Welfare reform has had far-reaching consequences for unmarried women and their children, including effects on their health insurance status. Those who would be receiving cash assistance absent welfare reform may have lower rates of health insurance if they failed to enroll separately for Medicaid (whose rules did not tighten over this time period), or if the new employment they entered did not provide health insurance. Administrative difficulties involved in accessing Medicaid separately from cash welfare may also have been a factor in the short run. Our research uses data from a large and nationally representative household survey that tracks the same individuals over time, the Survey of Income and Program Participation, to examine the effect of welfare reform (AFDC waivers and TANF implementation) on the health insurance status of unmarried mothers with High School completion or less (the population whose health insurance we expect would be affected by the welfare reform, or the “treatment group”) in the time period surrounding a particularly important life event, pregnancy. We look at the effects of these policies over the time period 1990-1999, as well as over the time period 1990-2003, to explore the short run vs. long run impact. Our “control group”, those who should not be affected by welfare reform itself but are expected to be affected by other national or state events that are happening contemporaneously, consists of married mothers with High School completion or less; the insurance experience of these women is used to control for the other forces that might otherwise lead us to attribute too little or too large an effect to welfare reform. Given the importance of access to health care at all points in the period surrounding and during pregnancy, we look at how welfare reform has affected insurance status before conception, during pregnancy, and after the birth of the child. We find that the Aid to Families with Dependent Children (AFDC) waivers of the 1990s as well as Temporary Assistance for Needy Families implementation have decreased access to Medicaid health insurance, increased access to employer health insurance, and led to a decrease in overall insurance, depending on the point in pregnancy considered and the time period of the study, with the largest effects found in coverage after the birth of a child. These findings have particular implications for the increasing emphasis on preconception and interconception care as a strategy to improve women's and infant's health.
Introduction and Background
The US health and welfare policies to improve maternal and infant health status have typically focused on the importance of adequate health care during pregnancy. Over the last 2 decades in particular, increasing access to health care during pregnancy has been the major focus of policies related to improving pregnancy outcomes (
Howell, 2001
), including expansions of the Medicaid program, which paid for 41% of all deliveries in the United States in 2002 (Family Foundation. (2007a and October, 2008
). More recently however, improving access to health insurance both before (preconceptional period) and after pregnancy (interconceptional period) has been gaining increased attention, because it is recognized that improving the health of a woman when she is not pregnant has significant potential to reduce adverse pregnancy outcomes and their associated short- and long-term costs (Kaiser Family Foundation. (2007a, October). Medicaid's role for women. Issue brief. An update on women's health policy. Available at: http://www.kff.org/womenshealth/upload/7213_03.pdf. Accessed February 14, 2008.
Johnson et al., 2006
).- Johnson K.
- Posner S.
- Bierman J.
- Cordero J.
- Atrash H.K.
- Parker C.S.
- et al.
Recommendations to improve preconception health and health care—United States. A report of the CDC/ATSDR Preconception Care Work Group and the Select Panel on Preconception Care.
Morbidity and Mortality Weekly Reports Recommendations and Reports. 2006; 55: 1-23
Having health insurance before pregnancy may be especially important for very low-income women who are at elevated risk for health problems regardless of pregnancy. In particular, uninsured women are less likely to access family planning and to receive sexually transmitted disease and HIV services, and are at increased risk of unintended pregnancies (
Kaiser Family Foundation and Alan Guttmacher Institute, 2005
). Becoming pregnant with unmanaged health conditions increases the likelihood of health problems during pregnancy, and therefore the likelihood of adverse pregnancy outcomes (Atrash et al., 2006
, Johnson et al., 2006
). Moreover, being uninsured before pregnancy may affect the timeliness of a woman's entrance into prenatal care (- Johnson K.
- Posner S.
- Bierman J.
- Cordero J.
- Atrash H.K.
- Parker C.S.
- et al.
Recommendations to improve preconception health and health care—United States. A report of the CDC/ATSDR Preconception Care Work Group and the Select Panel on Preconception Care.
Morbidity and Mortality Weekly Reports Recommendations and Reports. 2006; 55: 1-23
Egerter et al., 2002
, Braveman et al., 2003
); women without insurance before pregnancy may delay entry into prenatal care or forgo care altogether.Braveman, P., Marchi, K., Sarnoff, R., Egerter, S., & Rittenhouse, D. (2003). Promoting access to prenatal care: Lessons from the California experience. Washington DC: The Henry J. Kaiser Family Foundation. Available at: http://www.kff.org/womenshealth/loader.cfm?url=/commonspot/security/getfile.cfm&PageID=14290.
Welfare reform, brought on by the passage of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 and similar policies enacted by states during the early 1990s lead to a dramatic decrease in the welfare caseload (Temporary Assistance for Needy Families [TANF] replaced Aid to Families with Dependent Children [AFDC];
US Department of Health and Human Services, 2004
). Because of the close historic ties between cash assistance and Medicaid, a number of studies have analyzed how welfare reform has affected the health insurance status of women and children. Studies of women leaving welfare generally find that exiting welfare is associated with large decreases in Medicaid receipt (US Department of Health and Human Services. (2004). U.S. Welfare Caseloads Information.’' Available at: http://www.acf.dhhs.gov/news/stats/newstat2.shtml.
Acs and Loprest, 2001
, Guyer et al., 2002
). Two studies based on aggregate Medicaid rolls (Chavkin et al., 2000
, Garrett and Holohan, 2000
) also found declines in Medicaid after welfare reform.Several econometric studies of welfare reform and health insurance status using individual data also have been conducted.
Kaestner and Kaushal, 2003
found that the decline in caseloads overall lead to an increase in the uninsurance rate of 2%–9% among women and 6%–11% among children; however, the portion of this effect due to welfare reform was estimated to be smaller than the effect due to the improving economy during this time period. Bitler et al., 2005
examined how welfare reform (AFDC waivers and TANF implementation) affected health insurance status (as well as health care use and health status) among single women and found statistically insignificant effects on insurance status associated with AFDC waivers or TANF for African-American women and for their overall low-educated sample; however, there was a statistically significant negative association between health insurance status and TANF implementation among the sample of single Hispanic women (relative to married Hispanic women). Deleire et al., 2006
found that welfare reform increased the health insurance status of women with less than a high school degree, relative to higher educated women. Cawley et al., 2006
examined the effect of welfare reform on health insurance status for mothers and children using longitudinal monthly Survey of Income and Program Participation (SIPP) data from 1993 to 2000 that allowed them to track changes in insurance for the same individuals over time. Cawley et al., 2006
found that the negative impact of welfare reform on women and children in the SIPP was much larger than the estimates from prior studies.Of particular interest for health care policy for women of reproductive age is whether welfare reform had an impact on the health insurance of low-income pregnant women; this group theoretically should have been protected by the pregnancy-related Medicaid expansions of the late 1980s and early 1990s and by the fact that the very lowest-income women remained eligible for Medicaid even if not receiving cash assistance, if they met the eligibility standards in place for AFDC on July 16, 1996. Even though expansions in Medicaid eligibility rules were not reversed during the period after welfare reform, the changing welfare policy climate, administrative difficulties brought about by the policy changes (
Chavkin et al., 2000
, Greenstein and Guyer, 2001
, Hill and Lutzky, 2003
), the added marginal cost of enrolling in Medicaid without the automatic enrollment in cash assistance (Currie, 2004
), and the economic changes that resulted for those women making the shift from welfare to work may have led many low-income women to become eligible for or to seek Medicaid coverage only during pregnancy rather than during the entire period during which their children are <18 years old. To the extent that new jobs gained by women made then ineligible for Medicaid even during pregnancy (despite the existence of short-term transitional Medicaid that would still be available; Currie, J. (2004). The take-up of social benefits. NBER Working Paper #10488 Available at http://www.nber.org/papers/w10488.
Kaiser Commission on Medicaid and the Uninsured, 2002
), they may have remained uninsured during pregnancy if the new employment did not provide health benefits, or if there were preexisting conditions clauses for the coverage of a pregnancy that were not met. On the other hand, many of these same conditions may have led to an increase in employer health insurance for this group of women relative to the situation before reform.Kaiser Commission on Medicaid and the Uninsured. (2002). Transitional Medical Assistance (TMA): Medicaid issue update. Washington, DC: Kaiser Commission on Medicaid and the Uninsured, The Henry J. Kaiser Family Foundation. Available at: http://www.kff.org/medicaid/loader.cfm?url=/commonspot/security/getfile.cfm&PageID=14130.
Because of the particular nature of pregnancy-related insurance, there are several studies that have focused specifically on the effect of welfare reform on low-income women's insurance status in the prepregnancy and pregnancy/delivery periods.
Adams et al., 2003
, Adams et al., 2005
, Handler et al., 2006
, and Gavin et al., 2007
used PRAMS data from 8 (2005, 2006, 2007 studies) or 9 states (2003 study) to examine the prepregnancy insurance status and insurance transitions surrounding pregnancy in the periods before and after welfare reform. Adams et al., 2003
demonstrated that the percentage of low-income women uninsured before pregnancy increased in 5 (of the 9) states between 1996 and 1999, with significant increases in South Carolina and Washington. In their second paper, Adams et al., 2005
extended this work (using only 8 states) to a multivariate context and found that for “welfare-eligible” women (women eligible for Medicaid under 1996 welfare-related income levels), the odds of being Medicaid enrolled versus uninsured prepregnancy declined after welfare reform, with an absolute effect of a 7.9-percentage point decline in the probability of welfare-eligible women being insured before pregnancy. In a similar effort focused only on women reporting the receipt of cash assistance during their pregnancy, Handler et al., 2006
found that the prevalence of women in what they term the Medicaid Gap (having no prepregnancy coverage despite having Medicaid payment for delivery) increased from 16.1% in 1996 to 36.5% in 1998–2000; the adjusted odds ratio for falling into the Medicaid Gap for low-income women on cash assistance during their pregnancy in the 8 states was 4.5 (95% confidence interval, 2.1–9.6).Together these studies suggest that one of welfare reform's most notable effects associated with pregnancy was decreasing prepregnancy insurance coverage. Importantly, in a subsequent paper,
Rosenberg et al., 2007
found that in the period after welfare reform (1998–2000), the likelihood of delaying entry into prenatal care if a woman was in the Medicaid Gap was almost 3 times greater than for women not in the Medicaid Gap. Likewise, Gavin et al., 2007
found that welfare reform had a significant negative impact on the initiation of prenatal care in the first trimester among women eligible for Medicaid through their eligibility for cash assistance (estimated).Building on this prior work, the current study makes a number of contributions to the existing literature. First (to the best of the authors' knowledge), there has been no published tabulation of insurance coverage at multiple points surrounding a woman's pregnancy using nationally representative data since information published by the March of Dimes in 1999 (
Thorpe, 1999
). The SIPP is ideal for this task because it follows the same woman over a long period of time (over the length of a panel, each of which follows a different group of individuals) and, as such, is used for the analysis presented here. Second, there has been no nationally representative econometric study of the effect of welfare reform on the insurance status of women in the period surrounding and during pregnancy. Although the studies using PRAMS demonstrated a decrease in prepregnancy health insurance, these studies did not include a large sample of states, or provide information on insurance coverage during specific time points surrounding pregnancy. A final contribution is the longer time period of the study presented here; given the availability of more recent data, we are able to examine whether some of the impacts that were noticed soon after welfare reform still persist by extending our data horizon to 2003.Methods
To explore the effect of welfare reform on the health insurance status of women in the months before, during and after pregnancy, this analysis uses data from the 1990–1993, 1996, and 2001 panels of the SIPP, a survey conducted by the US Census Bureau that provides comprehensive information about individuals and households in the United States, including information about topics such as participation in government transfer programs, and health insurance coverage. Relative to surveys used in the prior studies described (Current Population Survey and Behavioral Risk Factor Surveillance System), the SIPP provides health insurance information that is specific to a point in time rather than referring to the previous year in general. The SIPP is also more detailed than the Behavioral Risk Factor Surveillance System in that the SIPP identifies whether a woman is a mother (and thus potentially eligible for welfare as a parent), and records the type of health insurance held (rather than simply recording whether someone is uninsured or not). In addition, SIPP data are available for the periods before and after welfare reform.
In the SIPP, each panel follows the same set of participants continuously for multiple years (typically 2.5 years; 4 years for the 1996 panel only). Interviews are conducted every 4 months, asking monthly information. This is essential for our analysis because this allows us to look at different points in time relative to the birth of the child. Using the longitudinal data within SIPP panels, we identify births to low-income women, and then examine changes in insurance status at different points around the pregnancy in the time periods before and after the implementation of welfare reform policy. Data from 1990 to 1999 are used to examine short-term effects; data from 1990 to 2003 are used to examine long-term effects. SIPP data produce national-level estimates for the US resident population and subgroups.
We use a standard “treatment-comparison group” approach; in this case, our treatment group is unmarried mothers aged 15–45 years with high school completion or less (surrogate for low-income women;
Dubay et al., 2001
). Our control group is married mothers aged 15–45 years with high school completion or less, a group theoretically not affected by welfare reform because of their expected higher incomes associated with marriage; in addition, recent studies suggest that welfare reform did not significantly affect rates of marriage (Bitler et al., 2004
, Kaestner and Kaushal, 2001
).The difference in differences approach allows us to control for factors that may have occurred in pregnancy-related health insurance/health insurance markets (be it at the national or state level) for all women during the time period under study, assuming that the experience of the control group reflects these effects as they would be experienced in the treatment group, had welfare reform not happened. We also control for other factors which may affect a woman's pre-pregnancy insurance status. This is similar to the assumptions and methods used in prior econometric studies of the effects of welfare reform on women and children in general; these prior papers provide evidence to support the assumptions made here (
Cawley et al., 2006
, Kaestner and Kaushal, 2003
). We also evaluate our results using alternate specifications and assumptions.We estimate difference-in-differences models of the following form:
where i indexes people, s states, and t time. Y stands for an indicator variable for insurance status (alternatively, any health insurance, Medicaid, own employer health insurance, and non-group health insurance), specifically, whether a person i has any health insurance coverage at time t (we look at 4 distinct points: 12 months before birth, 7 months before birth,1 month before birth, and 10 months post birth).X represents a set of individual characteristics (education, measured as high school dropout vs. high school completer [to differentiate between the 2 education groups included in a sample of those with high school completion or less], age and age squared [allowing a flexible form for the association between aging and health insurance], race and ethnicity [White, African-American, Hispanic, Asian, and other], and whether the pregnancy is associated with the woman's first child). Because marital status is what separates the treatment from the control group, it is not included in the X vector. In our dataset, a row of data represents 1 woman at a particular point in her pregnancy (e.g., in the regression for 7 months before delivery, 1 row represents a woman's data at the point she is 7 months before delivery).
where i indexes people, s states, and t time. Y stands for an indicator variable for insurance status (alternatively, any health insurance, Medicaid, own employer health insurance, and non-group health insurance), specifically, whether a person i has any health insurance coverage at time t (we look at 4 distinct points: 12 months before birth, 7 months before birth,1 month before birth, and 10 months post birth).X represents a set of individual characteristics (education, measured as high school dropout vs. high school completer [to differentiate between the 2 education groups included in a sample of those with high school completion or less], age and age squared [allowing a flexible form for the association between aging and health insurance], race and ethnicity [White, African-American, Hispanic, Asian, and other], and whether the pregnancy is associated with the woman's first child). Because marital status is what separates the treatment from the control group, it is not included in the X vector. In our dataset, a row of data represents 1 woman at a particular point in her pregnancy (e.g., in the regression for 7 months before delivery, 1 row represents a woman's data at the point she is 7 months before delivery).
2
The next item, vector Z, represents a set of state-level characteristics that vary over time, including a measure of Medicaid generosity for pregnancy-related insurance, the real minimum wage in the state, Earned Income Tax Credit generosity, the real benefit standard for cash assistance, the real per capita state income, and the unemployment rate and its 12-month lag (further descriptions and sources are available upon request). Vector P represents a set of variables reflecting welfare policy, specifically, an indicator for whether state s had an AFDC waiver at time t, and an indicator for whether state s had implemented TANF at time t. TREAT is an indicator that equals 1 if the respondent is a member of the treatment group, and equals 0 if the respondent is a member of the control group. The coefficient λ on the interaction term P∗TREAT is our difference-in-differences measure of the effect of welfare reform on prepregnancy, pregnancy, and postpartum insurance status. We include fixed effects for state, year, and panel.We calculate the timing before the birth of a child as follows. First, individuals are asked about the dates of birth of children in the household. That means we miss any pregnancies that did not produce live births or where the young child (one under about 2–3 years of age) does not live with the mother. On the other hand, some of the young children we might identify as birth children could be adopted (step or foster children are identified separately in the SIPP) because in pre-1996 panels, adopted children were not separately identified. However, our checks with the later panels suggest there are very few cases of miscoding that may occur in this way. We report results by month of pregnancy rather than by trimester because it involves no assumptions about exact length of gestation. For example, depending on the length of gestation being 10, 9, or 8 months, 7 months before birth could theoretically be the first or the second trimester. Nationally, almost 88% of babies are born at full term (
Martin et al., 2007
), although this rate is likely lower among women with less education given the inverse association between lack of education and adverse pregnancy outcomes (Kramer et al., 2000
). Further details of the data creation are available from the authors. 3
Other details of the estimation appear in notes under Table 3.The SIPP does not identify certain small states separately; thus, we are unable to use SIPP data for the following states and time periods: the 1990–1993 panels for Maine, Vermont, Iowa, North Dakota, South Dakota, Alaska, Idaho, Montana, and Wyoming; and the 1996 and 2001 panels for Maine, Vermont, North Dakota, South Dakota, and Wyoming. These states represent a very small fraction of the US population (Maine, Vermont, North Dakota, South Dakota, and Wyoming together account for 1.3% of the US nonelderly population in 2000 based on author calculations using US Census Bureau Population data).
Because the SIPP is not a survey of pregnant women, it is important to determine that the number of pregnancies found in the SIPP survey is representative of the number of pregnancies in the United States nationally. When we use the responses to a SIPP interview from early 2000 (this sample should include information about all individuals who were born in 1999) and produce a weighted frequency of those children with a birth year of 1999, there are 3,842,560.5 births, which is very close to the 3,959,417 infants born in 1999 (
Centers for Disease Control and Prevention, 2006
), indicating that the SIPP provides a representative sample of the annual number of US births. This provides confidence that the births in the other years are also representative samples of all births in that year as 2000 was picked at random.Centers for Disease Control and Prevention. (2006). Live births, birth rates, and fertility rates, by race: United States, 1909–99. Available at: http://www.cdc.gov/nchs/data/statab/t991x01.pdf. Accessed October 2006.
Results
We first show data on the insurance coverage of US women during different points relative to the birth of a child, for pregnancies occurring in 1990–2003 (Table 1). Of all women age 15–45 years who gave birth at some point during the SIPP 1990–2001 panels and were observed ≥12 months before that, 81.9% were insured 12 months before birth. Women's insurance coverage increased as their pregnancy progressed, and reached a peak at 3 to 1 months before birth. Their health insurance coverage even 10 months after birth was higher than it was a year before birth. The trend is toward increased rather than decreased Medicaid coverage, increased dependent employer coverage, and decreased own employer insurance over the course of the pregnancy–postpartum period, which likely primarily reflects reductions in labor force involvement that have health insurance consequences. The trends are in the same direction but of different magnitudes for women giving birth for the first time (data not shown). For these women, the initial level of “own employer health insurance” is higher, and the initial level of Medicaid coverage is lower. The final level of Medicaid coverage is also lower, presumably reflecting the differences in socioeconomic status between all first time mothers versus all higher birth order mothers. The trends are also similar by panel, although again exact magnitudes vary (data not shown).
Table 1Insurance Status: Points Relative to Childbirth, SIPP Panels, 1990–2001, (Data covering pregnancies 1990–2003), US Women 15 to 45 years of Age
| Any Health Insurance | Medicaid | Own Employer Coverage | Dependent Employer Coverage | Nongroup | n | |
|---|---|---|---|---|---|---|
| 12 months before birth | 0.819 | 0.117 | 0.320 | 0.332 | 0.040 | 7,422 |
| 9 months before birth | 0.828 | 0.130 | 0.314 | 0.335 | 0.036 | 8,605 |
| 7 months before birth | 0.856 | 0.172 | 0.309 | 0.334 | 0.035 | 9,421 |
| 6 months before birth | 0.874 | 0.198 | 0.302 | 0.336 | 0.036 | 9,902 |
| 3 months before birth | 0.904 | 0.244 | 0.284 | 0.339 | 0.037 | 11,475 |
| 1 month before birth | 0.901 | 0.253 | 0.274 | 0.341 | 0.037 | 12,596 |
| 1 month post birth | 0.898 | 0.267 | 0.254 | 0.345 | 0.038 | 13,428 |
| 3 months post birth | 0.867 | 0.236 | 0.236 | 0.356 | 0.042 | 13,280 |
| 10 months post birth | 0.838 | 0.206 | 0.224 | 0.364 | 0.039 | 10,475 |
Note: Uninsured not shown; complement of any health insurance. Sample weights are used. Source: Author calculations from SIPP panels 1990–2001.
Table 2 presents the same insurance information by treatment and control group separately. Treatment group women have insurance levels that are about 7 percentage points lower than control women 12 months before the birth of a child, but this differential shrinks to about 3 percentage points 1 month before birth. However, treatment group women are about 40 percentage points more likely to have Medicaid coverage at this point. In the regressions that follow, we examine employer health insurance through one's own employer as well as any health insurance coverage as 2 separate outcomes.
Table 2Insurance Status: Points relative to Childbirth, SIPP Panels, 1990–2001 (Data covering pregnancies 1990–2003), US Women 15 to 45 years of Age
| Treatment and Control Groups | ||||||
|---|---|---|---|---|---|---|
| Any Health Insurance | Medicaid | Own Employer Coverage | Dependent Employer Coverage | Nongroup | n | |
| Insurance status: points relative to childbirth | ||||||
| Control women aged 15–45 (married, high school or less) | ||||||
| All panels | ||||||
| 12 months before birth | 0.750 | 0.091 | 0.249 | 0.360 | 0.027 | 1,820 |
| 1 month before birth | 0.865 | 0.254 | 0.216 | 0.353 | 0.031 | 3,609 |
| 10 months post birth | 0.753 | 0.153 | 0.170 | 0.372 | 0.035 | 3,193 |
| Insurance status: points relative to childbirth | ||||||
| Treatment women aged 15–45 (unmarried, high school or less) | ||||||
| All panels | ||||||
| 12 months before birth | 0.680 | 0.337 | 0.162 | 0.179 | 0.026 | 1,704 |
| 1 month before birth | 0.837 | 0.634 | 0.111 | 0.136 | 0.016 | 2,611 |
| 10 months post birth | 0.784 | 0.609 | 0.106 | 0.088 | 0.016 | 2,042 |
Notes: Sample weights are used.
For the difference in differences estimates for the effect of welfare reform on the insurance status of women in the period surrounding pregnancy, we first generate results for the 1990–1999 time period. The effects are shown at four points relative to the pregnancy, for 2 different types of health insurance (Medicaid, own employer) as well as any health insurance. We show coefficients and standard errors for “Treatwaiver” and “Treattanf” corresponding to P∗Treat in our estimation equation. These show the causal impact of AFDC waivers and TANF on insurance status of the treatment group. For example, the coefficients and standard errors for the variable “Treatwaiver” in Table 3 show that there is no statistically significant relationship between AFDC waivers and insurance status at any of the points considered. The coefficients are usually in the direction one might expect (negative effects of welfare reform on Medicaid and positive effects on own employer health insurance), but standard errors are always fairly large leading to no precise statistical relationship.
Table 3Difference in Differences Results for the Relationship Between Welfare Reform and Insurance Status of Women 15–45 in the Period Surrounding Pregnancy 1990–1999
| Data 1990–1999 | n | ||
|---|---|---|---|
| 12 months before birth | 2,958 | ||
| Any health insurance | Medicaid | Own employer | |
| treatwaiver | 0.03 | 0.014 | 0.013 |
| [0.042] | [0.031] | [0.028] | |
| treattanf | −0.074∗ | −0.065∗∗ | 0.042 |
| [0.040] | [0.029] | [0.048] | |
| 7 months before birth | 3805 | ||
| Any health insurance | Medicaid | Own employer | |
| treatwaiver | 0.002 | −0.055 | 0.033 |
| [0.034] | [0.039] | [0.027] | |
| treattanf | −0.008 | −0.072∗∗ | 0.077∗ |
| [0.037] | [0.028] | [0.038] | |
| 1 month before birth | 5188 | ||
| Any health insurance | Medicaid | Own employer | |
| treatwaiver | −0.006 | −0.002 | 0 |
| [0.019] | [0.045] | [0.026] | |
| treattanf | −0.036 | −0.115∗∗∗ | 0.086∗∗ |
| [0.033] | [0.035] | [0.034] | |
| 10 months post birth | 4367 | ||
| Any health insurance | Medicaid | Own employer | |
| treatwaiver | 0.015 | −0.054 | 0.059 |
| [0.049] | [0.032] | [0.036] | |
| treattanf | −0.098∗∗ | −0.151∗∗∗ | 0.080∗∗∗ |
| [0.038] | [0.032] | [0.026] | |
Robust standard errors in brackets
∗Significant at 10%; ∗∗significant at 5%; ∗∗∗significant at 1%.
Note: Sample weights are used; standard errors are clustered at the state level. A woman could be in the dataset more than once. For example, if she gives birth twice during the 1996 panel (4 years), each birth is counted as a separate event; the impact of this on standard errors is already accounted for by clustering at the state level.
In contrast with the results for waiver implementation, TANF implementation is associated with statistically significant reductions in Medicaid coverage and increases in employer provided health insurance for multiple points under study. Medicaid reductions associated with TANF implementation are statistically significant at every point we consider. There is an increase in employer health insurance in 3 of the periods surrounding pregnancy (7 months before, 1 month before, 10 months after birth; Table 3). For example, 12 months before birth, TANF is associated with a 6.5-percentage-point decrease in Medicaid that is statistically significant, a small nonsignificant positive coefficient on employer health insurance, and a marginally significant negative coefficient on overall coverage. One month before birth, there is a 11.5-percentage-point reduction in Medicaid coverage and a 8.6-percentage-point increase in employer health insurance, both of which are significant, which leads to a statistically insignificant but negative effect on health insurance overall. Ten months after birth there is a 15-percentage-point decrease in Medicaid coverage, an 8-percentage-point increase in employer health insurance, and overall a decrease in any health insurance that is statistically significant.
Next we consider the long vs. short run effects of policy in Table 4. These results use the same model specifications, but the time period is now expanded to include data through 2003. When examining this longer time period, the results for AFDC waivers are in the same direction and of essentially the same magnitude as the results for the shorter time period; again, there are no significant effects. However, the results for TANF show effects that for the most part are not as robust as the short-term results for Medicaid coverage, but show stronger effects for own employer health insurance. For example, at 12 months before birth, there is now a significant increase in own employer health insurance of 7.5 percentage points (relative to a statistically insignificant effect in the short run). At 7 months before birth there is a 9.7-percentage-point increase in own employer health insurance (compared with 7.7 percentage points in the short run). At 1 month before birth, Medicaid drops 7.4 percentage points (relative to 11.5 percentage points in the short run); the effect on own employer health insurance is also larger at this time point than in the short-run results. The exception to this pattern of results is that at 10 months after birth, there are larger effects for Medicaid and any health insurance, and a slightly smaller effect (7.8 vs. 8 percentage points) for own employer health insurance in the longer term compared with the short-term results.
Table 4Difference in Differences Results for the Relationship Between Welfare Reform and Insurance Status of Women 15–45 in the Period Surrounding Pregnancy 1990–2003
| Data 1990–2003 | n | ||
|---|---|---|---|
| 12 months before birth | 3,524 | ||
| Any health insurance | Medicaid | Own employer | |
| Treatwaiver | 0.035 | 0.017 | 0.014 |
| [0.043] | [0.032] | [0.028] | |
| Treattanf | −0.012 | −0.021 | 0.075∗∗ |
| [0.041] | [0.024] | [0.034] | |
| 7 months before birth | 4552 | ||
| Any health insurance | Medicaid | Own employer | |
| treatwaiver | 0.005 | −0.052 | 0.032 |
| [0.033] | [0.039] | [0.027] | |
| treattanf | 0.017 | −0.038 | 0.097∗∗∗ |
| [0.031] | [0.023] | [0.029] | |
| 1 month before birth | 6220 | ||
| Any health insurance | Medicaid | Own employer | |
| treatwaiver | −0.008 | −0.003 | 0.002 |
| [0.019] | [0.045] | [0.026] | |
| treattanf | −0.008 | −0.074∗∗ | 0.092∗∗∗ |
| [0.026] | [0.032] | [0.026] | |
| 10 months post birth | 5235 | ||
| Any health insurance | Medicaid | Own employer | |
| treatwaiver | 0.016 | −0.05 | 0.054 |
| [0.048] | [0.033] | [0.034] | |
| treattanf | −0.102∗∗∗ | −0.171∗∗∗ | 0.078∗∗∗ |
| [0.031] | [0.032] | [0.021] | |
Robust standard errors in brackets.
∗Significant at 10%; ∗∗significant at 5%; ∗∗∗significant at 1%.
Note: Sample weights are used; standard errors are clustered at the state level. A woman could be in the dataset more than once. For example, if she gives birth twice during the 1996 panel (4 years), each birth is counted as a separate event; the impact of this on standard errors is already accounted for by clustering at the state level.
Specification checks
We estimated alternative (unreported) models to investigate the sensitivity of our results. First, we reestimated the models using data on first births; these mothers may be more socioeconomically advantaged than the cohort experiencing later births and thus not affected as much by welfare reform, or might be more affected if they have less of a previous connection to the labor force. Overall, the results for first births are a little larger than for non-first births for some specifications, possibly supporting the second alternative. Second, we estimated models that did not use a control group. This is to determine the extent to which the results we observe are due to comparisons of changes in insurance experienced by both the control versus treatment groups in the time periods surrounding the welfare reform incidents. In those specifications, the direction of the effects is consistent with the results that included the control groups, although there are substantial differences in the sizes of the coefficients. Depending on whether one believes there is a need for a control group (and whether the control group is adequate), this could be interpreted as indicating smaller effects than those resulting from models that do use a control group. We also reran our models including only women under 100% of the federal poverty level as 1 way to crudely approximate a welfare recipient or a welfare-eligible group using only the treatment group. Women in this group represent those most likely to have remained on welfare and/or to be eligible for Medicaid when nonpregnant owing to very low income. However, there are no statistically significant results for this group, even where we had earlier seen the largest changes in Table 3, Table 4, the period 10 months after birth. In our last 2 robustness checks, we first limited the sample to individuals who report always receiving AFDC or cash welfare; as with the group below 100% of the federal poverty level, we see few significant changes in their insurance status. However, when we limit the sample to those who were on welfare at the start of the SIPP but end their survey data off welfare (welfare leavers), we see a different pattern. Although few of the results are statistically insignificant owing to small sample sizes, the magnitudes of the coefficients are very large and in the expected direction (decreases in Medicaid and increases in own employer health insurance). In other words, those who are most at risk of the effects of welfare reform (those who were on cash assistance and subsequently moved off welfare) seem to be the low-income women most affected, although we note that this specification involves a small sample of welfare leavers.
Discussion
This study uses a nationally representative data set covering 1990–2003 to investigate the effect of welfare reform on the health insurance status of pregnant women. We compared the impact of welfare policy on a group of women expected to be affected (women who are single mothers with high school or less) compared with an otherwise similar group we do not expect to be affected by the policy (married mothers with similar education), and found that there is in general a negative effect of welfare reform on Medicaid coverage and an increase in own employer health insurance in the period surrounding pregnancy. This supports prior evidence that welfare reform increased labor force attachment and also reduced access to Medicaid. Our results depended somewhat on the time period used and whether the analysis focused on the implementation of TANF or the AFDC waivers. When examining the short run (1990–1999), we found larger negative (and significant) effects associated with TANF for Medicaid coverage before pregnancy relative to results using longer run data (1990–2003).
Our findings provide more current and more detailed information on the insurance coverage of women in the period surrounding pregnancy than has been available in the past (work conducted by
Thorpe, 1999
for the March of Dimes). From the descriptive statistics provided here, it is clear that pregnancy remains a pivotal point for women in terms of obtaining health insurance coverage. However, for women who are eligible for Medicaid coverage, increased coverage during pregnancy is followed by a decrease postpregnancy; this postpregnancy decrease is similar to the experience of women with their own employer coverage (who actually seem to lose coverage throughout the pregnancy period) and distinct from the experience of women with dependent employer coverage who seem to retain the coverage that they gained during pregnancy (Table 1). For the Medicaid population, these findings reflect the fact that although Medicaid now pays for >40% of all US births (and associated immediate postpartum care; Family Foundation. (2007a and October, 2008
), this coverage often is not available to women in the period before or between pregnancies unless they are extremely low income. Importantly, in the last 15 years, 26 states have adopted family planning waivers that allow Medicaid coverage for family planning and associated services for women not Medicaid eligible for other reasons (Kaiser Family Foundation. (2007a, October). Medicaid's role for women. Issue brief. An update on women's health policy. Available at: http://www.kff.org/womenshealth/upload/7213_03.pdf. Accessed February 14, 2008.
Family Foundation. (20007b and October, 2008
); however, significantly fewer states extend Medicaid funds to cover parents of children eligible for coverage through their Medicaid and State Children's Health Insurance Programs (Kaiser Commission on Medicaid and the Uninsured, 2008). With increasing financial pressures faced by the states associated with economic changes that began earlier in the decade, the shrinkage of the federal budget, and recent legislative and administrative changes in the rules governing Medicaid and State Children's Health Insurance Programs, the ability to sustain expanded coverage for nonpregnant adults is of concern (Kaiser Family Foundation. (20007b, October). Medicaid's role in family planning. An update on women's health policy. Available at: http://www.kff.org/womenshealth/upload/7064_03.pdf. Accessed February 21, 2008.
Kaiser Commission on Medicaid and the Uninsured. (2008), 2008
).Kaiser Commission on Medicaid and the Uninsured. (2008). Health coverage for children and families in Medicaid and SCHIP: State efforts face new hurdles: A 50-state update on eligibility rules, enrollment and renewal procedures, and cost-sharing practices in Medicaid and SCHIP in 2008. Washington, DC: Kaiser Family Foundation. Available at: http://www.kff.org/medicaid/upload/7740.pdf. Accessed February 21, 2008.
Similar to several prior studies focused on the effects of welfare reform on pregnant women in particular, it seems that welfare reform's most important impact on the health coverage of low-income pregnant women is not during the pregnancy itself. In the current study, the strongest effect of welfare reform for women with Medicaid coverage was on coverage in the months after a pregnancy (15.1% decline in the shorter term analysis, 17.1% decline in the longer term analysis). For policy makers who are increasingly focusing on the health care of nonpregnant women as a strategy to improve the health outcomes of mothers and infants, it is essential to not only consider the period before a first birth (strictest definition of “preconception” care) but the period after a pregnancy, which for many women may become an interconception period. Ensuring women's health in both the preconception and interconception periods requires a life-course approach to the financing of women's health care; the emphasis on coverage during pregnancy, which has been the major public policy strategy for improving pregnancy outcomes over the last 2 decades, although necessary has clearly not been sufficient. Studies of the effect of the Medicaid expansions of the late 1980s and early 1990s on improving pregnancy outcomes have shown equivocal results (
Baldwin et al., 1998
, Braveman et al., 1993
, Dubay et al., 2001
, Haas et al., 1993
, Howell, 2001
, Long and Marquis, 1998
, Piper et al., 1990
, Ray et al., 1997
), ushering in the current emphasis on interventions during the preconception and interconception periods. However, understanding the Medicaid expansion policy “failure” requires acknowledging that the Medicaid expansion legislation did not ensure prepregnancy coverage for expansion women or coverage early enough in pregnancy to allow women to reap many of the potential benefits of accessing prenatal care.The analysis presented herein also examined the experiences of women living in poverty at the time of childbirth—women ever on welfare as well as women who left welfare over the course of a SIPP panel—as a way to determine which group of very low-income women has been most affected by welfare reform. Prior research by
Adams et al., 2005
and Gavin et al., 2007
suggests that the effects of welfare reform have been greatest for these welfare-eligible rather than expansion-eligible women. Likewise, Handler et al., 2006
and Rosenberg et al., 2007
suggest that the experience of women who before welfare reform would have been on Medicaid before pregnancy because they were cash assistance recipients has become similar to that of the Medicaid expansion women, those who are only eligible for Medicaid when they are pregnant. In the current analysis, the effect of AFDC waivers/TANF on insurance status, although not statistically significant owing to the small sample size, seemed to be most important for those very low-income women who began their SIPP panel experience as cash assistance recipients but left over time.Our results on the effect of welfare reform on insurance coverage during, before, and after pregnancy are obtained from a robust research design in which we compare the women at multiple points surrounding a pregnancy using a nationally representative sample. To our knowledge, this is the first published data since 1999 to portray women's insurance coverage at multiple time points surrounding pregnancy. Importantly, because we use data from multiple SIPP panels beginning with the 1990–1993 panel and following through to the 2001 panel, we are able to examine the experiences of women in the pre- and post-welfare reform time periods.
There are several limitations to our study. First, sample size limitations prevented us from analyzing the results by race and ethnicity, which would have provided a national overview of differential rates of insurance in the period surrounding pregnancy for minority women who are at higher rates of adverse pregnancy outcomes (particularly, African Americans and Puerto Ricans) and whether their coverage has been affected by welfare reform. Second, there may be variables that affect insurance coverage that we do not include in our model, such as health status during pregnancy. Unless these omitted variables are correlated with the passage of welfare reform and are different between treatment and control groups, there is no reason to expect that our results are biased. The most significant limitation is the assumptions used in any study that attempts to draw out the causal effect of a policy change such as welfare reform using nonexperimental data. One is that we assume the policy affected women who are single mothers with a high school education or less, whereas married mothers with similar education were not affected. Moreover, we assume that the married women with similar education would be affected by any other state trends in insurance coverage that would have affected the unmarried mothers, absent welfare reform. This is an assumption made in most other econometric papers on welfare reform, but it is nevertheless important to keep in mind, because our robustness checks indicate the magnitude of our results are sensitive to this assumption.
Conclusion
This study demonstrates that pregnant women were not insulated from the negative effects on health insurance of welfare reform in either the short or long run. Although the impact on prepregnancy insurance seems to be less severe in the long run, of interest for further study and discussion are the effects of declines in insurance coverage 10 months after birth, particularly Medicaid coverage, as the interconceptional period gains increased attention as a point of intervention. In particular, as maternal and child health professionals put forth arguments for new health care financing policies such as reimbursement of medical/nursing providers for time spent offering preconception and interconception care, mandating Medicaid coverage for all women at a certain income or below after birth for a period of years, and/or increased private sector coverage of an annual obstetric/gynecologic visit (
Atrash et al., 2006
, Klerman(2005June 21),
), these calls for change must be considered within the context of previous policy changes that may have reduced access to insurance coverage for low-income women around the time of pregnancy.References
- Final synthesis report of findings from ASPE's ‘Leaver’ Grants.Urban Institute, Washington, DC2001
- Transitions in insurance coverage from before pregnancy through delivery in nine states, 1996–1999.Health Affairs. 2003; 22: 219-229
- Welfare reform, insurance coverage pre-pregnancy, and timely enrollment: An eight-state study.Inquiry. 2005; 42: 129-144
- Preconception care for improving perinatal outcomes: the time to act.Maternal and Child Health Journal. 2006; 10: S3-S11
- The effect of expanding Medicaid prenatal services on birth outcomes.American Journal of Public Health. 1998; 88: 1623-1629
- The impact of welfare reform on health insurance and health outcomes.Journal of Human Resources. 2005; 40: 309-334
- The impact of welfare reform on marriage and divorce.Demography. 2004; 41: 213-236
- Access to prenatal care following major Medicaid eligibility expansions.Journal of the American Medical Association. 1993; 269: 1285-1289
Braveman, P., Marchi, K., Sarnoff, R., Egerter, S., & Rittenhouse, D. (2003). Promoting access to prenatal care: Lessons from the California experience. Washington DC: The Henry J. Kaiser Family Foundation. Available at: http://www.kff.org/womenshealth/loader.cfm?url=/commonspot/security/getfile.cfm&PageID=14290.
- How did welfare reform affect the health insurance coverage of women and children?.Health Services Research. 2006; 41: 486-506
Centers for Disease Control and Prevention. (2006). Live births, birth rates, and fertility rates, by race: United States, 1909–99. Available at: http://www.cdc.gov/nchs/data/statab/t991x01.pdf. Accessed October 2006.
- State welfare reform policies and declines in health insurance.American Journal of Public Health. 2000; 90: 900-908
Currie, J. (2004). The take-up of social benefits. NBER Working Paper #10488 Available at http://www.nber.org/papers/w10488.
- Is welfare reform responsible for low-skilled women's declining health insurance coverage in the 1990's?.The Journal of Human Resources. 2006; 41: 495-528
- Changes in prenatal care timing and low birth weight by race and socioeconomic status: Implications for the Medicaid expansions for pregnant women.Health Services Research. 2001; 36: 373-403
- Timing of insurance and use of prenatal care among low-income women.American Journal of Public Health. 2002; 92: 423-427
- Health insurance coverage after welfare.Health Affairs. 2000; 19: 175-184
- The impact of welfare reform on insurance coverage before pregnancy and the timing of prenatal care initiation.Health Services Research. 2007; 42: 1564-1588
- Supporting work through Medicaid and Food Stamps.in: Blank R. Haskins R. The New World of Welfare. Brookings Institution Press, Washington, DC2001
- Millions of mothers lack health insurance coverage in the United States.International Journal of Health Services. 2002; 32: 89-106
- The effect of welfare reform on the pre-pregnancy insurance status of public aid recipients: Women in the Medicaid GAP.Journal of Health Care for the Poor and Underserved. 2006; 17: 162-179
- The effect of providing health coverage to poor uninsured pregnant women in Massachusetts.Journal of the American Medical Association. 1993; 269: 87-91
- Getting in, not getting in, and why: Understanding SCHIP enrollment. Assessing the New Federalism, Occasional Paper #66.Urban Institute, Washington, DC2003
- The impact of the Medicaid expansions for pregnant women: A synthesis of the evidence.Medical Care Research and Review. 2001; 58: 3-30
- Recommendations to improve preconception health and health care—United States. A report of the CDC/ATSDR Preconception Care Work Group and the Select Panel on Preconception Care.Morbidity and Mortality Weekly Reports Recommendations and Reports. 2006; 55: 1-23
- Welfare reform and health insurance families.Journal of Health Economics. 2003; 22: 959-981
Kaestner, R., & Kaushal, N. (2001). Immigrant and native responses to welfare reform. NBER Working Paper #8541.
Kaiser Commission on Medicaid and the Uninsured. (2002). Transitional Medical Assistance (TMA): Medicaid issue update. Washington, DC: Kaiser Commission on Medicaid and the Uninsured, The Henry J. Kaiser Family Foundation. Available at: http://www.kff.org/medicaid/loader.cfm?url=/commonspot/security/getfile.cfm&PageID=14130.
Kaiser Commission on Medicaid and the Uninsured. (2008). Health coverage for children and families in Medicaid and SCHIP: State efforts face new hurdles: A 50-state update on eligibility rules, enrollment and renewal procedures, and cost-sharing practices in Medicaid and SCHIP in 2008. Washington, DC: Kaiser Family Foundation. Available at: http://www.kff.org/medicaid/upload/7740.pdf. Accessed February 21, 2008.
Kaiser Family Foundation. (2007a, October). Medicaid's role for women. Issue brief. An update on women's health policy. Available at: http://www.kff.org/womenshealth/upload/7213_03.pdf. Accessed February 14, 2008.
Kaiser Family Foundation. (20007b, October). Medicaid's role in family planning. An update on women's health policy. Available at: http://www.kff.org/womenshealth/upload/7064_03.pdf. Accessed February 21, 2008.
Kaiser Family Foundation and Alan Guttmacher Institute. (2005). Medicaid: A critical source of support for family planning in the United states. Issue Brief KFF/AGI/ New York and Washington, April. 2005.
Klerman, L. (2005, June 21). Interconception care. Presentation to the National Summit on Preconception Care.
- Socio-economic disparities in pregnancy outcome: why do the poor fare so poorly?.Paediatric and Perinatal Epidemiology. 2000; 14: 194-210
- The effects of Florida's Medicaid eligibility expansion for pregnant women.American Journal of Public Health. 1998; 1998: 371-376
- Births: Final data for. 2007; 2005: 6
- Effects of Medicaid eligibility expansion on prenatal care and pregnancy outcome in Tennessee.Journal of the American Medical Association. 1990; 264: 2219-2223
- Effect of Medicaid expansions on preterm birth.American Journal of Preventive Medicine. 1997; 13: 292-297
- Prenatal care initiation among very low-income women in the aftermath of welfare reform: Does pre-pregnancy Medicaid coverage make a difference?.Maternal and Child Health Journal. 2007; 11: 11-17
- The distribution of health insurance coverage among pregnant women, 1990–1997.March of Dimes, Washington, DC1999
US Department of Health and Human Services. (2004). U.S. Welfare Caseloads Information.’' Available at: http://www.acf.dhhs.gov/news/stats/newstat2.shtml.
Biography
Kosali I. Simon is an Associate Professor in the Department of Policy Analysis and Management at Cornell University and Faculty Research Fellow of the National Bureau of Economic Research. She studies the impact of regulatory programs designed to make health insurance and health care more available for vulnerable populations. She is the 2007 recipient of the John D. Thompson Prize for Young Investigators.
Arden Handler, DrPH, is a Professor of Community Health Sciences and Maternal and Child Health at the University of Illinois School of Public Health (UIC-SPH). She is PI and co-director of the Maternal and Child Health Training Program at UIC-SPH. She has written extensively in the area of women's access to, use of, and satisfaction, with prenatal care, and other issues of concern surrounding the pregnancy period.
Article info
Publication history
Accepted:
August 15,
2008
Received in revised form:
August 15,
2008
Received:
March 20,
2008
Footnotes
The authors have no direct financial interests that might pose a conflict of interest in connection with the submitted manuscript.
Identification
Copyright
© 2008 Jacobs Institute of Women's Health. Published by Elsevier Inc. All rights reserved.
