Minimum Wages, the Earned Income Tax Credit, and Mental Health Around Pregnancy

ABSTRACT

This study estimates the effects on perinatal mental health of the state's minimum wage and earned income tax credit (EITC), controlling for other policies and state‐level factors. Using data from the Pregnancy Risk Assessment Monitoring System for 2012–2018 births we find robust evidence that minimum wages and EITC levels reduce depression before pregnancy and suggestive evidence of minimum wages reducing postpartum depression, at least for married respondents. Our estimates suggest that a one dollar increase in the minimum wage ($100 increase in the state EITC) reduces pre‐pregnancy depression by roughly 8.5% (1.5%). These findings stand up to standard robustness and falsification tests, including event study analyses, a wide array of alternative specifications, and finding no effect for those unlikely to benefit (e.g., college‐educated respondents). A supplementary analysis using data from the Behavioral Risk Factor Surveillance System suggests that state EITC levels may reduce mental distress during pregnancy. We investigate possible mechanisms by providing a descriptive analysis of the income and work behavior of such households, which shows the wide reach of these policies, and investigating a broad set of outcomes from the PRAMS, such as financial stressors, health insurance and birth outcomes.

1. Introduction

While mental health issues represent a major burden at any point in life, the medical and social changes associated with pregnancy and childbirth make the destabilizing threats of mental illnesses of particular concern during the perinatal period. Both young adults and women have considerably higher rates of depression than other subgroups (National Institute on Mental Health et al. ²⁰²³). Likewise, diagnoses for perinatal mood and anxiety disorders have increased substantially in the past 15 years (Zivin et al. 2024). One such disorder, postpartum depression (PPD) has a prevalence rate of around 13–19%, depending on the specific measure used, and is even higher for low‐income people (O’Hara and McCabe ²⁰¹³; Goyal et al. ²⁰¹⁰; Dearing et al. ²⁰⁰⁴). Yet, little research exists on the role that anti‐poverty policies play on perinatal mental health; this void is even more notable for labor market policies despite the high labor force participation of this population. Our study aims to fill that gap.

Policy advocates have highlighted the severity of perinatal depression and called on further government intervention (Vericker et al. 2010; Chester et al. ²⁰¹⁶). Several states have applied to extend public health insurance coverage 1‐year postpartum, with PPD cited as a major reason to do so (e.g., Boston Globe Editorial Board ²⁰²¹). Labor market policies also have the potential to affect health care access, both via employer‐sponsored health insurance and income more generally. Moreover, financial stress may create or exacerbate existing feelings of depression. Research suggests the mental health of young women is particularly harmed during economic downturns (Black et al. 2021) and that state minimum wages and Earned Income Tax Credit (EITC) levels have beneficial mental health impacts, especially for women (Horn et al. 2017; Kuroki ²⁰²¹; Dow et al. ²⁰²⁰; Gangopadhyaya et al. ²⁰²⁰). Depression prior to or during pregnancy is a key risk factor for PPD, suggesting that a comprehensive investigation explores mental health during the entire perinatal period.

To our knowledge, this study provides the first empirical evidence of the effect of minimum wages on mental health during any part of the perinatal period. In addition, it is the first to provide comprehensive evidence of the perinatal effects of state EITC policies. While pregnancy and childbirth may reduce labor market participation, the potential effects of these labor market policies remain strong, both through their partners' work experience as well as their own. We also investigate the effects of two programs that likely target this group specifically, the Temporary Assistance for Needy Families (TANF) and the Medicaid expansions that occurred as part of the Affordable Care Act (ACA).

Using data from the Pregnancy Risk Monitoring System (PRAMS) during a period of widespread policy changes (2012–18), we find robust evidence that state minimum wages and EITCs improve mental health before pregnancy. The estimated magnitudes are meaningful; for example, a $1 increase in the minimum wage ($100 increase in the state EITC) is estimated to reduce pre‐pregnancy depression 8.5% (1.5%). The findings for postpartum depression are less conclusive but suggestive that these effects may carry over there as well. Our supplementary analysis using the Behavioral Risk Surveillance System (BRFSS) (Centers for Disease Control and Prevention (CDC) ^2012–2018) provides evidence that the EITC has an impact on mental health during the pregnancy (a measure the PRAMS lacks) and investigates the impact of using a broader set of states than is available in the PRAMS. To explore the possible mechanisms, we bring in descriptive evidence from the American Community Survey that shows the wide reach of these policies. We also investigate a broad set of outcomes in the PRAMS, including financial stress, health care and insurance and birth outcomes and find suggestive results that are mostly consistent with past research. Our findings therefore suggest that labor market policies targeting low‐income workers have potentially strong beneficial effects that have to date been overlooked.

2. Background

Depression rates are higher for reproductive age women than the general population (National Institute on Mental Health et al. ²⁰²³). Pregnancy and childbirth may exacerbate this tendency, due to physiological changes as well as environmental factors and stress. Perinatal mood and anxiety disorders (Zivin et al. 2024) and PPD hospitalizations (Franca and McManus ²⁰¹⁸) have all increased in recent years. Perinatal depression “refers to depression occurring during pregnancy or after childbirth,” (https://www.psychiatry.org/patients‐families/peripartum‐depression/what‐is‐peripartum‐depression), while the more commonly used term “post‐partum depression” (PPD) is a specific type that occurs within the 12 months after childbirth and is distinct from the more prevalent and less serious “baby blues” (O’Hara and McCabe ²⁰¹³). Here we use PPD as a shorthand for depressive symptoms observed within 12 months after birth.

Depression before, as well as during, pregnancy leads to a higher risk of developing PPD (Stewart and Vigod 2016). Evidence further suggests that prenatal depression may be associated with poor infant outcomes, such as preterm birth or low birth weight (Accortt et al. 2015; Conway and Kennedy ²⁰⁰⁴). The consequences of postpartum mental illness and “maternal depression” more generally have been more widely studied and are found to include detrimental effects on maternal and child outcomes (Le and Nguyen 2018; Menta et al. ²⁰²³; Slomian et al. ²⁰¹⁹), parenting behaviors (Field 2009), food security (Noonan et al. 2016), housing (Corman et al. 2016; Curtis et al. ²⁰¹⁴) and long‐term economic well‐being (McGovern et al. ²⁰²²).

Efforts to combat perinatal mental illnesses include psychotherapies and medication, both of which have shown promising results (e.g., Stewart and Vigod ²⁰¹⁶; American Psychological Association ²⁰¹²). Anti‐poverty policies and other types of income support may also prevent or combat perinatal depression, both by improving access to these treatments and by alleviating stressors—as poverty (Ridley et al. 2020) and negative income shocks (Black et al. 2021) are risk factors for mental distress. Indeed, a growing literature finds promising mental health benefits of anti‐poverty and labor market policies, especially for women. However, the perinatal period has been mostly overlooked—with the exception of Medicaid, a policy that specifically targets the health of this group.

For Medicaid, evidence exists that the expansions of the 1990s that especially targeted pregnant women improved maternal health outcomes including mental health after the birth (Guldi and Hamersma 2021), while the more recent ACA expansions had considerably weaker effects on PPD (Austin et al. 2022; Margerison et al. ²⁰²¹). However, Margerison et al. (2021) finds that the ACA's Medicaid expansions led to a decrease in depression in the months leading up to pregnancy, which makes sense because these expansions extended coverage to those not pregnant and without children. Both ACA studies use the PRAMS and similar measures to ours. Their findings are supported by evidence that the ACA Medicaid expansions increased mental health treatment and, perhaps, mental health facilities accepting Medicaid (Ortega 2023). Another piece of evidence comes from a Randomized Control Trial in Nigeria in which cash incentives to use perinatal health services reduced PPD (Okeke 2021).

Research on the mental health effects of other safety‐net policies has investigated the broader measures of maternal, parental or general mental health. Temporary Assistance to Needy Families (TANF) targets very low‐income people with children, yet we are unaware of research that focuses on the effects of TANF—specifically—on mental health. However, Schmidt et al. (2023) finds that higher safety‐net levels, measured by a combination of EITC, TANF, and SNAP benefits, can reduce severe psychological distress for single mothers by 5.5% per $1000 in combined benefits. We therefore control for Medicaid and TANF in our analyses since both programs target this group and, like the EITC and minimum wage, vary across states. The 2021 expansion of the Child Tax Credit, which made it refundable and not conditional on earnings, motivated several studies that tend to find beneficial effects on parental mental health (e.g., Pignatti and Parolin ²⁰²⁴), although in their survey Gennetian and Gassman‐Pines (2024, 11) caution that “it is too early to take stock of evidence.” Past research suggests SNAP benefits alone can improve mental health outcomes for women (Munger et al. 2016), as can unemployment insurance benefits (Chen et al. 2023).

Our focus is on the EITC and minimum wage, which are distinct because their benefits depend on labor market participation. As such, they may not seem like obvious candidates for affecting perinatal mental health. In Section 3, however, we present evidence that labor force participation is widespread for this group as well as their partners. This distinction also means their potential effects on mental health operate not only through income and greater access to care but through time use and the (dis)utility of work. Both policies also have features that complicate their predicted impact. Still, research to date finds suggestive evidence of beneficial effects on mental health, especially for women. We now describe each policy and the evidence of its mental health effects.

2.1. Earned Income Tax Credit (EITC)

The EITC is a federal tax policy that is supplemented with similar policies in many states. The federal EITC is refundable (i.e., tax liability can be negative) and is equal to a specific percentage of earnings that increases with the number of children in the household up to three children. The dollar amount received therefore increases with earnings (the “phase‐in” range) until it reaches a maximum (the so‐called “plateau range”). Once earnings exceed a certain level, the credit is reduced as earnings increase (the “phase‐out” range). The federal EITC is sizable and reaches a wide range of income levels; for example, in the last year of our analyses, 2018, a family with three children could receive a maximum of $6431 and would be eligible for some amount of credit until their household income exceeded $54,884. The credit for those without children is substantially smaller than those with children; for example, the 2018 maximum is $519 for childless households and reaches those with incomes below $20,950.

A growing number of states have EITC policies, which are structured as a fraction of the federal EITC and, in 2018, ranged from 0.03 to 0.40. ¹ As summarized in Table 1 and shown in detail in Appendix Table A1, both the number of states offering EITCs and the fraction sizes increased during 2012–18. A distinctive feature of state EITCs is that they are not all refundable. Other nuances arise because the EITC is received as a one‐time payment upon filing taxes. This timing suggests that its effects may be more heavily concentrated in the spring and that new, first‐time parents may be less aware of it—and thus impacted differently—than existing parents. In our main specification, we use the maximum dollar amount of state + federal EITC possible given the respondent's number of children at the time of the outcome. Unlike the fraction, this measure captures the greater benefits accruing to larger families and the much smaller benefits for childless households. In alternative models, we use the state multiplier (fraction) instead and also include the federal EITC amount separately to better isolate the state policy's effects from that of family size or the federal policy. ² We also explore the nuances of refundability, timing and differences between first‐time and existing parents.

TABLE 1.

Summary of PRAMS state & year availability and policy variation during 2012–18.

State	Years in PRAMS	Change in min wage	Change in EITC multiplier
AK	2012–2018	$2.09
AR	2012–13, 2015–16	$2.25
CO	2012–13, 2015–18	$2.56	0.10 b
CT	2014–2018	$2.30	0.045 b
DE	2012–2018	$1.00	a , b
GA	2012–13, 2017–18
HI	2012–2016	$2.85	0.20
IL	2012–2018		0.13 b
IA	2012–2017		0.08 b
KY	2017–2018
LA	2015–2018		a
ME	2012–2018	$2.50	a , b
MD	2012–2017	$2.85	0.03 b
MA	2012–2018	$3.00	0.08 b
MI	2012–13, 2015–18	$1.85	a , b
MO	2012–2018	$0.60
MT	2017	$0.65
NE	2012–16, 2018	$1.75	a , b
NH	2013–2017
NJ	2012–2018	$1.35	0.17 b
NM	2012–2018		a , b
ND	2017–2018
OH	2012, 2014–15	$0.60	0.10
OK	2012–2017		b
OR	2012–13, 2015	$1.95	0.02 b
PA	2012–2018
RI	2012–14, 2016–18	$2.70	a , b
TN	2012–2015
UT	2012–2018
VA	2015–2018		a
WA	2012–2018	$2.46
WV	2012–2018	$1.50
WI	2012–2018		a , b
WY	2012–2018
Average change		$1.05	0.03
Average in 2018		$8.50	0.08
Max in 2018		$11.50	0.37

Note: States with a minimum wage below the federal level are considered to have the federal level of $7.25.

^a Denotes states with an EITC multiplier but did not change the multiplier level during our sample period.

^b Denotes a refundable EITC. 24 of these states also expanded Medicaid eligibility under the ACA (AK, AR, CO, CT, DE, HI, DE, IL, IA, KY, LA, MA, MD, MI, MT, ND, NH, NJ, NM, OH, OR, PA, RI, VA, WA, WV).

Economic research on the EITC first focused on employment related outcomes, as the theoretical effects on labor supply are ambiguous due to the three ranges of the program. Single parents have been consistently shown to increase their labor supply in response to the EITC, resulting in a reduction in poverty rates; the evidence is more limited and mixed for married households (see Hoynes ²⁰¹⁹ for full review). Michelmore and Pilkauskas (2021) is especially relevant with its finding that the most substantial positive maternal labor supply effects are concentrated in those with children under age 3. However, its focus is on unmarried mothers, and to our knowledge no research has focused on the perinatal period. Thus, while it seems likely the EITC improves household economic well‐being, its specific impacts on work behavior and earnings are less clear, at least for married households.

A growing list of other outcomes has been investigated as well, including education, health insurance, fertility, infant and child health, and, of particular interest here, mental health, where the findings have tended to be positive (Hoynes 2019). Two studies show that the federal expansion of the EITC in the early 1990s resulted in improvements in mental health measures and subjective well‐being for women (Evans and Garthwaite 2014; Boyd‐Swan et al. ²⁰¹⁶). Using variation in state EITCs, Gangopadhyaya et al. (2020) confirms these effects and shows they are largely through income and employment effects rather than insurance status changes, while Qian and Wehby (2021) find that only refundable EITCs have a statistically significant impact. State EITCs are also found to reduce deaths by suicide (Dow et al. 2020; Lenhart ²⁰¹⁹); however, the evidence for long‐term impacts on mental health is more mixed (Yoder 2022 vs. Jones et al. ²⁰²²).

For mental health during the perinatal period, the EITC's effect on the infant's health may be an especially critical mechanism, where past research finds mostly beneficial effects. Hoynes et al. (2015) provides evidence that the federal EITC increases birth weight, although subsequent analyses by Dench and Joyce (2020) cast doubt on its veracity. Focusing on state EITCs, Markowitz et al. (2017) finds positive effects on birth weight and gestation that are larger in more generous states. However, Qian and Wehby (2023) finds less robust results in their analysis of contiguous border counties. To our knowledge, the only work that considers the impact of state EITC levels on perinatal maternal health is the limited evidence provided by Morgan et al. (2022), which finds small reductions in PPD.

2.2. Minimum Wages

The bulk of economic studies on minimum wages focus on labor market outcomes, finding mixed results and leading to an ongoing debate in the field (see Neumark and Shirley ²⁰²²; Dube and Lindner ²⁰²⁴ for full reviews). These studies tend to rely on state‐level minimum wage changes, as the U.S. federal minimum wage has remained at $7.25 since 2009. The minimum wage imposed by individual states has increased and sometimes far exceeds the federal level, leading to ample variation (see Table 1 and Appendix Table A2). For example, in the last year of our analyses (2018) state minimum wages ranged from the federal minimum wage of $7.25 to $13.25.

Even more than the EITC, the effects on labor supply, income and economic well‐being are theoretically ambiguous, as the benefit of receiving higher pay per hour may be counteracted by the employment effects of possible layoffs and reduced hours. Most relevant for our study is Godøy et al. (2021), which focuses on the labor supply effects for parents of children under age five and finds positive effects on the employment of single mothers, while other groups' employment are mostly unaffected. Once again, to our knowledge the labor supply effects during the perinatal period have not been investigated.

Research examining the effects on other outcomes, including health insurance and outcomes, is extensive (see Leigh ²⁰²¹; Neumark ²⁰²³ for reviews). Most past work finds that minimum wages have beneficial effects on infant health (e.g., Wehby et al. ²⁰²⁰; see Leigh ²⁰²¹ Table 2 part IV for a summary of studies), a possible mechanism for affecting perinatal mental health. Evidence also suggests that minimum wages reduce financial strain and poverty‐related stress, while the effects on health insurance and health care is mixed (Leigh 2021, Table 2). Past work on the mental health impacts of minimum wages suggests a beneficial relationship, especially for women. Horn et al. (2017) find state minimum wages are associated with decreases in “not good” mental health days among women, but not men, using the BRFSS. Kuroki (2021) extends Horn et al. (2017) and suggests state minimum wages reduce “extreme mental distress.” Deaths by suicide are also shown to decrease with increases in state minimum wages (Dow et al. 2020). Andrea et al. (2020) finds a reduction in financial stress, such as trouble paying bills, following increases in sub‐minimum wages for people who have recently given birth, a potential pathway for mental health improvements. However, work focusing on the least educated, young adults, 18–25 years old, finds no significant effect in mental health outcomes (Allegretto and Nadler 2020).

TABLE 2.

Summary statistics of income and labor force information for less‐than college‐educated women, aged 18–45 during 2012–2018.

A. Across 3 data sources	PRAMS (12 months before birth)			BRFSS (currently pregnant)			ACS (with a child < 1 year old)
	Total	Married	Unmarried	Total	Married	Unmarried	Total	Married	Unmarried
Number of observations	108,260	48,962	59,298	10,522	5084	5438	137,031	80,601	56,430
Mean household income	$32,926	$42,854	$23,192	$33,366	$43,224	$24,972	$54,427	$64,489	$42,444
—With ACS top‐coded							$44,628	$51,778	$36,111
% currently working				46.2%	45.2%	46.8%	49.5%	46.6%	52.8%
% in labor force							57.3%	51.0%	64.8%
% worked any last year							61.8%	57.5%	66.8%
% worked any during pregnancy a	62.6%	60.8%	64.1%
% of men who worked any last year							93.6%	94.9%	89.9%

B. ACS for women working in last year	Total ACS			Hourly wage ≤ $8			Wage within $1 of State's MW
	Total	Married	Unmarried	Total	Married	Unmarried	Total	Married	Unmarried
As a % of women who worked in last year	100.0%	100.0%	100.0%	16.8%	11.8%	23.9%	27.4%	22.5%	34.3%
Median hours worked	1545	1648	1339	1184	1155	1188	1320	1320	1287
Median personal labor income	$15,000	$20,000	$12,600	$6000	$6000	$5800	$8000	$9600	$7800
Median household income	$42,050	$52,700	$30,000	$30,000	$38,630	$23,800	$36,600	$47,600	$27,000
Percent estimated to receive EITC b	57.9%	48.5%	69.1%	85.1%	71.0%	93.6%	76.5%	57.0%	92.0%
Amount of estimated EITC (if > 0) b	$3120	$3326	$2948	$2834	$3214	$2662	$2897	$3092	$2803

Note: Data comes from the PRAMS, BRFSS, and ACS. In all cases observations are limited to those without a 4‐year college degree and between the ages of 18 and 45. All numbers presented are weighted using survey weights.

^a The PRAMS contains a measure of working during the pregnancy for only 9 states and limited years.

^b The EITC measure is calculated by IPUMS using the NBER TAXSIM calculator.

Outside of the U.S., studies examining the introduction and subsequent increases in the UK's national minimum wage have yielded mixed results, with two finding a reduction in mental illness, citing stress relief and changes in health behaviors (Lenhart 2017; Reeves et al. ²⁰¹⁷) and two others finding no impact (Kronenberg et al. 2017; Maxwell et al. ²⁰²²). Exploiting the sub‐national variation in minimum wages in Canada, Bai and Veall (2023) provide evidence of beneficial effects, although mostly for men.

3. Potential Effects of Labor Market Policies on Perinatal Mental Health

The research to date on the mental health effects of the EITC and minimum wages is therefore promising but decidedly mixed; it is also nearly non‐existent for the time surrounding pregnancy. For these policies to have a direct effect during this period, either the person giving birth or a member of their household must be connected to the labor market in some way. Health impacts and additional time demands of pregnancy and childbirth may reduce labor supply and thus limit the effect on this group. We therefore first explore the likely reach of these policies before discussing their possible mechanisms for affecting mental health.

3.1. Descriptive Evidence of Perinatal Household Income and Work Behavior

Because our primary data sources, the PRAMS and BRFSS, contain little information on household income or labor market behavior, we turn to the American Community Survey (ACS) for additional evidence. The ACS has detailed information on income and work behavior but it does not identify pregnancy or childbirth. Instead, it asks if there is a child in the household under 1 year of age, which we use as a proxy for individuals/households that likely gave birth in the last year. In contrast, the PRAMS and BRFSS each have a single income measure that is categorical, vaguely defined and substantively top‐coded. ³ The BRFSS asks if the respondent is currently pregnant and if the person is currently working for wages. The PRAMS differs in that it does not contain all states and years (see Table 1) and the income measure refers to the 12 months before the child was born. In a small subset of states and years, it asks if the mother worked during pregnancy. ⁴

Given these substantial differences, we first explore the comparability of the key measures for our main analysis group (less than college educated women between the ages of 18 and 45), reported in the top panel of Table 2. ⁵ The mean of household income is reasonably similar between the PRAMS and BRFSS, especially given the measures' limitations as (different) categorical variables, whereas household income is considerably larger in the ACS. This difference is partly due to the top‐coding in the PRAMS & BRFSS; imposing the same restriction on the ACS cuts the difference almost in half, as shown in the third line. The rest is likely due to the more detailed and precise income questions in the ACS, the measurement error created by using the category midpoints in the PRAMS & BRFSS, and the different time intervals spanned (PRAMS is 12 months before birth, BRFSS is annual at time of pregnancy, ACS is proxying for up to 1 year after birth).

In both the BRFSS and ACS, nearly half report currently working. These percentages likely understate the potential reach of labor market policies, as an even higher percentage in the ACS report themselves as in the labor force and even more report working in the last year. The limited evidence in the PRAMS tells a similar story, with nearly 2/3rds reporting having worked during pregnancy. These policies can also manifest through their partner's or other household member's work behavior (if present). To explore this avenue, we report the percent of comparable men (same demographic group and with a child under 1 year old) by marital status who report working in the last year. The percentage rates are much higher than that of women and are 90% or more, which suggests that the vast majority of these households have at least one member who is working and thus potentially affected by these policies.

The bottom panel reports the median annual hours worked and earnings, as well as total household income and estimated EITC receipt, ⁶ for women who worked in the last year. We also report these measures for the subgroups of women most likely affected by minimum wages: (1) those whose average wage is less than $8 (close to the federal minimum), and (2) those whose average wage is within $1 of the minimum wage in that state and year. ⁷ 27.4% of women who worked in the last year have an average hourly wage within $1 of their state's minimum wage, and they report a median 1320 h of work. A $1 increase in the minimum wage could therefore increase total household income by 3.6% (=$1320/$36,600), holding hours of work constant. The impact is even larger if her labor supply increased, as found by Godøy et al. (2021) for parents of young children, or if her partner is also working a low‐wage job. Consistent with past research, the potential effects appear even stronger for unmarried women, who are more likely to be working at a low wage job and whose overall household income is much lower.

Finally, Table 2 shows the wide reach of the EITC among this population. More than half of these households are estimated to be eligible with an average, estimated payment of $3120; the percent eligible is much higher among unmarried households and those earning low wages. The estimated increase in household income is sizable at 7.4% overall (=3120/42,050) and 11.2% for the most disadvantaged group (unmarried women with wages < $8). Using the state multipliers (reported in Appendix Table A1) yields a corresponding average payment of up to 40% of that amount, or $1258.

These descriptive statistics suggest that these labor market policies are highly relevant for households during the perinatal period. A large proportion of our study population work during the perinatal period and likely live in a household where someone works. They also work a substantial number of hours, such that an increase in their wage has a sizable impact on household income. Most are also estimated to receive a substantial EITC payment. If the ACS sample is more economically advantaged than the PRAMS or BRFSS, as suggested by differences in household income, the potential impacts could be even larger in our analysis samples. We return to these statistics when interpreting the policy's estimated effects on mental health in our empirical analyses.

3.2. Possible Mechanisms for Mental Health Effects

Past research reveals that exactly how these policies affect these households' labor supply, income and other outcomes related to mental health is far from clear. The predicted mental health effects of the EITC seem more straightforward, as its impacts on reducing poverty are fairly well‐established (e.g., Hoynes ²⁰¹⁹), and poverty is a known risk factor. By construction, the policy expands the time‐income options available to the household. Even an induced reduction in labor supply may have beneficial mental health benefits in that it frees up time during a period when the demands on time increase sharply. Using the Current Population Survey (Annual Social and Economic Supplement, CPS‐ASEC) from 1990 to 2016, Michelmore and Pilkauskas (2021), Table 3 provides strong, robust evidence that the EITC increases the labor supply and pretax earnings for unmarried mothers without a college degree (our analysis sample), especially among those whose youngest child is under age 3. For this group, their estimates suggest that a $1000 increase in EITC leads to an 8.9% increase in the probability of working last week, a 6.6% increase in working at least 35 h and a $2443 increase in pre‐tax earnings. However, the study also finds associated, increased child care use and costs such that the impact on disposable income is likely smaller.

The effects of the minimum wage are more theoretically ambiguous. If the policy does not affect the household's ability to work their desired hours, then similar to the EITC it may expand the time‐income options available. However, to the extent that it places constraints on how much the household can work—through reduced hours or in the most extreme case of unemployment—the effects are less clear and could be detrimental (Zuelke et al. 2018). As noted above, the overall effects of the minimum wage remain highly debated, but the most relevant study for our analysis group yields robust evidence of beneficial—or at least not harmful—effects. Godøy et al. (2021) investigates the changes in employment in low wage jobs by parental and marital status and the ages of the youngest child. Using the CPS‐ASEC from 1982 to 2019, 152 minimum wage changes and a distributional difference‐in‐differences approach (plus other methods), they find positive labor supply and earnings effects, especially for unmarried women with children ages 0 to 5. For this group, increasing the minimum wage increases the employment in jobs above the minimum wage by more than twice as much as it reduces employment in jobs below the minimum wage (Godøy et al. 2021, Table 2, 427). The estimated employment elasticity is 0.21 overall and is much higher, at 1.12, for those at affected wages. Wages are also estimated to increase 13%. These beneficial effects spill over to reduced poverty levels and welfare receipt (Godøy et al. 2021, Figure 6, 435). For married women and men with young children, the effects are more muted; net‐employment is mostly unchanged and wages modestly increase. Still, no detrimental effects are evident.

Similar to Michelmore and Pilkauskas (2021), Godøy et al. (2021) also finds that this increased employment is associated with higher child care costs, which may again lead to smaller increases in disposable income. In comparing their (larger) estimated labor supply and income effects to Michelmore and Pilkauskas (2021), Godøy et al. (2021) makes the point that child care concerns and other fixed employment costs suggest that minimum wages may have a stronger labor supply effect for this group than the EITC. The lump‐sum, lagged effect of the EITC, combined with imperfect capital markets, may mean that a higher EITC is less helpful in overcoming the fixed costs of work than the sustained, steady increase provided by a wage increase.

The EITC and minimum wage may combine for stronger effects, as discussed in Neumark and Wascher (2011) and others. For example, the possibility that the wage subsidy effect of the EITC gets passed on to employers may be limited by a minimum wage; likewise, a minimum‐wage induced reduction in hours may be ameliorated by an EITC. Recent evidence comes from Lenhart and Chakraborty (2025) which finds that minimum wages reduce food insecurity only in states with EITCs and that the effect is stronger in high EITC states. We explore a possible joint effect in our empirical analyses as well.

Finally, focusing on labor supply/time use and income effects as the two main avenues ignores other pathways, such as a change in health insurance status and thus a lower cost for health care or the positive impacts these policies may have on infant and child health. As noted in Section 2, substantial evidence exists that these policies have beneficial effects on birth outcomes, which in turn may have an effect on perinatal mental healtgod. Likewise, poverty‐related stress may be reduced too, while the results for health insurance are far more mixed. Our empirical investigation into mechanisms in Section 5.4 explores the effects on health insurance coverage, health care access, job and income‐related stressors, and infant health outcomes.

4. Data and Methodology

Our primary analyses use data from the restricted access Pregnancy Risk Assessment Monitoring System (PRAMS) from 2012 to 2018 births, or phases 7 and 8. ⁸ The PRAMS is a survey conducted by the Center for Disease Control typically sent out to people who have given birth within 2–5 months, with the goal of collecting data on conception, pregnancy, and birth related health outcomes. The PRAMS is the ideal dataset as it contains measures of perinatal depression not available in other datasets, has ample sample size and includes state identifiers needed to connect to labor policies. We limit our sample to those aged 18 to 45 and drop those with extreme values. ⁹ We focus on respondents with less than a 4‐year college education because they seem the most likely to be affected by these policies. In robustness checks, we estimate the effects for other groups, including a college‐educated group and a less‐than‐high school group that are likely less/more affected.

The PRAMS is conducted at the state level and some states/years are unavailable, either due to failing to meet the PRAMS’ response rate threshold or because the state did not conduct the survey. We use the largest possible sample, which includes 34 states and 185 state‐year combinations. Estimating our main models using a fully balanced sample with the 14 states that have all 7 years of data yields qualitatively similar results. The years and states of our PRAMS analyses are reported in the first two columns of Table 1 and are indicated by shading in Appendix Tables A1 and A2, which also report the policy changes.

Identifying and isolating the empirical effects of these policies requires substantial independent variation, which the remaining columns of Table 1 show is considerable. Of the 34 states represented in the PRAMS, the minimum wage increased in 19 (often multiple times in a state), the state EITC multiplier increased in 10, and 26 expanded Medicaid eligibility at some point between 2012 and 2018. Appendix Figure A1 further depicts these changes and shows the geographic variation. Still, the short time period and unbalanced nature of the data challenges our ability to isolate the effects of these different policies. Regressing these policy variables on the other policies and state controls, individual characteristics, and state and year fixed effects, yields an R‐squared of 0.832 for the minimum wage and 0.988 for the EITC, suggesting very limited independent variation in the EITC. We explore this issue in several exercises, including studying each policy in isolation and in the BRFSS analysis that explores the impact of limiting the number of state‐years to that of the PRAMS.

4.1. Outcome Variables and Controls

The PRAMS provides two measures of depression that differ in timing and construction. The first is the respondent's self‐reported depression in the 3 months prior to pregnancy. As reported in Table 3, the rate of pre‐pregnancy depression is 14.1% for those without a 4‐year college education.

TABLE 3.

Summary statistics of respondents without a 4‐year college education.

	Mean (std.)	Definition
PRAMS mental health measures
Pre‐pregnancy depression	0.141	Equal to one if respondent checked “depression” to “during the 3 months before you got pregnant with your new baby, did you have any of the following health conditions?” for Phase 8 or “before you got pregnant with your new baby, did a doctor, nurse, or any other health care worker tell you that you had any of the following health conditions” for Phase 7.
	(0.348)
Postpartum depression	0.140	Equal to one if respondent answered “always” or “often” to either of “since your new baby was born, how often have you felt down, depressed, or hopeless?” and “since your new baby was born, how often have you had little interest or little pleasure in doing things you usually enjoyed?
	(0.347)
Postpartum well‐being	0.599	Equal to one if respondent answered “never” or “rarely” for both of “since your new baby was born, how often have you felt down, depressed, or hopeless?” and “since your new baby was born, how often have you had little interest or little pleasure in doing things you usually enjoyed?”
	(0.490)
BRFSS mental health measure during pregnancy
Number of days with “not good” mental health	4.976	Persons response to “now thinking about your mental health, which includes stress, depression, and problems with emotions, for how many days during the past 30 days was your mental health not good?”
	(11.80)
10 or more “not good” mental health day	0.163	Equal to one if the number of days with “not good” mental health out of past 30 days is 10 or more.
	(0.370)
All 30 “not good” mental health day	0.059	Equal to one if number of days with “not good” mental health out of past 30 days is 30.
	(0.235)
PRAMS labor measures
Worked during pregnancy	0.626	Equal to one if respondent reported working at any point during the pregnancy.
	(0.484)
Income	32.926	Reported total pre‐tax household income in the 12 months before the baby was born (in brackets and thousands of dollars).
	(21.647)
PRAMS financial stressors
Problems paying bills	0.236	Equal to one if respondent answered “yes” to this happening in the 12 months before the baby was born: “You had problems paying the rent, mortgage, or other bills.”
	(0.425)
Lost job, hours, or pay	0.321	Equal to one if respondent answered “yes” to any of the following happening in the 12 months before the baby was born: “My husband or partner lost his job”, “I lost my job even though I wanted to go on working”, and “my husband, partner, or I had a cut in work hours or pay” as questions.
	(0.467)
PRAMS other mechanisms
Pregnancy planned	0.436	Equal to one if respondent reported the pregnancy was planned.
	(0.496)
No insurance before pregnancy	0.212	Equal to one if respondent reported having no health insurance in the month before the pregnancy.
	(0.409)
No insurance during pregnancy	0.042	Equal to one if respondent reported having no health insurance during the pregnancy.
	(0.200)
No insurance date of interview	0.159	Equal to one if respondent reported having no health insurance at the date of interview.
	(0.365)
Postpartum checkup	0.879	Equal to one if respondent reported having a postpartum checkup visit.
	(0.326)
Talked to HCW about PPD	0.760	Equal to one if respondent reported a health care worker talked to them about postpartum depression during a prenatal visit.
	(0.427)
Low birth weight	0.072	Equal to one if respondent's baby weighed 2500 g or fewer at birth.
	(0.259)
Preterm birth	0.108	Equal to one if respondent's gestation period was fewer than 260 days.
	(0.311)

Note: Estimates are obtained using sample weights and data from 2012–2018.

The second is a measure of PPD constructed from depressive‐related symptoms reported at the time of the interview. Specifically, PPD is captured with a binary measure based on the person's response to two depression‐related screening questions, called a PHQ‐2. These two questions are “Since your new baby was born, how often have you felt down, depressed, or hopeless?” and “Since your new baby was born, how often have you had little interest or little pleasure in doing things you usually enjoyed?”. Each contains five possible answers: always, often, sometimes, rarely, and never. The typical threshold to qualify for depression, one used by the PRAMS as well, is answering often or always to at least one of these questions. PPD prevalence rates for those without a 4‐year college degree is also around 14%. Following Margerison et al. (2021), we also create an indicator for those least likely to have depression to capture both ends of the mental health spectrum.

The wording of the PHQ‐2 questions is different in Phase 6 of the PRAMS (spanning 2009–11 births), which requires limiting our sample to phases 7 and 8, or 2012–2018 births. Our two primary outcome measures are thus (1) self‐reported depression in pre‐pregnancy, and (2) a PHQ‐2 constructed measure for PPD. Note that these two measures are not directly comparable, which likely explains why the prevalence of depression does not appear to increase in the postpartum period.

Figure 1 reports the prevalence rates of both measures for different subgroups. As expected, depression in both periods is highest for the least educated and unmarried. While non‐Hispanic White respondents report higher rates prior to pregnancy, the gap closes and is reversed for non‐Hispanic Black respondents for PPD. This pattern suggests that the onset of PPD may differ by race/ethnicity and so we also report results stratified by race and ethnicity. Figure 1 also reveals the very high correlation between depression pre‐pregnancy versus post‐partum. This finding highlights a potential pathway for these policies to have an effect on PPD, by affecting mental health prior to pregnancy.

FIGURE 1.

Prevalence rates for measures of pre‐pregnancy and postpartum depression by subgroup. All rates are weighted using sampling weights using the PRAMS with 2012–2018 births. All subgroups are for respondents without a 4‐year college education unless otherwise stated. Pre‐pregnancy depression is self‐reported while postpartum depression is via a PHQ‐2. See Table 3 for more details on measures.

To fill in the gap between the 3 months prior to pregnancy and the postpartum period, we conduct a supplementary analysis using the BRFSS. The BRFSS does not identify those who have recently given birth or ask questions about mental health before the pregnancy, but it does identify people who are currently pregnant. The BRFSS is also available for all 50 states and D.C., allowing us to explore the impact of including only the more limited state‐years available in the PRAMS. However, the BRFSS measure of mental health is substantially different from the PRAMS, which limits its comparability. At the same time, its common use in mental health research more broadly (e.g., Horn et al. ²⁰¹⁷; Gangopadhyaya et al. ²⁰²⁰) facilitates comparison with existing work. The BRFSS asks how many out of the past 30 days has the person had “not good” mental health from which we create three measures following past work: number of days reported (following Horn et al. ²⁰¹⁷) and then two binary measures with cut offs at those reporting at least 10 days and those reporting all 30 days, similar to Kuroki (2021). Table 3 shows the two binary measures produce estimates that are similar in magnitude to the percentages of Americans living with any mental illness and a serious mental illness, at approximately 20 and 5% respectively (National Alliance on Mental Health ²⁰²³).

After investigating the effects of these policies on mental health around pregnancy, we then explore possible mechanisms available in the PRAMS, which includes income, financial stressors, health insurance and care received and birth outcomes. These are summarized at the bottom of Table 3 and, unfortunately, are often even more limited in their availability than our main outcomes.

Appendix Table A3 provides summary statistics and data sources for the individual and state‐level characteristics included in all models that are likely related to depression. Individual characteristics in the PRAMS include respondent's age, marriage status, race and ethnicity, number of people dependent on household income, months between birth and interview, and education attainment. The BRFSS contains these same, or very similar, measures. Controls for the overall state environment include the annual state unemployment rate, political party of the state governor, state GDP growth rate, and per capita mental health care supply from County Business Patterns data.

4.2. Empirical Strategy

We estimate, via OLS, ¹⁰ a generalized difference‐in‐differences model,

$$Y_{ist}={\beta }_1{MinWage}_{st}+{\beta }_2{EITCMax}_{ist}+{\beta }_3{TANF}_{st}+{{\beta }_{4}MedExp}_{st}+\psi X_{ist}+\theta A_{st}+{\delta }_s+{\tau }_t+{\epsilon }_{ist}$$ (1)

where $Y_{ist}$ is the outcome considered in each model for individual i in state s in year t ${MinWage}_{st}$ is the minimum wage in dollars in the state ¹¹ and ${EITCMax}_{ist}$ is the maximum federal + state EITC amount receivable for their family size, ¹² in thousands of dollars. As reported shortly, we also explore alternative forms of these variables, such as using logged or lagged values, the state EITC multiplier and the minimum wage relative to the state‐year median wage. We also explore possible joint effects of the minimum wage and the EITC, as well as the roles of refundability or seasonality, by adding interaction terms to Equation (1). ${TANF}_{st}$ is the maximum amount a family of three is eligible for in state s and year t ${MedExp}_{st}$ is a dummy variable for if state s expanded Medicaid under the ACA at least 9 months before the respondent gave birth. $X_{ist}$ is the vector of individual characteristics such as age, education, etc. as well as month of conception or interview fixed effects (in line with the outcome) and $A_{st}$ is the vector of state controls, which include the annual unemployment rate, GSP growth, political party of the governor and the supply of mental health care providers. Standard errors are clustered at the state level (Bertrand et al. 2004).

Our main analysis is limited to those without a 4‐year college education, defined as less than 16 years of schooling, as in Michelmore and Pilkauskas (2021), to focus on the group most likely impacted by these anti‐poverty policies. We conduct several event study analyses to verify that these models satisfy the pre‐trend assumption and also to investigate the dynamic effects—that is, whether the effects grow, diminish or are roughly constant as time passes. Besides having multiple policies to investigate, these analyses must also be adapted because minimum wages and EITC (1) are continuous, and (2) change multiple times in some states during our sample period. To address these complications, we follow Schmidheiny and Siegloch (2023) and conduct analyses for each policy separately (controlling for the others). ¹³

5. Results

We investigate each depression measure (pre‐pregnancy, PPD and the BRFSS during pregnancy) in a separate section that includes various robustness checks, sample stratifications and event studies. We conclude with a discussion of possible mechanisms.

5.1. Pre‐Pregnancy Depression

Table 4 reports pre‐pregnancy depression results, using Equation (1) and first including each policy separately and then including all at once. Our results suggest minimum wages and EITC levels are each associated with a reduction in depression before the pregnancy whether or not the other policies are controlled for. In contrast, neither TANF nor Medicaid is found to have an effect. Medicaid expansions approach statistical significance when no other policies are considered, echoing the findings of Margerison et al. (2021), but those effects are eliminated in the full model. ¹⁴

TABLE 4.

Marginal effects of safety‐net policies on pre‐pregnancy depression.

	Min wage only	EITC only	TANF only	Medicaid only	All
Min wage	−0.0119**				−0.0110**
	(0.0053)				(0.0048)
EITC max		−0.0221***			−0.0206***
		(0.0072)			(0.0074)
TANF			−0.0260		−0.0087
			(0.0972)		(0.0918)
Medicaid				−0.0065	0.0004
				(0.0060)	(0.0059)
N	105,672	105,672	105,672	105,672	105,672
Sample mean	0.141	0.141	0.141	0.141	0.141

Note: All models control for state and year fixed effects, month of conception fixed effects, respondent's race, ethnicity, age, education, marital status, parity, time between birth and interview, survey phase, state mental health care supply, state unemployment rate, state GDP growth rate, and political party of state governor. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. Policy variables are for year of conception. EITC Max and TANF are in thousands of dollars. All models are for observations without a 4‐year college education using PRAMS data for 2012–2018 births representing 185 state‐year combinations.

^*** p < 0.01.

^** p < 0.05.

*p < 0.1.

The estimated effects from our main model, column 5 in Table 4, suggest a one dollar increase in the minimum wage—about the average increase during our sample—is associated with a 1.1% point decrease in the likelihood of reporting being depressed in the months before the pregnancy. With a sample average prevalence of 14%, this reduction represents an 8.5% decrease in pre‐pregnancy depression for our main sample. A thousand dollar increase in the annual maximum EITC amount received is associated with a roughly 2.06% point decrease. Table 2 helps put these numbers in context by showing that more than half of women and over 90% of men with children under age 1 reported working in the last year. For the almost 50% of women currently working, their median annual hours worked is 1545 with an associated hourly wage of less than $10 (=$15,000/1545); their median household income is $42,050. While this median is above the minimum wage, it may be close enough to be affected (e.g., Cengiz et al. ²⁰¹⁹). Even if not, Table 2 suggests that more than a quarter of currently working women with a very young child earn within $1 of their state's minimum wage and are working a median 1320 h. It also seems likely that the fraction working at those wages (and the hours worked) is even greater in the 3 months prior to pregnancy. Thus, a $1 increase in the hourly wage could have a sizable impact on household income, even without considering the impacts on other household members.

Table 2 also shows that 57.9% of these households are estimated to be eligible for the EITC, who would receive $3120 in EITC payments on average. However, state EITC's tend to be much smaller. The average state multiplier is 0.13 for the PRAMS state‐years and many are not refundable. A $1000 increase in the EITC, due to state policy alone would therefore be equivalent to enacting a refundable EITC multiplier of 0.285, well above what is in our data. Using the average change in our data of 0.03 (Table 1) suggests an increase in EITC payments of only about $100 and predicted decline in pre‐pregnancy depression of 0.206% points (or about 1.5%), a sensible yet nontrivial impact.

To verify the validity of these results, Figures 2a and 2b report the event study analyses for the minimum wage and EITC, respectively, following Schmidheiny and Siegloch (2023); see footnote 13 for details. For both policies, the pre‐treatment estimates are centered around and not statistically different from zero, suggesting that the parallel pre‐trends assumption is not violated. ¹⁵ Besides satisfying the pre‐trends assumption, these figures also reveal that the estimated policy effects are fairly stable in the years following the change. As is now well‐known, difference‐in‐differences models can suffer from bias due to comparing treated to previously treated units (e.g., Goodman‐Bacon ²⁰²¹). While several methods exist to adjust for it, the unbalanced nature of our data combined with having two continuous policies with multiple changes makes these adjustments particularly challenging to implement. To explore possible bias, we conduct several exercises. Specifically, we re‐estimate the model (1) studying the impact of “big” changes (>$1) in the minimum wage only, (2) limiting the sample to states with only one “big” change, (3) removing the unemployment rate, as including it may be shutting down a possible avenue of effect, (4) removing all state controls, to explore the extent of confounding factors, and (5) limiting the sample to states with PRAMS data in every year for a balanced panel. Each exercise yields reasonably similar results to our main model and are reported in Appendix Table A4. These findings, plus the relative stability of the post‐treatment coefficients (Figure 2), provide reassurance.

FIGURE 2.

(a) Event study for changes in minimum wages on pre‐pregnancy depression for main sample using PRAMS data for 2012–2018 births.(b) Event study for changes in EITC max on pre‐pregnancy depression for main sample using PRAMS data for 2012–2018 births. Event study models follow Schmidheiny and Siegloch ²⁰²³ for minimum wages while controlling for EITC Max levels in 2a (then for EITC Max while controlling for minimum wages in 2b), state and year fixed effects, month of conception fixed effects, survey phase, respondent's race, ethnicity, age, education, marital status, parity, time between birth and interview, state mental health care supply, state unemployment rate, state GDP growth rate, state TANF levels for a family of 3, Medicaid expansion, and political party of state governor. The gray dashed lines are the 95% confidence intervals. All observations are weighed using sample weights and standard errors are clustered at the state level using data from the PRAMS with 2012–2018 births. Policy variables are for year of conception.

Table 5 reports the results from splitting the sample into different subgroups and exploring alternative policy measures. We report the results for only the minimum wage and EITC coefficients because the Medicaid and TANF coefficients continue to be statistically insignificant. The first column in the top panel repeats the results for our main model for comparison. The rest of the first panel and the second report the estimated effects when the sample is split into different groups. As expected, those with less than a high school education have substantially larger effects, while those with a college education show no effect for minimum wages and a counter‐intuitive positive effect for EITC. ¹⁶ Married households have at least two potential earners and thus may be more strongly affected by the minimum wage than unmarried households. The results offer some support, especially given the lower incidence of pre‐pregnancy depression among married women which suggests an even larger proportional effect. The reverse appears to be true for the EITC, as the point estimates and associated proportional effects are estimated to be larger for unmarried households. This difference also makes sense given the lower household income and likely larger proportionate effect of the EITC on these households, as shown in Table 2.

TABLE 5.

Alternative samples and specifications for pre‐pregnancy depression.

	Main sample	Less than HS	College plus	Married	Unmarried
Min wage	−0.0110**	−0.0220**	−0.0009	−0.0112*	−0.0083
	(0.0048)	(0.0085)	(0.0034)	(0.0058)	(0.0071)
EITC max	−0.0206***	−0.0374	0.0109*	−0.0141*	−0.0340***
	(0.0074)	(0.0235)	(0.0055)	(0.0078)	(0.0100)
N	105,672	18,901	54,713	50,406	55,266
Sample mean	0.141	0.153	0.071	0.114	0.168

	Previous birth	No previous birth	Non‐Hisp white	Non‐Hisp black	Hispanic
Min wage	−0.0111*	−0.0114	−0.0122	−0.0033	−0.0077
	(0.0056)	(0.0077)	(0.0088)	(0.0110)	(0.0087)
EITC max	−0.0208*	−0.0160	−0.0240*	−0.0143*	−0.0220**
	(0.0117)	(0.0126)	(0.0120)	(0.0081)	(0.0100)
N	68,498	37,174	48,394	21,672	19,919
Sample mean	0.136	0.143	0.163	0.105	0.081

	18–25 year‐olds	Lagged measures	Logged measures	Relative min wage	EITC multiplier
Min wage	−0.0191*	−0.0102**	−0.0884**	−0.1431*	−0.0106**
	(0.0096)	(0.0044)	(0.0408)	(0.0802)	(0.0045)
EITC max	−0.0233*	−0.0209**	−0.1704**	−0.0212***	−0.1526***
	(0.0119)	(0.0084)	(0.0645)	(0.0073)	(0.0467)
N	35,540	105,672	105,672	105,672	105,672
Sample mean	0.154	0.141	0.141	0.141	0.141

Note: All models control for state and year fixed effects, month of conception fixed effects, respondent's race, ethnicity, age, education, marital status, parity, time between birth and interview, survey phase, state mental health care supply, state unemployment rate, state GDP growth rate, state TANF levels for a family of 3, state Medicaid expansion, and political party of state governor. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. Policy variables are for year of conception. EITC Max and TANF are in thousands of dollars. All models are for observations without a 4‐year college education, unless otherwise noted, using PRAMS data for 2012–2018 births 185 state‐year combinations.

^*** p < 0.01.

^** p < 0.05.

^* p < 0.1.

The second panel explores stratifying the sample by parity and race/ethnicity. EITC payments are much larger for households with a child, which our specification takes account of by including the maximum amount for that specific household. It nonetheless is reassuring to confirm that first time parents are not driving our results, as their pre‐birth EITC payments are substantially lower than those with previous children. Overall, the estimated magnitude of the effects are similar by parity, although splitting the sample reduces the statistical significance, as expected. Splitting by race similarly reduces the statistical power, especially for the minimum wage where none are statistically significant. The estimated EITC effects are reasonably similar across groups, although appear strongest for Hispanic respondents. The lower incidence of pre‐pregnancy depression among Hispanic respondents suggests that the EITC may have an especially powerful effect for this group.

The last panel in Table 5 explores alternative samples and policy measures frequently used in past work. We first limit the sample to respondents ages 18 to 25, as previous work argues it is the age group most likely to be impacted by the minimum wage (Allegretto and Nadler 2020). Our findings for both policies are similar for this age group and the magnitudes are larger, if anything. Using lagged and then logged measures of both minimum wages and EITC levels yield similar results. Wage levels vary a great deal across states, suggesting that the same minimum wage may be more binding in some states than others. We therefore redefine the minimum wage to be relative to the state's median wage in that year ¹⁷ to capture these differences. This exercise yields a negative and statistically significant effect with a similar but larger magnitude as the proportionate wage effect (logged). ¹⁸

EITC studies often use the state EITC multiplier (fraction) instead of the dollar amount (which adjusts for household's size and includes the federal payment). In the last entry of Table 5 we replace the federal + state maximum EITC payment with the state EITC multiplier as a fraction of the federal EITC. The beneficial mental health effects of the EITC remain and the magnitude is about double that of the main model; the average increase (0.03) leads to a decrease of 0.03 × 0.1526 = 0.46% points. Appendix Table A5 summarizes the results of our investigation into several other nuances of the EITC, as discussed in Section 2. Separating the EITC into its federal and state components yields an even larger and more statistically significant effect of the state EITC (and a null effect for the federal one), further verifying that our main results are driven by the variation in state policy. We then add interactions of the EITC with (1) an indicator for being refundable, (2) an indicator for a pre‐conception period in the spring (when EITC payments are received), and (3) the minimum wage, to explore possible joint effects. None of these interaction terms are statistically significant, although the minimum wage interaction is negative (as expected) and the three coefficients are jointly statistically significant, lending some weak support for a joint effect. The limited number of state‐years available in the PRAMS may not be rich enough to identify precisely these nuances.

5.2. Postpartum Depression (PPD)

We follow a similar approach and use the same model, with variables adjusted for the different timing of the outcome, in our empirical investigation of PPD. Given the much higher prevalence of PPD for those with pre‐pregnancy depression, our main estimates may capture the continued effects on pre‐pregnancy depression, a possibility we explore shortly.

Table 6 follows Table 4's format of considering each policy separately and then when all are included. The estimates for minimum wages and EITC are again negative but about one‐third in size compared to pre‐pregnancy and neither is statistically significant. The diminished effect makes intuitive sense as PPD is more associated with biological changes that may be less influenced by income. In addition, most interviews take place 2–5 months after birth, a time when household labor force participation is likely lower, especially for unmarried women. The event studies once again yield no evidence that the parallel pre‐trends assumption is violated, but also yield little evidence of an effect. A counterintuitive result is the positive and marginally statistically significant estimate for TANF when all policies are included. While puzzling, removing TANF as a control does not change our main results for minimum wages and EITC levels. Additionally, modeling TANF as the maximum amount a family can receive given their number of dependents, as we do for EITC, yields a null effect.

TABLE 6.

Marginal effects of safety‐net policies on postpartum depression.

	Min wage only	EITC only	TANF only	Medicaid only	All
Min wage	−0.0034				−0.0027
	(0.0037)				(0.0034)
EITC max		−0.0072			−0.0064
		(0.0072)			(0.0071)
TANF			0.0926		0.1280*
			(0.0737)		(0.0647)
Medicaid				−0.0083	−0.0075
				(0.0057)	(0.0056)
N	107,228	107,228	107,228	107,228	107,228
Sample mean	0.140	0.140	0.140	0.140	0.140

Note: All models control for state and year fixed effects, month of interview fixed effects, respondent's race, ethnicity, age, education, time between birth and interview, parity, state mental health care supply, state unemployment rate, time between birth and interview, state GDP growth rate, and political party of state governor. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. Policy variables are for year of birth. EITC Max and TANF are in thousands of dollars. All models are for observations without a 4‐year college education, unless otherwise noted, using PRAMS data for 2012–2018 births 185 state‐year combinations.

***p < 0.01.

**p < 0.05.

^* p < 0.1.

Table 7 reports PPD results for the same variations as Table 5 did for pre‐pregnancy depression, plus some new ones unique to PPD in the bottom panel. For PPD, we see much stronger differences across demographic groups. For less than high school, the effect of minimum wages is even a bit larger and more statistically significant than it was for this group before pregnancy; the EITC is still statistically insignificant but the magnitude is also more similar to before pregnancy.

TABLE 7.

Alternative samples and specifications for postpartum depression.

	Main sample	Less than HS	College plus	Married	Unmarried
Min wage	−0.0027	−0.0227***	−0.0018	−0.0104***	0.0044
	(0.0034)	(0.0068)	(0.0019)	(0.0028)	(0.0056)
EITC max	−0.0064	−0.0229	−0.0038	0.0059	−0.0218**
	(0.0071)	(0.0190)	(0.0077)	(0.0094)	(0.0092)
N	107,228	19,295	55,239	51,168	56,060
Sample mean	0.140	0.156	0.075	0.093	0.162

	Previous birth	No previous birth	Non‐Hisp white	Non‐Hisp black	Hispanic
Min wage	−0.0090***	0.0081	−0.0064	0.0071	0.0013
	(0.0032)	(0.0071)	(0.0039)	(0.0122)	(0.0040)
EITC max	−0.0104	−0.0041	−0.0166	−0.0188*	0.0128
	(0.0145)	(0.0088)	(0.0154)	(0.0097)	(0.0090)
N	69,517	37,711	48,929	21,987	20,300
Sample mean	0.137	0.142	0.137	0.172	0.109

	18–25 year‐olds	Lagged measures	Logged measures	Relative min wage	EITC multiplier
Min wage	−0.0029	−0.0007	−0.0183	−0.0760	−0.0031
	(0.0076)	(0.0032)	(0.0310)	(0.0645)	(0.0036)
EITC max	−0.0047	−0.0010	0.0537	−0.0065	0.0935
	(0.0099)	(0.0082)	(0.0841)	(0.0070)	(0.0685)
N	39,481	107,228	107,228	107,228	107,228
Sample mean	0.168	0.140	0.140	0.140	0.140

	Controlling for DB	Did have DB	Did not have DB		Postpartum well‐being
Min wage	−0.0010	0.0141	−0.0027		0.0100**
	(0.0036)	(0.0126)	(0.0034)		(0.0044)
EITC max	−0.0033	−0.0348	0.0009		−0.0020
	(0.0069)	(0.0391)	(0.0055)		(0.0103)
N	105,672	15,622	90,050		107,228
Sample mean	0.140	0.304	0.113		0.599

Note: All models control for state and year fixed effects, month of interview fixed effects, respondent's race, ethnicity, age, education, time between birth and interview, parity, state mental health care supply, state unemployment rate, time between birth and interview, state GDP growth rate, state TANF levels for a family of 3, state Medicaid expansion, and political party of state governor. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. Policy variables are for year of birth. EITC Max is in thousands of dollars All models are for observations without a 4‐year college education, unless otherwise noted, using PRAMS data for 2012–2018 births 185 state‐year combinations.

^*** p < 0.01.

^** p < 0.05.

^* p < 0.1.

The difference between the responses of married (stronger for minimum wages) and unmarried respondents (stronger for EITC) uncovered in Table 5 is much more extreme here, so much so that pooling the two together seems to mask any effects. We therefore re‐estimate all of the specifications in Table 7 also stratified by marital status (Appendix Tables A6 and A7). For minimum wages and married respondents, the effect is consistently negative, often statistically significant and displays the same pattern across the specifications as it did for the entire sample in Table 5 (before pregnancy). This effect is also mostly robust to the exercises in Appendix Tables A4 and A5. The effect of the EITC on unmarried respondents, shown in Appendix Table A7, is less robust and is especially sensitive to using the EITC multiplier, where the effect is eliminated. ¹⁹

The bottom panel of Table 7 explores the role of pre‐pregnancy depression, alternately controlling for it and stratifying by it. Pre‐pregnancy depression is likely endogenous and so these models should be viewed with caution. (Recall that PPD is measured differently, which precludes modeling the change in depression status.) Including pre‐pregnancy depression as a control has little impact, suggesting that reducing pre‐pregnancy depression is not the primary mechanism. Splitting the sample into those who do versus do not report pre‐pregnancy depression lends support to this interpretation, as the minimum wage results, including those for married women, are driven by those without prior depression. However, the much smaller sample size for those reporting depression, along with the aforementioned endogeneity, cautions against drawing a conclusion.

Lastly, we consider the opposite end of the PPD spectrum, following Margerison et al. (2021) to create an indicator for those that answer “never” or “rarely” to both questions of the PHQ‐2. Reported in the last column, higher minimum wages lead to improved mental well‐being in the postpartum period. EITC has no effect, not even for unmarried women. Taken together, these results suggest that the beneficial effect of the minimum wage found for pre‐pregnancy carry over in a more limited way in the postpartum period. The results for the EITC are even more diminished, although may still be modestly beneficial for unmarried women.

5.3. Supplemental Analysis of Mental Health During Pregnancy

Our main analysis investigates depression before pregnancy and after birth using the PRAMS. To fill the gap in between—mental health during pregnancy—we use 2012–18 BRFSS mental health data from all available states for respondents reporting they are pregnant at the time of the survey. We attempt to match our PRAMS models as closely as possible, including the same state policy and control variables, the same individual characteristics and year, state and month fixed effects. Notably, however, the BRFSS mental health measures are quite different, being based on the number of days that mental health is reported as “not good”.

Results using these BRFSS measures are summarized in Table 8 for three possible samples: (1) one that includes all states (except California) and years (50 × 7) ²⁰ , (2) one that includes all years for the 34 states used in our main PRAMS analyses, and (3) the exact state‐year observations used in our PRAMS analyses. These samples reveal the possible effects of having to use the more limited state‐years available in the PRAMS, in addition to the impacts to mental health during pregnancy. As Table 1 and Appendix 1 show, the EITC experienced far fewer changes during this time period than the minimum wage, and several are not captured by the PRAMS due to missing years, and so we may expect its results to be especially sensitive.

TABLE 8.

Marginal effects of safety‐net policies on BRFSS “not good” mental health days during pregnancy.

	All states & years			PRAMS states & all years			PRAMS states & PRAMS years
	# of days	10+ days	All 30 Days	# of days	10+ day	All 30 Days	# of days	10+ days	All 30 Days
Min wage	0.1069	−0.0036	−0.016	0.2030	−0.0058	−0.0010	0.4381	−0.0176	−0.0069
	(0.3148)	(0.0121)	(0.0080)	(0.3707)	(0.0182)	(0.0096)	(0.5372)	(0.0201)	(0.0101)
EITC max	−1.2644***	−0.0497***	−0.0195**	−1.511*	−0.0359	−0.0118	−1.4721	−0.0401	−0.0129
	(0.4438)	(0.0168)	(0.0093)	(0.7622)	(0.0357)	(0.0156)	(0.9086)	(0.0429)	(0.0159)
N	10,153	10,153	10,153	6188	6188	6188	5011	5011	5011
# of states	50	50	50	34	34	34	34	34	34
Sample mean	4.976	0.164	0.059	5.377	0.186	0.066	5.316	0.185	0.064

Note: All models control for state and year fixed effects, month of interview fixed effects, respondent's race, ethnicity, age, education, parity, state mental health care supply, state unemployment rate, state GDP growth rate, state TANF levels for a family of 3, state Medicaid expansion, and political party of state governor. California observations are not included due to the extreme value of its state EITC multiplier (0.85); results are qualitatively similar with its inclusion. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. Policy variables are for year of interview. EITC Max in is thousands of dollars. All models are for pregnant observations without a 4‐year college education, using BRFSS data for 2012–2018 data.

^*** p < 0.01.

^** p < 0.05.

^* p < 0.1.

That is in fact what we find; the EITC has a negative and strongly statistically significant effect on all three mental health measures when all states and years are included, showcasing possible mental health benefits during pregnancy. ²¹ As the states and years become more limited (the last 6 columns), the results are steadily diminished. These findings suggest that the somewhat weaker results found for the EITC before pregnancy and, especially, post‐partum may be due to a lack of policy variation. ²² In contrast, the minimum wage does not have a statistically significant effect in any specification. Given the much richer variation in minimum wage policy, the diminished effect of minimum wages here seems likely due to the different time period (during pregnancy) or the different mental health measure used.

5.4. Mechanisms

Lastly, we explore possible mechanisms through which the minimum wage and EITC can improve mental health during the time around pregnancy. However, we caution that the PRAMS is even more limited in its availability for these types of variables, summarized at the bottom of each panel of Table 9, especially those not related to health outcomes. In addition to the limited information on income and work behavior reported in Table 2, ²³ the PRAMS also contains some information on financial stressors (as used in Andrea et al. ²⁰²⁰). The first is if the respondent “had problems paying the rent, mortgage, or other bills” in the 12 months leading up to birth. The second asks if the respondent or their partner lost their job or had a cut in hours or pay in the same time period. However, nine of the 34 states are missing these two variables for 2016–18, a period that saw many policy changes. Moreover, as discussed in section 3b, the expected impact on some of these variables is not clear. For instance, households may choose to work less in response (and incur fewer work‐related costs) while leaving their disposable income mostly unchanged. For minimum wages a concern is that it could lead to reduced hours or layoffs; hence, a null effect is a reassuring outcome. Health care variables most directly related to maternal depression are also sparse. Two questions about mental health care are only available in five states and only between 2012 and 2016, which we deem insufficient to study. ²⁴ Whether the respondent spoke with a health care worker about PPD is available for more state‐years but is still missing for about one‐third. These limitations are magnified when we attempt to limit by education, marital status or parity. The bottom line is that the PRAMS is not well‐suited for this purpose.

TABLE 9.

Marginal effect of safety‐net policies on possible mechanisms.

	Worked during preg	Income (in $1000's)	Problem with bills	Lost job/pay/hours
1. Labor and finances
Min wage	−0.0033	0.2352	−0.0030	−0.0145*
	(0.0129)	(0.2780)	(0.0049)	(0.0081)
EITC max	0.0294	−0.2795	−0.0016	−0.0124
	(0.0184)	(0.5693)	(0.0145)	(0.0112)
N	36,940	100,124	95,624	93,397
State‐year obs	59	185	162	162
Sample mean	0.626	32.92	0.236	0.321

	Planned	No insurance before	No insurance during	No insurance now
2. Intention and insurance
Min wage	0.0054	0.0104	−0.0078**	−0.0054
	(0.0072)	(0.0094)	(0.0034)	(0.0101)
EITC max	−0.0268**	0.0430***	0.0003	0.0142
	(0.0131)	(0.0127)	(0.0115)	(0.0114)
N	98,016	107,122	93,806	107,063
State‐year obs	168	185	185	185
Sample mean	0.436	0.212	0.042	0.159

	Postpartum checkup	HCW Talked PPD	Low birth weight	Preterm birth
3. Care and outcomes
Min wage	0.0007	−0.0029	−0.0008	0.0011
	(0.0030)	(0.0123)	(0.0016)	(0.0034)
EITC max	0.0065	0.0085	−0.0067*	−0.0014
	(0.0072)	(0.0107)	(0.0037)	(0.0053)
N	106,839	71,160	107,228	107,228
State‐year obs	185	119	185	185
Sample mean	0.879	0.760	0.072	0.108

Note: All models control for state and year fixed effects, month of interview fixed effects, respondent's race, ethnicity, age, education, time between birth and interview, parity, state mental health care supply, state unemployment rate, time between birth and interview, state GDP growth rate, state TANF levels for a family of 3, state Medicaid expansion, and political party of state governor. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. EITC Max and is in thousands of dollars. All models are for observations without a 4‐year college education, using PRAMS data for 2012–2018 births.

^*** p < 0.01.

^** p < 0.05.

^* p < 0.1.

With these caveats in mind, Table 9 reports estimated effects for several different possible pathways and the number of state‐years available for each measure. The first panel reports the work, income and financial stressors. As expected given the paucity of data, the work and financial outcomes are mostly statistically insignificant although at least often the expected sign. The EITC has a positive and almost statistically significant effect on working during pregnancy, mirroring the findings of Michelmore and Pilkauskas (2021). The effects on financial or labor market distress are all negative but mostly statistically insignificant. The one statistically significant result, the negative effect of minimum wages on lost pay or hours, alleviates the concern that the minimum wage could induce layoffs or reductions in hours, in line with Godøy et al. (2021).

The rest of the table explores other possible outcomes including insurance status, health care received, pregnancy intention and birth outcomes, again with few statistically significant effects. The mixed findings for health insurance align with past work (Hoynes 2019; Leigh ²⁰²¹), as does the negative (beneficial) effect of the EITC on low birth weight (e.g., Markowitz et al. ²⁰¹⁷). Overall, the analysis of possible mechanisms in the PRAMS is suggestive at best.

6. Discussion

This research offers new evidence on the effects of a range of labor market and safety‐net policies on mental health during a particularly impactful time—before, during and after pregnancy. While a growing body of research suggests that minimum wages and the EITC have beneficial mental health effects on the broader population, our study is the first to our knowledge to focus on the period around pregnancy. We also consider more commonly studied policies targeting this group, namely the ACA Medicaid expansions and TANF.

Our primary analyses using the PRAMS provide robust evidence that both state minimum wages and EITCs substantially reduce depression before pregnancy. Our baseline estimates for those without a college education imply that a one dollar increase in the minimum wage led to an 8.5% reduction and that a $100 increase in the state EITC leads to a 1.5% reduction. These estimates are substantial but within the general range suggested by past research for different interventions and samples (Schmidt et al. 2023; Horn et al. ²⁰¹⁷). Some segments of the population experience even larger effects, such as less educated groups and Hispanic respondents. Our ability to study mechanisms is quite limited by what is available in the PRAMS and, likely as a consequence, the outcomes we can study yield weak results that are suggestive at best. We also provide a descriptive analysis using labor supply and income data on similar households in the ACS to demonstrate the wide reach and potentially large impacts of these policies, but we do not attempt to estimate any behavioral effects. Careful analyses of the possible mechanisms using better‐suited data sources is a worthwhile extension to this research.

The evidence for the effects during pregnancy and in the early postpartum period are less clear. Estimated effects on postpartum mental health continue to suggest a beneficial effect and show similar patterns to pre‐pregnancy depression, but are smaller, often statistically insignificant and less robust, especially for the EITC. The limited states and years available in the PRAMS have a much larger impact on the residual variation of the EITC than the minimum wage, which could be responsible. To explore this data limitation as well as the effects on mental health during pregnancy, we conduct supplementary analyses using the BRFSS. Those results suggest that the EITC improves mental health during pregnancy and that the ability to empirically identify its effects is impeded by the limited number of state‐years in the PRAMS. The minimum wage appears to have no effect, which could be due to either the timing (during pregnancy) or the different measure of mental health used in the BRFSS.

Our study therefore offers robust evidence that labor market policies improve mental health in the months before pregnancy and may be a promising approach to improving mental health during and after pregnancy as well, but several questions and limitations remain. While the PRAMS is arguably the best data for studying perinatal mental health, the survey is limited in its ability to study first stage effects. Moreover, the short time period that mental health is consistently measured limits the policy variation observed, especially for EITC. Future research could address both limitations by using richer data that is better suited to studying a wide range of outcomes, including labor market decisions and the overall experience of these vulnerable households. Why these policies appear to affect different racial/ethnic groups and through what mechanisms is also not clear. Finally, the Covid‐19 pandemic has changed the mental health landscape and the labor market environment, which suggests research in this area should be updated and extended as post‐Covid data become available.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

FIGURE A1.

Changes in state minimum wages and EITC multiplier, 2012–2018.

TABLE A1.

State EITC multiplier levels.

	2012	2013	2014	2015	2016	2017	2018
AL	0.000	0.000	0.000	0.000	0.000	0.000	0.000
AK	0.000	0.000	0.000	0.000	0.000	0.000	0.000
AZ	0.000	0.000	0.000	0.000	0.000	0.000	0.000
AR	0.000	0.000	0.000	0.000	0.000	0.000	0.000
CA	0.000	0.000	0.000	0.000	0.850	0.850	0.850
CO	0.000	0.000	0.100	0.100	0.100	0.100	0.100
CT	0.300	0.250	0.275	0.275	0.275	0.230	0.230
DE	0.200	0.200	0.200	0.200	0.200	0.200	0.200
DC	0.400	0.400	0.400	0.400	0.400	0.400	0.400
FL	0.000	0.000	0.000	0.000	0.000	0.000	0.000
GA	0.000	0.000	0.000	0.000	0.000	0.000	0.000
HI	0.000	0.000	0.000	0.000	0.000	0.000	0.200
ID	0.000	0.000	0.000	0.000	0.000	0.000	0.000
IL	0.050	0.050	0.100	0.100	0.100	0.100	0.180
IN	0.060	0.060	0.090	0.090	0.090	0.090	0.090
IA	0.070	0.070	0.140	0.140	0.150	0.150	0.150
KS	0.180	0.180	0.170	0.170	0.170	0.170	0.170
KY	0.000	0.000	0.000	0.000	0.000	0.000	0.000
LA	0.035	0.035	0.035	0.035	0.035	0.035	0.035
ME	0.050	0.050	0.050	0.050	0.050	0.050	0.050
MD	0.250	0.250	0.250	0.250	0.260	0.270	0.280
MA	0.150	0.150	0.150	0.150	0.230	0.230	0.230
MI	0.060	0.060	0.060	0.060	0.060	0.060	0.060
MN	0.330	0.330	0.330	0.330	0.330	0.350	0.350
MS	0.000	0.000	0.000	0.000	0.000	0.000	0.000
MO	0.000	0.000	0.000	0.000	0.000	0.000	0.000
MT	0.000	0.000	0.000	0.000	0.000	0.000	0.000
NE	0.100	0.100	0.100	0.100	0.100	0.100	0.100
NV	0.000	0.000	0.000	0.000	0.000	0.000	0.000
NH	0.000	0.000	0.000	0.000	0.000	0.000	0.000
NJ	0.200	0.200	0.200	0.300	0.350	0.350	0.370
NM	0.100	0.100	0.100	0.100	0.100	0.100	0.100
NY	0.300	0.300	0.300	0.300	0.300	0.300	0.300
NC	0.050	0.045	0.000	0.000	0.000	0.000	0.000
ND	0.000	0.000	0.000	0.000	0.000	0.000	0.000
OH	0.000	0.000	0.050	0.050	0.100	0.100	0.100
OK	0.050	0.050	0.050	0.050	0.050	0.050	0.050
OR	0.060	0.060	0.080	0.060	0.080	0.080	0.080
PA	0.000	0.000	0.000	0.000	0.000	0.000	0.000
RI	0.150	0.150	0.150	0.100	0.125	0.150	0.150
SC	0.000	0.000	0.000	0.000	0.000	0.000	0.208
SD	0.000	0.000	0.000	0.000	0.000	0.000	0.000
TN	0.000	0.000	0.000	0.000	0.000	0.000	0.000
TX	0.000	0.000	0.000	0.000	0.000	0.000	0.000
UT	0.000	0.000	0.000	0.000	0.000	0.000	0.000
VT	0.320	0.320	0.320	0.320	0.320	0.320	0.360
VA	0.200	0.200	0.200	0.200	0.200	0.200	0.200
WA	0.000	0.000	0.000	0.000	0.000	0.000	0.000
WV	0.000	0.000	0.000	0.000	0.000	0.000	0.000
WI	0.040	0.040	0.040	0.040	0.040	0.040	0.040
WY	0.000	0.000	0.000	0.000	0.000	0.000	0.000

Note: Data comes from the Center for Poverty Research. State‐year combinations that are provided by the PRAMS are shaded gray. Years with a change in the state EITC level are bolded. Please note, models for pre‐pregnancy depression utilize state ETIC levels in the year of conception while models for PPD utilize state ETIC levels during the year of interview.

TABLE A2.

State minimum wage levels.

State	2012	2013	2014	2015	2016	2017	2018
AL	7.25	7.25	7.25	7.25	7.25	7.25	7.25
AK	7.75	7.75	7.75	8.75	9.75	9.80	9.84
AZ	7.65	7.80	7.90	8.05	8.05	10.00	10.50
AR	7.25	7.25	7.25	7.50	8.00	8.50	8.50
CA	8.00	8.00	9.00	9.00	10.00	10.00	11.00
CO	7.64	7.78	8.00	8.23	8.31	9.30	10.20
CT	8.25	8.25	8.70	9.15	9.60	10.10	10.10
DE	7.25	7.25	7.75	8.25	8.25	8.25	8.25
DC	8.25	8.25	9.50	10.50	11.50	11.50	13.25
FL	7.67	7.79	7.93	8.05	8.05	8.10	8.25
GA	7.25	7.25	7.25	7.25	7.25	7.25	7.25
HI	7.25	7.25	7.25	7.75	8.50	9.25	10.10
ID	7.25	7.25	7.25	7.25	7.25	7.25	7.25
IL	8.25	8.25	8.25	8.25	8.25	8.25	8.25
IN	7.25	7.25	7.25	7.25	7.25	7.25	7.25
IA	7.25	7.25	7.25	7.25	7.25	7.25	7.25
KS	7.25	7.25	7.25	7.25	7.25	7.25	7.25
KY	7.25	7.25	7.25	7.25	7.25	7.25	7.25
LA	7.25	7.25	7.25	7.25	7.25	7.25	7.25
ME	7.50	7.50	7.50	7.50	7.50	9.00	10.00
MD	7.25	7.25	7.25	8.25	8.75	8.75	10.10
MA	8.00	8.00	8.00	9.00	10.00	11.00	11.00
MI	7.40	7.40	8.15	8.15	8.50	8.90	9.25
MN	7.25	7.25	8.00	9.00	9.50	9.50	9.86
MS	7.25	7.25	7.25	7.25	7.25	7.25	7.25
MO	7.25	7.35	7.50	7.65	7.65	7.70	7.85
MT	7.65	7.80	7.90	8.05	8.05	8.15	8.30
NE	7.25	7.25	7.25	8.00	9.00	9.00	9.00
NV	8.25	8.25	8.25	8.25	8.25	8.25	8.25
NH	7.25	7.25	7.25	7.25	7.25	7.25	7.25
NJ	7.25	7.25	8.25	8.38	8.38	8.44	8.60
NM	7.50	7.50	7.50	7.50	7.50	7.50	7.50
NY	7.25	7.25	8.00	8.75	9.00	9.70	10.40
NC	7.25	7.25	7.25	7.25	7.25	7.25	7.25
ND	7.25	7.25	7.25	7.25	7.25	7.25	7.25
OH	7.70	7.85	7.95	8.10	8.10	8.15	8.30
OK	7.25	7.25	7.25	7.25	7.25	7.25	7.25
OR	8.80	8.95	9.10	9.25	9.75	9.75	10.75
PA	7.25	7.25	7.25	7.25	7.25	7.25	7.25
RI	7.40	7.75	8.00	9.00	9.60	9.60	10.10
SC	7.25	7.25	7.25	7.25	7.25	7.25	7.25
SD	7.25	7.25	7.25	8.50	8.55	8.65	8.85
TN	7.25	7.25	7.25	7.25	7.25	7.25	7.25
TX	7.25	7.25	7.25	7.25	7.25	7.25	7.25
UT	7.25	7.25	7.25	7.25	7.25	7.25	7.25
VT	8.46	8.60	8.73	9.15	9.60	10.00	10.50
VA	7.25	7.25	7.25	7.25	7.25	7.25	7.25
WA	9.04	9.19	9.32	9.47	9.47	11.00	11.50
WV	7.25	7.25	7.25	8.00	8.75	8.75	8.75
WI	7.25	7.25	7.25	7.25	7.25	7.25	7.25
WY	7.25	7.25	7.25	7.25	7.25	7.25	7.25

Note: Data comes from the Center for Poverty Research. State‐year combinations that are provided by the PRAMS are shaded gray. Years with a change in the state minimum wage are bolded. Please note, models for pre‐pregnancy depression utilize minimum wages in the year of conception while models for PPD utilize minimum wages during the year of interview.

TABLE A3.

Controls statistics.

	Mean (std.)
State controls
Medicaid expansion	0.444
	(0.497)
TANF for family of 3	$711
	(124)
Unemployment rate	6.02
	(1.73)
Democratic governor	0.40
	(0.49)
MH care supply per 1000	1.07
	(4.01)
State GDP growth rate	0.036
	(0.022)
Individual controls
Age	28.06
	(5.52)
Married	0.485
	(0.500)
White NH	0.562
	(0.496)
Black NH	0.169
	(0.375)
Hispanic	0.186
	(0.389)
Parity	0.647
	(0.478)
Dependents	3.02
	(1.39)
Months from birth to interview	4.16
	(1.31)

Note: Data for state controls comes from the University of Kentucky center for poverty research, other than MH care supply which comes from the county business patterns.

TABLE A4.

Robustness test with alternative specifications.

	Base model	Big MW change	Only 1 big MW changestates	Drop unemployment rate control	Drop all state controls	Balanced panel
Pre‐pregnancy depression
Min wage	−0.0110**	−0.0152	−0.0118	−0.0111**	−0.0156***	−0.0108*
	(0.0048)	(0.0122)	(0.0103)	(0.0047)	(0.0041)	(0.0056)
EITC max	−0.0206***	−0.0220***	−0.0203**	−0.0208***	−0.0181***	−0.0140
	(0.0074)	(0.0070)	(0.0094)	(0.0074)	(0.0063)	(0.0116)
N	105,672	105,672	63,575	105,672	105,672	55,750
State‐year obs	185	185	110	185	185	98
PPD
Min wage	−0.0027	−0.0099**	0.0086	−0.0027	−0.0037	−0.0070*
	(0.0034)	(0.0047)	(0.0064)	(0.0036)	(0.0031)	(0.0036)
EITC max	−0.0064	−0.0069	−0.0061	−0.0039	−0.0027	−0.0044
	(0.0071)	(0.0070)	(0.0082)	(0.0070)	(0.0066)	(0.0100)
N	107,228	107,228	64,436	107,228	107,228	56,525
State‐year obs	185	185	110	185	185	98

Note: All models control for state and year fixed effects, month of conception (or birth) fixed effects, demographics, time between birth and interview, mental health care supply, unemployment rate, state GDP growth rate, state TANF levels for a family of 3, state Medicaid expansion, and political party of state governor. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. EITC Max is in thousands of dollars. Policy variables are for year of birth for pre‐pregnancy depression and year of interview for PPD. All models are for observations without a 4‐year college education, unless otherwise noted, using PRAMS data for 2012–2018 births 185 state‐year combinations.

^*** p < 0.01.

^** p < 0.05.

^* p < 0.1.

TABLE A5.

EITC exercises.

	Base model	Fed EITC max separately	EITC state* refundable	EITC max* Mar–June	EITC max* min wage
Depression before
Min wage	−0.0110**	−0.0113**	−0.0111**	−0.0110**	−0.0063
	(0.0048)	(0.0046)	(0.0048)	(0.0048)	(0.0085)
EITC max	−0.0206***	−0.0229***	−0.0248***	−0.0211***	−0.0117
	(0.0074)	(0.0073)	(0.0091)	(0.0073)	(0.0137)
Additional EITC variable	N/A	−0.0174	0.0060	0.0011	−0.0009
		(0.0269)	(0.0154)	(0.0010)	(0.0010)
PPD
Min wage	−0.0027	−0.0028	−0.0029	−0.0027	0.0009
	(0.0034)	(0.0034)	(0.0034)	(0.0033)	(0.0072)
EITC max	−0.0064	−0.0032	−0.0147*	−0.0072	−0.0002
	(0.0071)	(0.0077)	(0.0082)	(0.0073)	(0.0121)
Additional EITC variable	NA	−0.0494	0.0131*	0.0017	−0.0006
		(0.0380)	(0.0068)	(0.0013)	(0.0009)

Note: All models control for state and year fixed effects, month of conception (or birth) fixed effects, demographics, time between birth and interview, mental health care supply, unemployment rate, state GDP growth rate, state TANF levels for a family of 3, state Medicaid expansion, and political party of state governor. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. EITC Max is in thousands of dollars. Policy variables are for year of birth for pre‐pregnancy depression and year of interview for PPD. All models are for observations without a 4‐year college education, unless otherwise noted, using PRAMS data for 2012–2018 births 185 state‐year combinations.

^*** p < 0.01.

^** p < 0.05.

^* p < 0.1.

TABLE A6.

Samples and specifications for postpartum depression (from Table 7)—married only.

	Main sample	Less than HS	College plus
Min wage	−0.0104***	−0.0277*	−0.0030*
	(0.0028)	(0.0150)	(0.0017)
EITC max	0.0059	0.0151	−0.0001
	(0.0094)	(0.0311)	(0.0057)
N	51,168	7146	49,893
Sample mean	0.093	0.135	0.072

	Previous birth	No previous birth	Non‐Hisp white	Non‐Hisp black	Hispanic
Min wage	−0.0120**	−0.0081	−0.0129***	0.0025	−0.0077
	(0.0048)	(0.0082)	(0.0036)	(0.0157)	(0.0065)
EITC max	−0.0047	0.0177	−0.0052	−0.0394	0.0345
	(0.0118)	(0.0175)	(0.0174)	(0.0363)	(0.0280)
N	36,397	14,771	27,838	5475	9868
Sample mean	0.119	0.110	0.111	0.156	0.099

	18–25 year‐olds	Lagged measures	Logged measures	Relative min wage	EITC multiplier
Min wage	−0.0233*	−0.0047	−0.0862***	−0.2026***	−0.0106***
	(0.0242)	(0.0035)	(0.0271)	(0.0590)	(0.0027)
EITC max	0.0298	0.0175*	0.0749	0.0051	0.1083**
	(0.0242)	(0.0089)	(0.0540)	(0.0093)	(0.0439)
N	12,194	51,168	51,168	51,168	51,168
Sample mean	0.150	0.093	0.093	0.093	0.093

	Controlling for DB	Did have DB	Did not have DB		Postpartum well‐being
Min wage	−0.0090***	0.0263*	−0.0125***		0.0121
	(0.0031)	(0.0146)	(0.0030)		(0.0077)
EITC max	0.0105	−0.0371	0.0145*		−0.0108
	(0.0080)	(0.0391)	(0.0083)		(0.0105)
N	50,406	6147	44,259		51,168
Sample mean	0.093	0.269	0.097		0.636

^*** p < 0.01.

^** p < 0.05.

^* p < 0.1.

TABLE A7.

Samples and specifications for postpartum depression (from Table 7)—unmarried only.

	Main sample	Less than HS	College plus
Min wage	0.0044	−0.0201**	0.0104
	(0.0056)	(0.0094)	(0.0094)
EITC max	−0.0218**	−0.0383	−0.0169
	(0.0092)	(0.0276)	(0.0244)
N	56,060	12,149	5346
Sample mean	0.162	0.187	0.114

	Previous birth	No previous birth	Non‐Hisp white	Non‐Hisp black	Hispanic
Min wage	−0.0051	0.0182**	0.0007	0.0106	0.0090
	(0.0068)	(0.0082)	(0.0050)	(0.0142)	(0.0092)
EITC max	−0.0230	−0.0156	−0.0344*	−0.0196*	−0.0051
	(0.0218)	(0.0098)	(0.0182)	(0.0105)	(0.0128)
N	33,120	22,940	21,091	16,512	10,432
Sample mean	0.160	0.164	0.170	0.177	0.117

	18–25 year‐olds	Lagged measures	Logged measures	Relative min wage	EITC multiplier
Min wage	0.0057	0.0024	0.0439	0.0429	0.0040
	(0.0081)	(0.0057)	(0.0533)	(0.0931)	(0.0062)
EITC max	−0.0154	−0.0270***	−0.0002	−0.0216**	0.0453
	(0.0113)	(0.0093)	(0.1293)	(0.0092)	(0.1060)
N	27,287	56,060	56,060	56,060	56,060
Sample mean	0.176	0.162	0.162	0.162	0.162

	Controlling for DB	Did have DB	Did not have DB		Postpartum well‐being
Min wage	0.0062	0.0079	0.0070		0.0076
	(0.0058)	(0.0159)	(0.0057)		(0.0048)
EITC max	−0.0172*	−0.0497	−0.0125*		0.0073
	(0.0088)	(0.0415)	(0.0063)		(0.0133)
N	55,266	9475	45,791		56,060
Sample mean	0.162	0.326	0.128		0.565

Note: All models control for state and year fixed effects, month of interview fixed effects, respondent's race, ethnicity, age, education, time between birth and interview, parity, state mental health care supply, state unemployment rate, time between birth and interview, state GDP growth rate, state TANF levels for a family of 3, state Medicaid expansion, and political party of state governor. All observations are weighted using sample weights and standard errors clustered at the state level are in parenthesis. Policy variables are for year of birth. EITC Max is in thousands of dollars. All models are for observations without a 4‐year college education, unless otherwise noted, using PRAMS data for 2012–2018 births 185 state‐year combinations.

^*** p < 0.01.

^** p < 0.05.

^* p < 0.1.

Data Availability Statement

The data that support the findings of this study are available from PRAMS. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the author(s) with the permission of PRAMS.