indicating that in most instances there have been comparatively greater gains for Black women and for Hispanic women than for white women. Although respondents overall are less likely to report excellent health or excellent or very good health, white women were even less likely to report these favorable health outcomes than are Black or Hispanic women. Relative to white women, Black women are 6 percentage points more likely to report excellent health and 3 percentage points to report excellent or very good health. When Hispanic women are compared to white women, the estimates reflect the same pattern but with slightly smaller magnitudes, with a 5 percentage points difference for excellent health and 2 percentage point difference for excellent or very good health.

The most consistently strong result for other health outcomes is that for smoking status. Compared to white women, the post-Title IX decline in smoking behavior is greater for women who are Black or who are Hispanic. The Black-white difference is 12 percentage points, and the Hispanic-white difference is 5 percentage points. Smoking is perhaps the most well documented risky behavior, boosting the probability of a diverse array of health outcomes, increasing mortality risks by one-sixth to one-third, and shortening life expectancy by 3.6–7.2 years (Viscusi, Reference Viscusi2002). Smoking behavior generally begins when people are in high school or of college age. Few people start smoking in middle age. The age groups most directly affected by Title IX are consequently well timed in terms of promoting athletic activities that will lead to the healthier long-term behavior of being a nonsmoker.

The post-Title IX interaction effects for the other health outcomes are more mixed in terms of whether there are greater gains for Black and Hispanic women relative to white women. Compared to white women, Black women show a greater reduction in being obese, with a 4 percentage point differential. But Hispanic women show a greater increase in being obese or being obese or overweight. The estimated Hispanic-white differential is 5 percentage points for obesity and 8 percentage points for being obese or overweight. There are no differential effects for being diabetic or for being diabetic and borderline diabetic. Compared to Black women, white women show a small relative reduction in hypertension, but an increase relative to Hispanic women.

The most consistent differential impacts of Title IX on Black and Hispanic women compared to white women are with respect to decreased smoking and increased measures of self-assessed health status, both in terms of whether health is excellent or whether health is excellent or very good. There is less consistent evidence of impacts of Title IX on specific health conditions of obesity, diabetes, or hypertension, which are ailments related to overall health status. The pronounced decrease in smoking rates should, however, have a consequential long-term impact on health given the increased rates of cancer, heart disease, and other ailments due to smoking.

The regression estimates in Table 3 control for a detailed set of personal characteristics but do not include controls for personal characteristics that one might view as comorbidities. The regressions in Table 4 add as explanatory variables whether the respondent has ever smoked and whether the respondent is obese, which are two health measures that may also reflect in part the influence of Title IX. The inclusion of these variables in the regressions in Table 4 results in an overall temporal shift in the post-Title IX period that no longer indicates an overall decline in health status once trends in smoking rates and obesity are taken into account. For the comparisons of Black women and Hispanic women with white women, the principal health status measure of interest is whether the respondent reports being in excellent health. Those coefficients show statistically significant improvements in excellent health relative to white women, with a 7 percentage point difference for Black women and a 6 percentage point difference for Hispanic women. The estimates for excellent or very good health follow the same pattern but with smaller coefficients. The results in Table 4 indicate no other statistically significant changes in health compared to white women during the post-Title IX period.

Table 4.

Difference-in-differences regressions controlling for obese and smoking

Note: Sample is comprised of female respondents to the NHIS who were aged 42–52 when surveyed and were born between 1951–1955 or 1966–1970. In addition to indicators for obese or ever smoked, all regressions include indicators for married and education, as well as year born, survey year, and age fixed effects. All values are calculated using the NHIS person weight. Robust standard errors clustered by year born are reported in parentheses.

⁺ significant at 10%;

* significant at 5%;

^** significant at 1%.

The health conditions frequently overlap and jointly contribute to how one assesses overall health status. A useful parsimonious approach to assessing the implications of the various results in these tables is to consider the summary measures of health status, whether the respondent rates her health status as excellent or rates it as either excellent or very good. While the differential effect of Title IX on minorities’ overall health status cannot be monetized readily, they do serve as a prominent measure of health status in the literature. The greater beneficial effects of Title IX for Black women and Hispanic women are consistent with policy efforts to ensure that policies benefit historically disadvantaged groups and address health disparities, so that these impacts are supportive of equity concerns. The greater marginal effects of Title IX on the health status of minorities are also attractive from an efficiency standpoint if the cost of providing access to athletic opportunities to Black and Hispanic women do not differ by race or ethnicity.

6 Cost implications of Title IX’s effect on smoking

The reduction in smoking rates resulting from Title IX has potential cost implications with respect to both health care expenditures and the valuation of health status. Smoking rates were reduced more for Black women and Hispanic women relative to white women, with Title IX reducing Black women’s smoking rate by an additional 12 percentage points and Hispanic women’s smoking rate by an additional 8 percentage points. Given the substantial health risks posed by smoking, these reductions have implications for both medical cost outlays and the valuation of health loss due to smoking. The following calculations quantify the benefits of smoking reduction. However, we are not able to fully take into account race-specific differences in the benefits of the greater reduction in smoking among Black and Hispanic women, as race-specific values of the financial and health costs of smoking are not available.

The proper assessment of the cost of smoking involves a comparison between the trajectory of lifetime costs for smokers and nonsmokers. The appropriate nonsmoker reference point is what has been termed the nonsmoking smoker (Manning et al., Reference Manning, Keeler, Newhouse, Sloss and Wasserman1991). In particular, the task is to assess the present value of the trajectory of costs for a person who has the same personal characteristics as smokers, such as education and income. The personal health care costs of smoking are unambiguous in that smoking increases a wide range of health risks and, on balance, has an expected adverse effect on health. The financial cost implications of smoking are less clear-cut, since cigarette smoking has financial cost implications that are both positive and negative. Because of their shorter lifetimes, the trajectory of medical costs ends sooner, as does their receipt of pension benefits and social security payments.

Consider first the expected health loss internality associated with smoking. The traditional valuation of these costs in the public health literature had been to assign an arbitrary value on each year of lost life, such as $100,000, and use that figure to monetize the health loss. The mortality cost assessment used by Viscusi and Hersch (Reference Viscusi and Hersch2008) monetized the health risks after estimating the value of a statistical life (VSL) and the VSL year (VSLY) for men and women smokers. They then applied these values to the change in the lifetime mortality rate trajectory due to smoking.Footnote ⁶ Based on a discount rate of 3%, the present value of the mortality cost per pack for a female smoker is $80.09 (USD 2000), or $136.61 (USD 2022). For a woman who smokes an average of 250 packs per year for 50 years, the lifetime mortality cost of smoking is $1.7 million.

The financial savings from reduction in smoking differ depending on the study, but are certainly much more modest given the competing financial implications of cigarettes. Estimates reported by Viscusi (Reference Viscusi2002) indicate that there are some financial externalities associated with smoking, the most important of which are for medical care ($0.58 per pack), along with cost increases for sick leave, group life insurance, forgone taxes on earnings, and fires. Smoking also generates cost savings in terms of reduced retirement and pension benefits of $1.26 per pack, as well as lower costs for nursing home care. Based on these estimates, the financial externalities of smoking generate cost savings of $0.32 per pack.

This result would seem to be inconsistent with the conclusion by Sloan et al. (Reference Sloan, Ostermann, Conover, Taylor and Picone2006) that the financial cost of smoking for a 24-year-old woman would be $86,000 (USD 2000). The largest component of their cost calculation is the estimated mortality cost of smoking of $52,385, which is not a financial cost and does not incorporate the use of the VSL. The financial cost categories of medical care costs, social security outlays, social security taxes, defined benefit private pensions, and life insurance lead to a net lifetime financial cost of $1,465, which is a negligible amount. This value is not too different from the estimate in Viscusi (Reference Viscusi2002) using a somewhat different set of financial categories. The main cost reduction associated with decreased smoking rates is in terms of reducing the considerable health cost of smoking.