Abstract
Despite plausible mechanisms, little research has evaluated potential changes in health behaviors in response to expansions in public insurance coverage of the 1980s and 1990s targeted at low-income families. In this paper, we provide the first national study of the effects of Medicaid expansions on health behaviors for pregnant women, which is a group of particular interest given evidence of the importance of prenatal health to later life outcomes. In doing so, we also add to the sparse literature on ex ante moral hazard, which is nearly always mentioned as a theoretical consequence of health insurance, though relatively few empirical studies have assessed its importance. We exploit exogenous variation from the Medicaid income eligibility expansions for pregnant women during late-1980s through mid-1990s to examine the effects of these policy changes on smoking, weight gain, and other maternal health indicators. We find that the 13 percentage point increase in Medicaid eligibility during the study period was associated with approximately a 3% increase in smoking and a small increase in pregnancy weight gain for most of the sample. The increase in smoking, which is a significant cause of poor infant health, may partly explain why Medicaid expansions have not been associated with substantial improvement in infant health.
Similar content being viewed by others
Notes
A recent literature has developed that examined the long-term effects of Medicaid eligibility during childhood. Much of this literature finds beneficial effects of childhood Medicaid eligibility. See for example: Brown et al. (2015), Miller and Wherry (2015), Wherry and Meyer (2016), and Wherry et al. (2017).
The expansion of Medicaid may also cause a change in the composition of women who give birth and this may affect health behaviors. We assess empirically whether such compositional change occurred and find no evidence that the types of women giving birth changed significantly in response to the Medicaid expansions.
There was little expansion of Medicaid for pregnant women post 1996.
In the Appendix, we report results obtained using an alternative way of constructing eligibility: using fixed (time-invariant) state-specific samples instead of a fixed national sample of pregnant women to construct eligibility. Results are not sensitive to this choice.
We control for state excise taxes on cigarettes to account for the possibility that some states may be raising these taxes over the sample period to fund the Medicaid expansions. As we discuss later, estimates and conclusions are not materially affected if we exclude all time-varying state controls.
Estimates and patterns are similar if we alternately employ indicators for eligibility quintiles.
Medicaid participation among less than high school-educated pregnant women is 2.4 times greater relative to higher educated pregnant women, and 6.1 times greater among unmarried pregnant women relative to those who are married, based on the Current Population Survey.
These outcomes are not reported by some states (for instance, CA, IN, NY, SD, OK) over all or part of our sample period. We exclude births occurring in these states when analyzing these behaviors.
The threat to validity from measurement error stems from the measurement error being correlated with Medicaid expansions. While this is possible if greater contact with physicians (more prenatal care), induced by the expansions, lead expectant mothers to increase their reporting of smoking, we do not find any evidence of this when we account for the joint changes in prenatal care and health behaviors.
These guidelines were revised in 2009, in reflection of the obesity epidemic, for pre-pregnancy obese women, recommending their pregnancy-related weight gain to be limited to 11–20 lbs. The premise was that heavier women could gain less weight and still deliver a normal-weight infant.
Approximately 1% of the sample reported zero weight gain during pregnancy. We exclude these observations, as they may represent reporting and measurement error. Estimates are virtually unchanged by this restriction.
Prior to states using revised birth certificates starting in 2003, the data did not differentiate between pre-pregnancy and gestational diabetes. However, based on disaggregated data, almost 90% of these diabetes cases represent gestational diabetes.
We match Medicaid eligibility based on year of pregnancy for those records where gestation straddles adjacent years. This choice was motivated by evidence that prenatal smoking is most responsive during the first trimester (Colman et al. 2003; Colman and Joyce 2003). Alternately, we imputed pregnancy year based on birth year and a standard gestation of 40 weeks for all births. Results are not sensitive to using this alternate measure.
We also used a time-invariant, state-specific sample to exploit differences in state income distributions that will affect eligibility. Given that the measure of eligibility is based only on state variation in timing and magnitude of Medicaid expansions, using state samples introduces no additional variation that would bias estimates. Any endogeneity from the state income distribution is purged through the state fixed effects. Our estimates are robust to using a state instead of a national sample (reported in the Appendix).
We chose to define the group cells for Medicaid eligibility by race and age (in addition to state and year) because of the large racial and age-related differences that exist in insurance coverage and health behaviors. Furthermore, the race- and age-specific samples exploit differences in the income distribution across these factors. If there is a different distribution of income by race, then that variation helps identify the association between Medicaid eligibility and outcomes and results in more precise estimates. For instance, a given shift in federal poverty line (FPL)-based eligibility in a specific state may lead to differential shifts for different races due to differences in race-specific and age-specific income distributions. Thus, Medicaid eligibility constructed for race- and age-specific samples provides a measure of the policy instrument with greater and more accurate variation when matched to individual records, which raises the precision of estimates. Again, using this measure of eligibility does not alter the threats to internal validity of the measure.
The increase in prenatal smoking may reflect a decrease in quitting among women who smoked prior to pregnancy and/or an increase in initiation. However, it is unlikely that results are driven by initiation. The majority of smokers initiate smoking prior to age 18, and virtually all initiate prior to age 21. Our results are robust to excluding pregnant women ages 18–20. Furthermore, very few women start smoking during pregnancy.
It is worth noting that the standard errors are also relatively comparable between the basic and the extended models (Tables 2 and 3). In order to evaluate collinearity, we regressed the Medicaid eligibility instrument on maternal characteristics and state and year fixed effects, which yielded an R-square of 0.883, and a variance inflation factor (VIF) of 8.5. Adding time-varying state covariates to the regression yielded an R-square of 0.886 and a VIF of 8.8. Finally, adding the interactions of age and race with the state and year fixed effects (state*race, state*age, year*race, year*age) to the regression yielded an R-square of 0.888 and a VIF of 8.9. While the R-square and VIF from these auxiliary regressions are somewhat high, standard errors of estimates remain small enough to reliably detect small effect sizes. Furthermore, adding the full set of time-varying state covariates and the interactions between state and year fixed effects and age and race does not materially reduce the identifying variation in eligibility (R-square increases only slightly from 0.883 to 0.888). These results indicate that we do not lose identifying variation much with the addition of the state controls and extended fixed effects. The reason why the standard errors either remain the same or even decrease in the fuller specification is because the added controls also reduce the residual variance. In the end, there is sufficient independent variation in the eligibility measure to obtain estimates precise enough to be informative of economically and clinically important effects.
We also note that we find a significant, negative effect of the state cigarette excise tax on maternal smoking, which is consistent with the literature (Colman et al. 2003).
Models for weight gain do not control for gestation because gestation is potentially impacted by Medicaid eligibility and therefore endogenous. However, we re-estimated all models including gestation. Estimates are similar to those reported in Table 3, and remain significant, but somewhat smaller.
It should be noted that differential effects are possible and may reflect heterogeneity in the behavioral response across the education distribution even if Medicaid take-up rates are similar.
This contrasts with the study by Dave and Kaestner (2009), who find that insurance improves health behaviors through greater physician contact. Part of the reason why there may not be strong counteracting effects in our case relates to the difference in study populations. Dave and Kaestner (2009) investigate the effects of Medicare and their population of interest is the elderly upon receiving public insurance at age 65, whereas we study the Medicaid-eligible population of low-educated single mothers ages 18–39. For instance, Dave et al. (2008) study the direct effects of these expansions on prenatal care and find that a 20 percentage-points increase in eligibility is associated with an insignificant 0.06 additional prenatal visit, about half a percentage point increase (also statistically insignificant) in the probability of receiving adequate care, and a significant 0.3 percentage point decrease in late (3rd trimester) prenatal care initiation, among low-educated mothers. Hence, it does not appear that the expansions resulted in significantly higher contact between pregnant women and the medical care community, at least for the average mother, which may explain why the results are not sensitive to controlling for prenatal care.
Specifically, we assessed movement of the probability density across four categories comprising combinations of adequate prenatal care and prenatal smoking (1: 1st trimester care + No smoking; 2: 1st trimester care + Smoking; 3: No 1st trimester care + No smoking; 4: No 1st trimester care + Smoking), and similar combinations of adequate prenatal care and weight gain thresholds. With respect to smoking, the marginal effects suggest an increase in the joint probability of prenatal care and smoking (category 2) and the probability of prenatal care and no smoking (category 1) relative to the other two categories—thus an increase in early initiation, conditional on smoking. A 13 percentage points increase in eligibility would move about 1.5% late initiators into early care, based on the MNL estimates. We also find that the expansions are associated with an increase in the probability of prenatal care and smoking (category 2) and the probability of no prenatal care and smoking (category 4) relative to the other two categories; thus Medicaid is associated with an overall increase in prenatal smoking (and, of more than 5 cigarettes daily), and it is validating that these effect magnitudes are highly similar to the OLS estimates (0.07 and 0.1) in Tables 2 and 4 for the two sub-populations. Turning to the MNL models that assess non-linear effects of eligibility, we find as before that the patterns and effect magnitudes diminish at higher levels of eligibility. MNL estimates are not reported, and results are available from the authors upon request.
The low prevalence of smoking for this group limits the statistical power of this falsification test.
In contrast, Joyce et al. (1998), using pooled cross-sections of states, find that expansions in the income thresholds for Medicaid eligibility between 1987and 1991 are associated with a 5% increase in the birthrate among white women, but not among black women.
We noted earlier that a 13% increase in Medicaid eligibility is associated with 0.9 percentage points increase in any prenatal smoking. Scaling this up by about 7.7 (100/13) yields the estimate reported in Table 2 (6.8 percentage points increase in prenatal smoking associated with a 100% increase in eligibility). Assuming a Medicaid take-up rate of 50%, this effect then needs to be further scaled up by 2 to arrive at the TOT effect, of about 13 percentage points, of actually obtaining Medicaid. Thus, the TOT can be derived by scaling the effect of a 13% increase in Medicaid eligibility by a factor of about 15 (0.9 percentage point times 15 = 13 percentage points).
These estimates are based on average out-of-pocket spending for non-Medicare Medicaid/publicly insured families with a health problem of $250, compared to about $1100–1200 with employer/other private coverage, and $550 without coverage (Tables 6 and 8 of Merlis 2002).
Average annual personal income from all sources over the sample period was $7650 for low-educated pregnant women identified in the CPS.
We assume that Medicaid has no effect on the pregnancy decision, which is consistent with some of the prior literature and which we confirm with our data (discussed in the text).
We do not treat medical care in the first period, for example maternal prenatal care, as uncertain and affected by insurance because such care is preventive and its effects are assumed known—i.e., there is no uncertainty. Including Medicaid coverage for maternal care prior to birth would not change any of the predictions described below.
We incorporate the demand for prenatal health by allowing first-period medical care to influence the probability of an adverse event at birth.
Here, we analyze the case where consumption has a negative effect on infant health, but as noted, some forms of consumption may have beneficial effects.
References
Adams EA, Markowitz S, Dietz PM, Tong VT (2013) Expansion of Medicaid covered smoking cessation services: maternal smoking and birth outcomes. Medicare Medicaid Res Rev 3(3):mmrr.003.03.a02
Almond D, Currie J (2011) Killing me softly: the fetal origins hypothesis. J Econ Perspect 25(3):153–172
Angrist JD, Pischke J-S (2009) Mostly harmless econometrics: an Empiricist's companion. Princeton and Oxford
Baicker K, Finkelstein A, Song J, Taubman S (2013) The impact of Medicaid on labor force activity and program participation: evidence from the Oregon health insurance experiment. National Bureau of Economic Research Working Paper # 19547, Cambridge
Bhattacharya J, Bundorf MK, Pace N, Sood N (2011) Does health insurance make you Fat? In: Grossman M, Mocan N (eds) Economic aspects of obesity. University of Chicago Press, pp 35–64
Brown DW, Kowalski AE, Lurie IZ (2015) “Medicaid as an investment in children: what is the long-term impact on tax receipts?” Cambridge. National Bureau of Economic Research Working Paper # 20835, MA
Card D, Dobkin C, Maestas N (2008) The impact of nearly universal insurance coverage on health care utilization and health: evidence from Medicare. Am Econ Rev 98(5):2242–2258
Centers for Disease Control and Prevention (2002) Infant mortality and low birth weight among black and white infants—United States, 1980—2000. Morb Mortal Wkly Rep 51(27):589–592
Chiappori PA (2000) Econometric models of insurance under asymmetric information. In: Dionne G (ed) Handbook of insurance, North Holland
Cohen A, Dehejia R (2004) The effect of automobile insurance and accident liability laws on traffic fatalities. J Law Econ 47(2):357–393. https://doi.org/10.1086/422978
Colman GJ, Joyce T (2003) Trends in smoking before, during, and after pregnancy in ten states [PRAMS]. Am J Prev Med 24:29–35
Colman G, Grossman M, Joyce T (2003) The effect of cigarette excise taxes on smoking before, during, and after pregnancy. J Health Econ 22:1053–1072
Courbage C, Coulon A (2004) Prevention and private health insurance in the U.K. Geneva Pap Risk Ins 29(4):719–727. https://doi.org/10.1111/j.1468-0440.2004.00313.x
Courtemanche C, Marton J, Ukert B, Yelowitz A, Zapata D (2017) Early effects of the Affordable Care Act on health care access, risky health behaviors, and self-assessed health. NBER Working Paper 23269
Currie J (2009) Healthy, wealthy, and wise: socioeconomic status, poor health in childhood, and human capital development. J Econ Lit 47(1):87–122
Currie J, Gruber J (1996a) Saving babies: the efficacy and cost of recent changes in the Medicaid eligibility of pregnant women. J Polit Econ 104(6):1263–1296. https://doi.org/10.1086/262059.
Currie J, Gruber J (1996b) Health insurance, eligibility, utilization of medical care, and child health. Q J Econ 111(2):431–466
Curtin SC, Osterman M, Uddin SF, Sutton SR, Reed PR (2013) Source of payment for the delivery: births in a 33-state and District of Columbia reporting area, 2010. Natl Vital Stat Rep 62(5):1–20
Cutler DM, Gruber J (1996) Does insurance crowd out private insurance? Q J Econ 111(2):391–430
Cutler DM, Zeckhauser R (2000) Insurance markets and adverse selection. In: Cutler AJ, Newhouse JP (eds) Handbook of health economics, North Holland
Dave D, Kaestner R (2009) Health insurance and ex ante moral hazard: evidence from Medicare. Int J Health Care Finance Econ 9:367–390
Dave D, Decker SL, Kaestner R, Simon KI (2008) “Re-examining the effects of Medicaid expansions for pregnant women” National Bureau of Economic Research Working Paper Series, No 14591
Dave D, Decker SL, Kaestner R, Simon KI (2011) The effect of Medicaid expansions on health insurance coverage of pregnant women: an analysis using deliveries. Inquiry 47:315–330
Dave D, Decker SL, Kaestner R, Simon KI (2014) The effect of Medicaid expansions in the late 1980s and early 1990s on the labor supply of pregnant women. Am J Health Econ 1(2):165–193
DeLeire T, Lopoo LM, Simon KI (2011) Medicaid expansions and fertility in the United States. Demography 48:725–747
Dubay L, Kenney G (1997) Did the Medicaid expansions for pregnant women crowd-out private insurance? Health Aff 16(1):185–193
Fortin B, Lanoie P (2000) Incentive effects of worker’s compensation: a survey. In: Dionne G (ed) Handbook of insurance, North Holland
Gruber J, Simon K (2008) Crowd-out ten years later: have recent public insurance expansions crowded out private health insurance? J Health Econ 27:201–217
Gruber J, Yelowitz A (1999) Public health insurance and private savings. J Polit Econ 107(6):1249–1274
Hadley J (2003) Sicker and poorer—the consequences of being uninsured: a review of the research on the relationship between health insurance, medical care use, health, work, and income. Med Care Res Rev 60(2 Suppl):3S–75S
Howell EM (2001) The impact of the Medicaid expansions for pregnant women: a synthesis of the evidence. Med Care Res Rev 58(1):3–30
Howell EM, Kenney GM (2012) The impact of the Medicaid/CHIP expansions on children: a synthesis of the evidence. Med Care Res Rev 20:1–25
Institute of Medicine (2009) Weight gain during pregnancy: reexamining the guidelines, report brief, May 2009
Jarlenski M, Bleich SN, Bennett WL, Stuart SA, Barry CL (2014) Medicaid enrollment policy increased smoking cessation among pregnant women but had no impact on birth outcomes. Health Aff 33:997–1005
Joyce T, Kaestner R, Kwan F (1998) Is Medicaid Pronatalist? The effect of eligibility expansions on abortions and births. Fam Plan Perspect 30:108–113
Kaestner R, Carroll A (1997) New estimates of the labor market effects of workers’ compensation insurance. South Econ J 63(3):635–651. https://doi.org/10.2307/1061099
Kelly IR, Markowitz S (2009) Incentives in obesity and health insurance. Inquiry 46(4):418–432
Kenkel DS (2000) Prevention. In: Culyer AJ, Newhouse JP (eds) Handbook of health economics, North Holland
Kenkel DS, Schmeiser M, Urban C (2014) Is smoking inferior?: evidence from variation in the earned income tax credit. J Hum Resour 49(4):1094–1120
Klick J, Stratmann T (2007) Diabetes treatments and moral hazard. J Law Econ 50:519–538
Kotelchuck M (1994) The adequacy of prenatal care utilization index: its US distribution and association with low birthweight. Am J Public Health 84(9):1486–1489
Leininger L, Levy H, Schanzenbach D (2010) Consequences of SCHIP expansions for household well-being. Forum Health Econ Policy 13(1):1–30
Levy H, Meltzer D (2004) What do we really know about whether health insurance affects health? In: McLaughlin C (ed) Health policy and the uninsured. Urban Institute Press, Washington D.C, pp 179–204
Maclean JC, Pesko MF, Hill SC (2017) The effect of insurance expansions on smoking cessation medication prescriptions: evidence from ACA Medicaid expansions. NBER Working Paper # 23450
Merlis M (2002) Family out-of-pocket spending for health services: a continuing source of financial insecurity, The Commonwealth Fund, www.cmwf.org
Miller S, Wherry L (2015) The long-term effects of early life Medicaid coverage. Working paper. University of Michigan
National Commission to Prevent Infant Mortality (1988) Death before life: The tragedy of infant mortality, Washington, D.C.: National Commission to Prevent Infant Mortality
Newhouse JP (1993) Free for all? Lessons from the RAND health insurance experiment. In: A RAND study. Harvard University Press, Cambridge
Nielsen SS, Dills RL, Glass M, Mueller BA (2014) Accuracy of prenatal smoking data from Washington state birth certificates in a population-based sample with cotinine measurements. Ann Epidemiol 24(3):236–239
Ruser JW (1985) Workers’ compensation insurance, experience-rating, and occupational injuries. RAND J Econ 16(4):487–503. https://doi.org/10.2307/2555508
Ruser JW (1991) Workers’ compensation and occupational injuries and illnesses. J Labor Econ 9(4):325–350. https://doi.org/10.1086/298272
Simon K, Soni A, Cawley J (2017) The impact of health insurance on preventive care and health behaviors: evidence from the first two years of the ACA Medicaid expansions. Journal of Policy Analysis and Management 36(2):390–417
Sommers B, Kronick R, Finegold K, Po R, Schwartz K, Glied S (2012a) Understanding participation rates in Medicaid: implications for the Affordable Care Act, ASPE issue brief, Office of the Assistant Secretary for Planning and Evaluation, https://aspe.hhs.gov/basic-report/understanding-participation-rates-medicaid-implications-affordable-care-act
Sommers B, Baicker K, Epstein A (2012b) Mortality and access to care among adults after state Medicaid expansions. N Engl J Med 367:1025–1034
Stanciole AE (2008) Health insurance and life style choices: identifying the ex-ante moral hazard. Geneva Pap 33:627–644
Tong VT, Dietz PM, Farr SL, D’Angelo DV, England LJ (2013) Estimates of smoking before and during pregnancy, and smoking cessation during pregnancy: comparing two population-based data sources. Public Health Rep 128(3):179–188
Wherry LR, Meyer BD (2016) Saving teens: using a policy discontinuity to estimate the effects of Medicaid eligibility. J Hum Resour 51:556–588
Wherry LR, Miller S, Kaestner R, Meyer BD (2017) Childhood Medicaid coverage and later life health care utilization. Review of Economics and Statistics. Forthcoming
Zavodny M, Bitler M (2010) The effect of Medicaid eligibility expansions on fertility. Soc Sci Med 71:918–924
Acknowledgements
The authors would like to thank Briggs Depew, Brad Humphreys, Christina Marsh, Joshua Pinkston, Jennifer Trudeau, Wen You, Joshua Graff Zivin, seminar participants at Montana State University, University of Gothenburg, and the Danish Institute for Local and Regional Government Research, and the anonymous referees for helpful comments and suggestions.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Responsible editor: Junsen Zhang
Appendix: Theoretical model
Appendix: Theoretical model
The summary of the theoretical predictions provided in the text are derived from the following model. In this model, mothers care about consumption, leisure, and child health. There are two periods that span the pre- to post-birth period.Footnote 31 In this model, child health in the post-birth period is uncertain. With probability (π), the child may experience an adverse health shock (z) that lowers child health and can be offset (repaired) with medical care (m1). Medical care in the first period can also be used to alter the probability of an adverse health event in period two.Footnote 32 First-period maternal consumption, for example, nutrition and smoking, may also affect the probability of an adverse health shock in period two. Health insurance is particularly important because it is used to buy medical care after the birth in the case of an adverse outcome.
A woman’s expected utility is described by:
Equation (3) reflects the fact that there are two periods: prior to birth (t = 0) and after birth (t = 1). Utility is a function of consumption (x), leisure (l), and child health (c) in each period, although in period 0 the child is not born and so child health does not enter the utility function.Footnote 33 The discount rate is denoted by (β).
The woman’s budget constraint is given by:
Lifetime income is spent on: consumption (x) in periods 0 and 1; medical care (m) in periods 0 and 1 with price of medical care denoted by pm; and the quantity of health insurance (α) in period one. The interest rate is denoted by (r). Health insurance reduces the price of medical care and is financed out of earnings (w). The cost of health insurance also includes a loading charge (f/α) where f is a fixed cost of administering health insurance.
The constrained choice problem is given by:
The first-order condition for consumption (health behavior) in the first period is given by:
Equation (6) is the usual equilibrium condition. The left-hand side is the benefit of first period consumption and the right hand side is the cost, which includes the higher probability of lower second period utility because of an adverse health shock to child caused by first period consumption and the greater spending on second period medical care due to a higher probability of an adverse health shock \( \Big({\tilde{u}}_1 \) denotes period 1 utility if sick).Footnote 34
Equation (6) also illustrates the problem of ex ante moral hazard. Health insurance (α) lowers the price of medical care, and a lower price of medical care reduces the cost of first period consumption that is related to the probability of a child health shock. The first-order conditions for medical care are:
First-period medical care is valued because it reduces the probability of an adverse event and raises expected utility in the second period (\( {\tilde{u}}_1 \) denotes period 1 utility if sick). A lower probability of an adverse event also reduces second-period medical expenditures. Both of these marginal benefits are equated to the price of medical care. Second-period medical care raises utility by improving child health and this marginal benefit is equated to the price of medical care.
The first-order condition for health insurance is:
According to (8), a woman chooses insurance to equate the expected medical expenditures to the cost of insurance, which is the wage offset plus loading cost. Again, the expansion of Medicaid makes insurance free and there is an income effect associated with this change. The income effect may also be driven by Medicaid-induced shifts in labor supply.
Rights and permissions
About this article
Cite this article
Dave, D.M., Kaestner, R. & Wehby, G.L. Does public insurance coverage for pregnant women affect prenatal health behaviors?. J Popul Econ 32, 419–453 (2019). https://doi.org/10.1007/s00148-018-0714-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00148-018-0714-z