Unconditional Cash Transfers and Maternal Employment: Evidence from the Baby’s First Years Study

Abstract

How the labor force participation of mothers of young children responds to unconditioned cash support remains an open question in policy debates. Using data from Baby’s First Years, a large-scale randomized controlled study, we generate new estimates of the impact of an unconditional monthly cash transfer on maternal employment behavior through a child’s first four years of life. We find no overall statistically detectable differences in whether mothers participated in the paid workforce or on total household earnings. Receipt of the cash transfer appears to have reduced hours of maternal work during the height of the pandemic in 2020–21.

1. Introduction

Relative to the OECD average, the U.S. spends one-third as much of its GDP on cash and in-kind benefits for families with children (OECD, 2021). In the case of unconditional cash assistance, concerns in the U.S. about their potential adverse behavioral effects on adults have outweighed consideration of their potential benefits for children and led to a shift toward work-conditioned benefits such as earned income tax credits (Aizer et al., 2022; Hoynes and Schanzenbach, 2018). However, both the COVID-19 pandemic and more than 100 state and local basic income initiatives have revived interest in understanding how people in the U.S. may benefit from a basic level of income provided as unconditional cash transfers (Stanford Basic Income Lab, 2024).

Most empirical studies of labor supply responses of low-income parents to increased income are of programs like the Earned Income Tax Credit (EITC). The EITC leverages income gains that phase in at lower income levels and phase out at higher income levels, making it difficult to separate income and substitution effects.¹ Additionally, the EITC is disbursed annually through a lump sum payment rather than monthly. For these reasons, it is unclear whether findings based on the EITC would be found in programs that transfer cash monthly without conditions or a phase-out.

The main contribution of this study is to estimate the labor market behavior of one important subpopulation of low-income individuals—mothers with young children— in response to an unconditional monthly cash transfer that does not count against income eligibility for receipt of benefits from most existing U.S. safety net programs.²

Economic theory predicts that unconditional cash transfers should reduce labor supply owing to a pure income effect. For the general population, several recent studies have estimated the magnitude of the responses to unconditional income to be small. A review by Marinescu (2018) concludes that the marginal propensity to earn (MPE) out of unearned income derived from lottery studies and negative income tax experiments is −0.10, i.e., an income increase of $100 leads to a $10 decrease in earnings. Data used in recent work based on U.S. lottery winners by Golosov et al. (2024) supported estimates of treatment effect heterogeneous by income level. The authors found a MPE of −0.31 for adults in the lowest quartile of the income distribution (whose pre-win annual earnings averaged $12,000), with no significant differences by gender. However, none of these studies focus on parents with young children. Bibler et al. (2023) analyzed the short-term labor supply effects of the Alaska Permanent Fund, a universal cash transfer, and found that women were more likely to reduce hours worked following the transfer receipt. Noteworthy for this paper is that they found larger reductions for mothers of children under age five, but did not estimate separate labor supply responses for mothers in higher- vs. low-income households.

Our analysis focuses on a racially and geographically diverse sample of low-income mothers with younger children (under four years old) in the United States. The labor supply responses of this group are particularly relevant and expected to differ from those of other populations. Parents of young children are typically in the early stages of their labor market careers, with higher economic instability and increased care burdens. Moreover, women might face challenges in their labor market participation, especially in the first few years after birth, when non-parental childcare is more expensive and less available (Herbst, 2022). However, the lack of access to paid maternity leave and pressing economic needs may force them to quickly return to the labor market following a birth. In 2019, more than half of single mothers with infants under one year old reported working (Bureau of Labor Statistics, 2021). On the other hand, a long delay in the return to work after birth can be costly: the child penalty that mothers experience when they interrupt their labor trajectories after the birth of a child may be an important cause of long-run gender gaps in earnings (Angelov et al., 2013; Kleven et al., 2019). For example, Kuka and Shenhav (2020) find that returning to work sooner after childbirth has long-term positive consequences in the labor market outcomes of low-income mothers, as shown by EITC-driven changes in work incentives.

There is no clear consensus on the size of low-income mothers’ labor supply and behavioral responses to a pure income effect. Exploiting an increase in TANF benefits in New Hampshire in 2017, Freedman and Kim (2022) estimated a 6% decrease in the employment of low-income single mothers, translating into an elasticity of −0.14. However, these benefits are conditional on studying or working. A few studies have measured the effects of unconditional support for low-income parents in the European context. Del Boca et al. (2021) and Aparicio Fenoll and Quaranta (2022) analyzed a cash transfer experiment with low-income parents of 0 to 6-yea-rold children in Italy, which had both conditional and unconditional treatment arms. They found no effects of receiving an unconditional cash transfer on labor market outcomes, while the employment participation was higher for males when the cash transfer was conditional on attending a job search training. Mother’s labor supply did not respond to either of the cash transfer treatment arms. On the other hand, Jensen and Blundell (2024) estimated a sizeable income elasticity for mothers with young children as a response to an (unconditional) child benefit reduction in Denmark, particularly at the extensive margin. They estimate an elasticity between −2.7 and −0.9.

Baby’s First Years (BFY) is a randomized control study of an unconditional cash transfer made available to 1,000 low-income mothers who gave birth in 12 U.S. hospitals in four cities in 2018–19. Upon consent, mothers were randomized to receive a monthly unconditional cash gift of either $333 or $20, starting at their child’s birth, for a duration of 40 months (subsequently extended to 76 months). The high-cash gift amount corresponds to 17% of baseline household income. Take-up of the cash gift was nearly universal (Gennetian et al., 2023), and the study has achieved 90+% response rates in each of its four annual waves of survey data collection.

Overall, we found mostly negative but not statistically significant effects of the high-cash gift on the extensive and intensive margins of labor supply, as well as self-reported annual earnings, over the four years of the study analysis period —which spans the interval between the focal child’s birth and fourth birthday.

At the extensive margin, we rule out effects larger than 6 percentage points. Our estimated four-year pooled impact on the probability of mothers working for pay was −1 percentage point, or −2.8%.³ This point estimate translates into an income elasticity of −.16, which is similar to findings on single heads of households in the negative income tax experiments (Robins, 1985) (ϵ=−.16), low-income single mothers in New Hampshire (Freedman and Kim, 2022) (ϵ=−.14), and married women and single mothers’ responses to tax changes across the 70s and 80s (Blundell and Macurdy, 1999; National Academies of Sciences, 2019) (between ϵ=−.085 and ϵ=−.12). It is also significantly smaller than the effect estimated for Danish mothers (Jensen and Blundell, 2024). While still statistically insignificant, the estimated extensive margin coefficient was larger at the time of the child’s first birthday (a 4-percentage-point group difference, corresponding to a 9% reduction in employment), converging to zero in the following years.⁴ Additionally, the cash gift did not affect the probability of living in a household with no employed adults (measured as a household with no earned income), which is the case for about 15% of our sample.

Our pooled estimates also allow us to rule out a decrease of more than 4.7 hours worked per week at the intensive margin. However, looking at each wave separately, we found a statistically significant decrease of 6.0 hours worked per week (−39%) at the time of the child’s second birthday and in the midst of the COVID-19 pandemic, and null results at the time of the first, third, and fourth birthday. Our exploratory analysis suggests that the BFY cash gift facilitated behavioral adjustments to the pandemic.

Finally, we rule out changes in maternal annual earnings of more than $1500 (i.e., a marginal propensity to earn out of the cash gift of −0.4). We estimated a statistically insignificant $451.21/year reduction in maternal earnings and $177.74/year reduction in nonmaternal household earnings, corresponding to a marginal propensity to earn out of the cash gift of −0.15 and −0.10, respectively. These point estimates are quite similar to those found for the general population in other settings (Marinescu, 2018), although smaller than lottery-based estimates for low-income individuals (Golosov et al., 2024).

In sum, although our estimates lacked precision to detect small effects, we can rule out moderate and large responses (e.g., those estimated by Jensen and Blundell (2024) or Golosov et al. (2024)). Additionally, our results have important policy implications for discussions around child allowances or a permanent, fully refundable Child Tax Credit, addressing concerns about potentially large employment losses.⁵

The COVID-19 pandemic occurred during the BFY study, which raises some concerns. The contraction of the economy and job losses during the early months of the pandemic might have attenuated labor supply responses to the cash gift, limiting the extent to which our results can generalize to other contexts. However, because we observed maternal employment over four years (from July 2019 to July 2023), the pandemic period prior to the widespread availability of vaccinations comprised only about one-quarter of the total reporting period. Compared with other recessions, the labor market recovered from the COVID-19 pandemic very quickly. The labor force participation rate, particularly for the “prime-age” group (ages 25–54), was already at the pre-pandemic levels in mid-2022 (The White House, 2023). Hence, our study enables us to estimate treatment effects across very different economic contexts.

The data also allowed us to explore the extent to which the pandemic interacts with the cash-gift effect, which is theoretically ambiguous. A higher cash gift could have allowed mothers to offset work participation or hours during the pandemic, at a time when in-person work may have been dangerous and child care was scarce. On the other hand, the cash gift could have supported mothers’ work participation or hours during the pandemic because it could be used to cover pandemic-related costs of working (e.g., engaging in more protective behaviors, improving internet/technology, paying for alternative child care arrangements when schools closed, etc.).

As previously indicated, we found that mothers in the high-cash gift group reduced working hours and were significantly less likely to work full-time than mothers in the low-cash gift group (an effect of −6 p.p. in full-time work, or −27%) at the time of their child’s second birthday, which was early in the COVID-19 pandemic when public health protocols and vaccination availability were rapidly evolving. Labor supply differences were not apparent in the two following waves of data collection. Moreover, mothers in the high-cash gift group were significantly more likely to report making major behavioral changes to limit contact with people outside their homes and losing income due to COVID. Finally, the reduction in weekly hours was not associated with their site-specific labor market conditions during the interview month. Together, these results suggest a possible role of cash support in addressing health and family concerns during the COVID-19 pandemic.

2. The Baby’s First Years Study

Between May 2018 and June 2019, the Baby’s First Years study recruited 1,000 low-income mothers in New Orleans, New York City, Omaha Metropolitan area, and the Twin Cities of Minneapolis and St. Paul. Potential participants at twelve hospitals were approached on postpartum wards after giving birth, assessed for study eligibility, and, if eligible, offered enrollment in the study.⁶ Study recruitment was designed to provide an even flow of participants across the 12-month recruiting period. Details about the study design can be found on the study website, in Noble et al. (2021), and are pre-registered (Duncan et al., 2019).⁷. After completing a baseline survey, participants were invited to receive a cash gift with random assignment into a high-cash gift group (n = 400) or a low-cash gift group (n = 600). The former is receiving $333 per month ($3,996/year) and the latter $20 per month ($240/year) for an initially planned duration of 40 months, then extended for a total duration of 76 months, with no restrictions regarding how the money could be spent. The cash gifts were loaded onto a debit card with a “4MyBaby” logo printed on the card. The card was activated and given to the mothers in the hospital. Each month, either $333 or $20 is loaded on the same day of the month as the child’s birth with an accompanying text message. A hotline was available to assist in using the BFY card, and most issues that arose were easily and quickly able to be resolved (e.g., replacement cards). Mothers were able to use the card effectively. For the subsample of mothers who consented to share card transaction information (n=900), as of July 2022, only 7 mothers in the low-cash gift group had failed to ever use their cards (and none in the high-cash group failed to do so), suggesting that high implementation success of the cash gift disbursement approach. Additional relevant information about the study implementation can be found in Gennetian et al. (2023).

Our ability to detect small income effects is hampered by the study’s sample size. We have statistical power to identify treatment effects of a 6-percentage-point change in the probability of working for pay at a .05 significance level and 80% power (Bloom, 1995), given the control group employment rate of 48%, averaged across waves. Given that the cash gift represents a 17% increase in household income, our minimum detectable elasticity is 0.7. We can also detect a change of $1,500 in maternal annual earnings and $2,800 in household annual earnings, given a control group base of $10,970 and $23,320, respectively. This corresponds to minimum detectable propensities to earn (MPE) out of the cash transfer of −.40 and −.76 at the individual and household level, respectively. On the other hand, the sample size does enable us to rule out some of the larger labor supply reductions suggested in policy discussions, as well as the elasticities found by Jensen and Blundell (2024) for Danish mothers, which were between −2.7 to −0.9.

3. Empirical Strategy

3.1. Participants

Some 976 out of 1,000 participants responded to at least one of the four follow-up surveys, which were scheduled to take place around the time of the children’s 1st, 2nd, 3rd, and 4th birthdays.⁸ Because recruitment was spread across a 12-month period, these follow-up surveys provide a fairly continuous time series of labor supply responses across the entire follow-up period.

Response rates among eligible mothers were consistently very high: 94% at Age 1 (931 respondents), 93% at Age-2 (922 respondents), 93% at Age-3 (922 respondents), and 90% at age 4 (888 respondents). More details about sample construction and reasons for non-response can be found in the CONSORT diagram in Appendix A1. Reasons for non-response included a small number of sample exclusions (either the child or mother was deceased), ineligibility to complete a survey wave (when the mother was incarcerated or did not have custody of the child), as well as a number of participants who were not found or were unavailable or refused to participate in the survey.⁹

Age-1 interviews were scheduled to be conducted in person between July 2019 and June 2020. The COVID-19 pandemic began in March 2020; as a result, in-home interviews were suspended on March 12, and the study transitioned to phone interviews beginning on March 14. A total of 326 phone interviews (35 percent of the Age-1 total) were conducted, which means that nearly two-thirds of the Age-1 labor market information was gathered before the onset of the pandemic. All Age-2 and Age-3 interviews were completed over the phone. At Age-4, participants were scheduled for in-person data collection at university labs. The Age-4 outcomes for this paper come from a short survey that was conducted during the lab visit. For mothers who were not able to bring their children to a lab, the survey was conducted over the phone.

Descriptive statistics measured at baseline (i.e., pre-randomization) are included in Appendix B3, for the sample of participants who responded to any of the follow-up surveys (n=976), as well as each wave separately. The participating BFY study mothers were diverse along several dimensions, but well-balanced between the two cash-gift groups. Characteristics for the full sample of 1,000 enrolled mothers can be found in Noble et al. (2021). At the time of enrollment in the study, mothers in both the low- and high-cash gift groups reported similar levels of labor market participation and intentions to work in the future; about 57 percent of mothers in both groups worked while pregnant, 85 percent planned to work at some point during the next year, and more than half were planning to work within three months of their child’s birth.¹⁰ Joint significance tests of all the listed covariates predicting cash group assignment suggested no treatment-group differences across respondents in any of the four follow-up survey samples. Nevertheless, as described in Section 3.3, baseline variables were included in the regressions to improve the precision of the impact estimates.

3.2. Measures

We have several variables designed to capture maternal and household labor market participation, which we organize into three groups: whether mothers participate in the paid labor market (extensive margin), hours of paid work (intensive margin), and maternal and household annual earnings.

Mothers were asked about paid employment at the time of the survey interview: whether they were working for pay, whether they were self-employed¹¹, and how many hours they had worked in an average week in the last month. We create an indicator for working at least 35 hours per week.

Mothers were also asked to report their annual earnings from work as well as income from other sources during the previous calendar year. We construct a measure of earnings for the mother (available for all four waves) and a measure of earnings from all other adults in the household (available for waves 1 to 3).¹² To minimize the impact of outliers, all earnings measures are truncated at the 99th percentile. We contrast the results with and without truncation as a sensitivity test in Appendix B4. Calendar-year earnings correspond to either 2018 or 2019 at Age-1, 2019 or 2020 at Age-2, 2020 or 2021 at Age-3, and 2021 or 2022 at Age-4.¹³ To assess whether unconditional cash support might increase the likelihood of children residing in households without any employed adults (Rachidi and Doar, 2019; Winship, 2021), we also estimate impacts on an indicator of whether the household has any positive earnings in the prior calendar year. While the amount of non-maternal earnings is not reported at Age-4, we do have an indicator variable for whether there were any nonmaternal earnings in the household in the previous week.

Finally, the second year of follow-up survey data included a module of questions related to how the pandemic affected their income, health, and general behavior changes, which we explore in section 4.3.¹⁴. We provide more details on the alignments between the timing of relevant measures, data collection waves, and calendar time in Appendix A2.

3.3. Analysis

Because assignment to the high-cash gift group was random by design, securing an unbiased estimate of the impact of the higher cash gift (a difference of $313/month between groups) is relatively straightforward.

Our estimates pooling the four waves of data are derived from the following model:

$$Y_{isw}={\beta }_0+{\beta }_1{\mathit{HighCashGift}}_{isw}+X_{isw}+{\gamma }_w+{\delta }_s+{\epsilon }_{isw}$$ (1)

where $i$ indexes the mother, $s$ indexes the four sites, $w$ indexes the wave of data collection, $Y$ is the outcome of interest, $HighCashGift$ is an indicator for being assigned to receive $333 as opposed to $20 per month, $X$ is a vector of covariates, ${\gamma }_w$ is a vector of wave indicators, ${delta}_s$ is a set of four site fixed effects, and $\epsilon$ is the error term. Site fixed effects are needed since randomization occurred within site. Standard errors were clustered at the individual level. The coefficient ${\beta }_1$ provides an estimate of the causal effects on $Y$ of being assigned to the high as opposed to low-cash gift group. Given mothers’ nearly universal use of the money in both groups, the intention-to-treat estimates can also be interpreted as the effects of receiving a monthly cash gift of $313, i.e., treatment-on-the-treated (TOT) estimates. Regressions were estimated using Ordinary Least Squares, while the effects on the number of hours worked were also estimated using a tobit model. For annual earnings, we also control for the calendar year for which earnings are reported. Finally, we also estimate the same specification for each wave separately, plus a fully interacted model to assess whether the estimates differ across waves.

The pre-registered plan specified the following baseline covariates to be used to improve the precision of causal estimates: mother’s age, maximum education level attained, race and ethnicity, marital status, general health, maternal depressive symptoms, cigarette and alcohol consumption during pregnancy, number of children born to mother, number of adults in the household, father living with the mother, household income, household net worth, and baby’s weight and gestational age at birth. We added the following pre-birth (and therefore pre-random assignment) maternal employment measures: worked while pregnant, continued working until birth, and plans to return to work after birth. We also control for two post-randomization variables: a calendar month indicator and the time distance to the child’s birthday within wave in months (i.e., how far away they were from the scheduled data collection).

4. Results

4.1. Descriptive Results

We first present unadjusted means of employment by month and cash-gift group (Figure 1). Because each mother is interviewed only once per year, around the time of her child’s birthday, the within-wave monthly variation in Figure 1 comes from the fact that BFY sample recruiting and subsequent follow-ups were spread across roughly 12-month periods. Recruitment began in May 2018, and the Age-4 reinterview period ended in July 2023. Figure 1 shows the four annual data collection periods mapped against child age and three pandemic periods: pre-pandemic (before March 14, 2020, when data collection switched to phone interviews), pandemic pre-vaccine (between March 14, 2020, and April 19, 2021; when vaccines became available to all adults in the U.S.), and pandemic after vaccine (post April 19, 2021). As might be expected, the proportion of mothers reporting working for pay decreased during the pandemic months of 2020 but began to show a positive trend for both groups by the beginning of 2021. It appears that the high-cash gift group mothers were less likely to be working for pay during the first year, particularly those interviewed before COVID. Employment rates seem to converge after this point. Approximately between 40% and 60% of the mothers reported working for pay across this period. As a reference, the employment-population ratio of all women with children under 3, regardless of income, was .59 in 2021 (US Census Bureau, 2021).

Figure 1: Maternal employment by interview month.

Notes: Markers in Panel A illustrate the percentage of participants who responded “yes” to the question “Are you currently working for pay?”, by interview date, while those in Panel B represent the average hours worked per week in the four weeks prior to the interview. Hours are truncated at the 99th percentile. Each bin represents a calendar month. To reduce the noise caused by some sample unbalance at the beginning and the end of each wave, the first and last months are winsorized within waves. I.e., the first bin in each wave combines the interviews conducted before August and the last bin combines the interviews conducted after May. Lines represent a LOWESS smoothed trend in the variable with a 0.6 bandwidth. Each wave of data collection is indicated by the gray background rectangles. Vertical black lines illustrate three main “pandemic periods”, based on the date when interviews switched from in-person to phone-based (March 14, 2020) and the date when all adults in the U.S. became eligible for COVID-19 vaccines (April 19, 2021). Data for Panel B is missing for the first months of data collection due to an error in the survey implementation.

4.2. Main Estimates of the BFY Cash Gift on Employment

Our main results summarizing the effects of the BFY cash gift on maternal and household employment and earnings estimated using Equation 1 are reported in Tables 1 and 2, with robustness checks shown in Appendix B2. Table 1 shows the impacts on maternal employment. In all four waves, group differences at the extensive employment margin (columns 1–2) were not statistically significant. As suggested in Figure 1, the coefficient size is larger at wave 1 (b=−0.04). However, we cannot reject that the estimates for all waves are equal. In the pooled analysis that includes data from all four follow-up surveys (Panel A), we estimate a 1 percentage-point (−3%) lower proportion of high- vs. low-cash mothers in the likelihood of working for pay but cannot reject the null hypothesis of no difference (p = 0.55): the 95% confidence interval ranges from −5.6 and +2.9 percentage points. While null effects do not imply the absence of a labor supply response, our estimates rule out effects larger than 6 percentage points in the probability of working for pay.

Table 1:

Main effects of BFY cash gift on maternal employment

	Working for pay	Self-employed	Working full-time	Weekly Hours (OLS)	Weekly Hours (Tobit)	Annual earnings
	(1)	(2)	(3)	(4)	(5)	(6)
Panel A. Pooled results
High cash gift	−0.01 (0.02)	0.01 (0.02)	−0.03 (0.02)	−1.62+ (0.90)	−2.08 (1.70)	−451.21 (532.78)
Observations	3661	3323	3283	3194	3194	3582
Low-cash gift group mean	0.48	0.15	0.31	17.91	17.91	10967.79
Effect in %	−2.76%	4.73%	−9.48%	−9.03%	−11.61%	−4.11%
Panel B. Wave 1
High cash gift	−0.04 (0.03)	0.04 (0.03)	−0.01 (0.03)	−0.46 (1.53)	−0.51 (3.20)	204.10 (508.07)
Observations	931	593	582	582	582	922
Low-cash gift group mean	0.45	0.11	0.19	13.41	13.41	7254.38
Effect in %	−9.29%	39.46%	−7.42%	−3.44%	−3.79%	2.81%
Panel C. Wave 2
High cash gift	−0.02 (0.03)	0.01 (0.02)	−0.07* (0.03)	−3.27** (1.18)	−6.02* (2.49)	−229.53 (676.47)
Observations	921	921	908	908	908	907
Low-cash gift group mean	0.42	0.13	0.24	15.41	15.41	9868.70
Effect in %	−4.81%	7.10%	−27.56%	−21.21%	−39.05%	−2.33%
Panel D. Wave 3
High cash gift	0.01 (0.03)	0.01 (0.02)	0.01 (0.03)	−0.32 (1.26)	0.77 (2.23)	−816.22 (829.99)
Observations	922	922	910	910	910	913
Low-cash gift group mean	0.50	0.13	0.30	18.46	18.46	11391.83
Effect in %	2.39%	7.12%	1.88%	−1.74%	4.17%	−7.16%
Panel E. Wave 4
High cash gift	−0.01 (0.03)	−0.02 (0.03)	−0.04 (0.03)	−2.04 (1.41)	−2.48 (2.23)	−1021.38 (984.82)
Observations	887	887	883	794	794	840
Low-cash gift group mean	0.56	0.21	0.46	23.49	23.49	15816.61
Effect in %	−1.58%	−8.39%	−7.79%	−8.67%	−10.57%	−6.46%
Panel F. Additional Tests
All waves are equal (p-val)	0.615	0.485	0.162	0.148	0.089	0.279
All waves are equal to zero (p-val)	0.670	0.602	0.091	0.067	0.096	0.426

Notes:

⁺p<.10

^*p<.05

^**p<.01.

Robust standard errors in parentheses. All regressions include site fixed effects and the following covariates: mother’s age, maximum education level attained, race and ethnicity, marital status, general health, maternal depressive symptoms, cigarette and alcohol consumption during pregnancy, number of children born to mother, number of adults in the household, father living with the mother, household income, household net worth, baby’s weight and gestational age at birth, mother worked while pregnant, continued working until birth, plans to go back to work after birth, calendar month indicator, and the time distance to the child birthday within wave, in months (i.e., how far away they were from the scheduled data collection). Pooled estimates also include wave fixed effects and standard errors are clustered at the individual level. Working full-time is defined as working at least 35 hours per week.

Table 2:

Main effects of BFY cash gift on household earnings

	HH non-maternal earnings	Total HH earnings	HH has zero earnings (indicator)	HH has non-maternal earnings (prev.week)
	(1)	(2)	(3)	(4)
Panel A. Pooled (waves 1–3)
High cash gift	−177.74 (853.71)	−406.90 (1024.13)	−0.01 (0.02)
Observations	2654	2638	2638
Low-cash gift group mean	12554.16	22322.04	0.15
Effect in %	−1.42%	−1.82%	−4.66%
Panel B. Wave 1
High cash gift	−574.68 (1004.71)	−467.77 (1140.05)	0.00 (0.02)
Observations	892	887	887
Low-cash gift group mean	12995.78	20272.44	0.12
Effect in %	−4.42%	−2.31%	0.74%
Panel C. Wave 2
High cash gift	−618.94 (1199.28)	−674.21 (1513.98)	−0.01 (0.02)
Observations	879	873	873
Low-cash gift group mean	12777.73	22906.13	0.15
Effect in %	−4.84%	−2.94%	−6.16%
Panel D. Wave 3
High cash gift	549.88 (1202.54)	−206.72 (1528.99)	−0.01 (0.03)
Observations	883	878	878
Low-cash gift group mean	11883.79	23788.13	0.18
Effect in %	4.63%	−0.87%	−6.65%
Panel E. Wave 4
High cash gift				0.01 (0.03)
Observations				878
Low-cash gift group mean				0.38
Effect in %				3.89%

Notes:

⁺p<.10

^*p<.05

^**p<.01.

Standard errors in parentheses. All regressions include site fixed effects and the following covariates: mother’s age, maximum education level attained, race and ethnicity, marital status, general health, maternal depressive symptoms, cigarette and alcohol consumption during pregnancy, number of children born to mother, number of adults in the household, father living with the mother, household income, household net worth, baby’s weight and gestational age at birth, mother worked while pregnant, continued working until birth, plans to go back to work after birth, calendar month indicator, the time distance to the child birthday within wave, in months (i.e., how far away they were from the scheduled data collection, and interview year. Pooled regressions also include wave fixed effects and standard errors are clustered at the individual level.

At the intensive margin (columns 3–5), at the time of the Age-2 survey, which overlapped substantially with the pre-vaccine phase of the pandemic, there was a statistically significant 7 percentage-point reduction (−27%) in the probability of working full-time (more than 35 hours per week) (p<0.05) and a reduction of 6 hours worked per week (−39%). As shown by the hours’ histograms in Appendix A3, much of this reduction was concentrated at the 40-hour threshold. This decline is robust to different specifications related to covariates, adjustments to non-response, and effect bounds based on Lee (2009) (Appendix B2). Impacts on the intensive margin were null at the time of the Age-1, Age-3, and Age-4 surveys.

Finally, the effects on maternal annual earnings were also small and not statistically significant at conventional levels. We estimated a decrease of $450 in annual earnings, which implies a marginal propensity to earn out of the cash gift of −.15.¹⁵ Additional tests in Appendix B4 show that these results are not sensitive to the inclusion of outliers. We also rule out a MPE larger than −0.4.

The effects of the BFY cash gift on household earnings are reported in Table 2. For the age-1 to age-3 pooled analysis, point estimates were −178 (−1.4%) in non-maternal earnings, and −407 in total household earnings, implying a marginal propensity to earn out of the cash gift of −.10 at the household level.

The differences in the timing of measurement of employment status and earnings do not allow us to precisely estimate whether receiving the high cash gift had an effect on wage rates. However, since annual earnings are reported for the calendar year preceding the interview, we can approximate a wage rate using annual earnings reported in a given year and the hours reported in the previous year. We do not find a significant effect of the cash gift on this estimated wage rate.

Finally, we tested for differences in whether mothers reported that no members of their households had positive earnings in the calendar year prior to the Age-1 to Age-3 interviews. Some 15% of both low- and high-cash gift group households fell into this category. The point estimate of the BFY impact on this measure of the absence of work is negative (i.e., the high-cash gift group is less likely to have no earnings from work) but far from any threshold of statistical significance. Moreover, at the time of the Age-4 survey, there were no significant differences in the probability of living in a household with non-maternal earnings.

In sum, we do not detect statistically significant reductions in maternal employment as a response to the BFY cash gift, except for a reduction in the hours worked at the time of the Age-2 survey. This is further explored in the next section. Moreover, the results are robust to the exclusion of covariates or the inclusion of non-response weights (Appendix B2). To further address attrition concerns, we provide Lee bounds to our estimates. Finally, we explored whether our results were hiding some relevant heterogeneity by population subgroup (Appendix B6). In particular, we considered some variables that are arguably related to labor market participation outcomes: level of education, presence of a partner in the household, number of (older) children besides the BFY-focal child, and previous labor market involvement (during pregnancy). We found suggestive evidence that mothers who had a partner at the time of the baby’s birth might have reduced their employment significantly more than those who did not have a partner, and mothers without high school completion also experienced a larger reduction in earnings.

4.3. The COVID-10 Pandemic, Cash Gifts, and Maternal Employment

The pandemic was an important context for the families in the study. While the timing of the pandemic is a valid concern regarding the external validity and interpretation of the results, the four-year duration of the data collection allows us to offer some insights into how the pandemic interacted with the cash gift.

Having four waves of data, we were able to observe the employment outcomes of mothers throughout a considerably long period of time. Survey interviews were completed: i) during the pre-pandemic period (between July 2019 and March 14, 2020; when data collection switched to phone interviews); ii) during the pandemic period before vaccines became widely available (between March 14, 2020, and April 19, 2021; and iii) the period after vaccines became widely available (between April 19, 2021, and July 2023). While the pandemic might have had an important influence during the end of wave 1 and wave 2, wave 3 and wave 4 were collected in the post-vaccine era, allowing us to better disentangle the incidence of the COVID-19 pandemic and increasing the external validity of our study. In fact, when the age-4 data collection started, employment levels across the country were already at the pre-pandemic levels.

Table 1 showed that group differences in maternal labor force participation were negative and statistically significant at Age-2 but had disappeared completely at the time of the Age-3 and Age-4 surveys. Because much of the Age-2 data collection overlapped with the pandemic pre-vaccine period, these findings suggest that the high-cash gift mothers reduced their labor market hours during the heightened and potentially riskiest time of the pandemic. Next, we further explored whether we can attribute the Age-2 results to the pandemic.

COVID-specific questions asked at Age-2 provide an additional opportunity to analyze whether mothers’ employment behavior was associated with the pandemic. Table 3 shows that mothers in the high-cash gift group were 11.7% and 19.7% more likely than mothers in the low-cash gift group to live in a household where at least one person lost income or received unemployment benefits due to the pandemic, respectively. Both of these estimates are statistically significant at conventional levels. Moreover, although high-cash-gift mothers were not statistically significantly less likely to get COVID, they were 7 percentage points (9.8%) more likely to report having made major changes to their behavior, such as not going to school or work or limiting contact with people outside their home.

Table 3:

Effects of the BFY cash gift on COVID-19-related outcomes

	Lost income due to COVID	Unemployment benefits	COVID diagnosis	Major changes to behavior
	(1)	(2)	(3)	(4)
Panel A: Anyone in the household
High cash gift group	0.07* (0.03)	0.07* (0.03)	0.03 (0.03)
Observations	918	911	916
Low-cash gift group mean	0.60	0.36	0.16
Effect in %	11.70%	19.72%	18.50%
Panel B: Mother
High cash gift group	0.06+ (0.03)	0.04 (0.03)	0.01 (0.02)	0.07* (0.03)
Observations	918	912	916	913
Low-cash gift group mean	0.42	0.27	0.13	0.69
Effect in %	15.05%	14.40%	4.72%	9.73%
Panel C: Other household member (not mother)
High cash gift group	0.03 (0.03)	0.03 (0.03)	0.03 (0.02)
Observations	918	911	918
Low-cash gift group mean	0.30	0.17	0.11
Effect in %	9.91%	16.30%	29.04%

Notes:

⁺p<.10

^*p<.05

^**p<.01.

Standard errors in parentheses. All regressions include site fixed effects and the following covariates: mother’s age, maximum education level attained, race and ethnicity, marital status, general health, maternal depressive symptoms, cigarette and alcohol consumption during pregnancy, number of children born to mother, number of adults in the household, father living with the mother, household income, household net worth, baby’s weight and gestational age at birth, mother worked while pregnant, continued working until birth, plans to go back to work after birth, calendar month indicator, and the time distance to the child birthday within wave, in months (i.e., how far away they were from the scheduled data collection). The outcome in Column (4) refers to the third option in the following question: ”Thinking back to when your community was most affected by the coronavirus pandemic, which one of the following statements describes how much you changed your behavior in response to social distancing and stay-at-home recommendations? 1- I made no changes to my behavior; 2-I made minor changes to my behavior (for example, going out less, seeing fewer friends); 3-I made major changes to my behavior (for example, not going to school or work or limiting my contact with people outside our home)

We then explore whether the employment conditions during the pandemic might be driving the Age-2 results. To do this, we interact the high-cash gift coefficient with the unemployment rate in the local area during the month of data collection. Arguably, variations in monthly unemployment rates reflect pandemic-related changes in employment conditions (see Appendix A4).¹⁶

As observed in Table 4, a 1-percentage-point increase in local unemployment was associated with a 1-percentage-point decrease in maternal employment rate, and this association is similar in high- and low-cash gift groups. However, the coefficient on the interaction between being in the high cash gift group and unemployment rates is quite small and not statistically significant. It appears, then, that reductions in work for high-cash gift mothers when their children were two years of age may have been more responsive to health and family concerns than to local employment conditions.¹⁷

Table 4:

Effects of the BFY cash gift interacted with local monthly unemployment rate

	Working for pay		Self-employment		Working full-time
	(1)	(2)	(3)	(4)	(5)	(6)
High cash gift	−0.015 (0.022)	0.007 (0.034)	0.007 (0.017)	−0.002 (0.028)	−0.031 (0.020)	−0.008 (0.034)
Monthly unemployment rate (site-specific)	−0.013** (0.003)	−0.012** (0.004)	−0.003 (0.003)	−0.003 (0.003)	−0.006* (0.003)	−0.005 (0.003)
Unemployment rate X High Cash Gift		−0.004 (0.005)		0.002 (0.004)		−0.004 (0.004)
Observations	3661		3323		3283
Low-cash gift group mean	0.481		0.148		0.305

Notes:

⁺p<.10

^*p<.05

^**p<.01.

Standard errors in parentheses. All regressions include site and wave fixed effects and the following covariates: mother’s age, maximum education level attained, race and ethnicity, marital status, general health, maternal depressive symptoms, cigarette and alcohol consumption during pregnancy, number of children born to mother, number of adults in the household, father living with the mother, household income, household net worth, baby’s weight and gestational age at birth, mother worked while pregnant, continued working until birth, plans to go back to work after birth, calendar month indicator, and the time distance to the child birthday within wave, in months (i.e., how far away they were from the scheduled data collection). Standard errors are clustered at the individual level.

In sum, based on the timing of the observed reductions in hours worked during the height of the pandemic, together with the COVID-specific questions collected at Age-2, and the additional analysis interacting the cash gift group with local employment conditions, we believe that the high cash gift allowed mothers to reduce their hours of paid work as a response to pandemic-driven disruptions in daily life.

5. Conclusion

This study examined the labor supply behavior of mothers with young children in response to a pure income effect created by the receipt of a monthly unconditional cash gift. Findings showed that receiving a $333 monthly cash gift (as opposed to $20) over the first four years after the child’s birth led to null to small declines in maternal employment and household earnings. A reduction in maternal work hours was limited to the period around the child’s second birthday which also coincided with the height of the COVID pandemic. Our study’s exploratory analysis of potential COVID “contamination” suggests that the cash gift may have allowed low-income mothers to reduce work-based exposure to COVID or stay home altogether. At the time of the age-3 or age-4 surveys, which were implemented between July 2021 and July 2023, there were no differences in employment between the two groups at any margin.

A potential concern that would limit the external validity of our study is the design of the card, labeled ”4MyBaby”.¹⁸ While we cannot directly test whether there is a labeling effect, a look at the transaction data (i.e., categories of expenditures directly derived from the use of the card), as well as a qualitative sub-study, suggest that mothers used the BFY cash gift in a large variety of ways, and mostly to cover existing expenses (Gennetian et al., 2022; Halpern-Meekin et al., 2024). Additionally, it is important to note that mothers in both the high- and the low-cash gift group received a similar card, with the only difference being the gift amount. In this sense, concerns regarding potential psychological responses triggered by the label, such as conveying the significance of investing in children, should be alleviated.

Even though we could not detect small income effects, we can rule out a 6 percentage point decrease in paid work participation, a 4.7 reduction in hours worked per week, and a $1500 reduction in maternal earnings (i.e., $125 per month). The main specification in our analysis yields a point estimate of −2.8% employment reduction over the study’s first four years. Given that the cash gift represents a 17% increase in household income, our estimate corresponds to an elasticity of −.16. Although not statistically significant, this result is qualitatively similar to elasticities estimated with other income changes such as the −.16 elasticity found for single female heads of households in Negative Income Tax experiments (Robins, 1985), the −.14 elasticity of low-income single mothers in New Hampshire as a response to a change in TANF benefits (Freedman and Kim, 2022), or rough midpoints of −.12 for married women, and −.085 for single mothers used in the NAS report (National Academies of Sciences, 2019) based on Blundell and Macurdy (1999) review of tax changes across the 1970s and 1980s in the U.S. and Europe. While also not statistically significant with our sample, the estimated marginal propensity to earn out of unearned income was −.15 for maternal earnings and −.10 for household earnings. These estimates align with the prior literature on lottery winners and negative income tax experiments, and are smaller than the estimates reported by Golosov et al. (2024) for US lottery winners with similar annual incomes. We also ruled out the large elasticities found by Jensen and Blundell (2024).

Finally, by tracking both labor supply responses as well as impacts related to child development, the BFY study of unconditional cash benefits contributes to our understanding of both the benefits for children and the possible behavioral costs of such a policy (Aizer et al., 2022). And, despite concerns about its statistical power, this study makes an important contribution by estimating income effects for the sample population of low-income mothers with young children, which has received little attention in the literature and is subject to increasing public attention in the context of child allowance discussions, such as the expansion of child tax credits.¹⁹

Highlights.

• We estimate income effects for low-income mothers with children from birth to age 4.

• Mothers were randomized to receive $333 or $20 per month after their child’s birth.

• We find small and statistically insignificant effects on maternal labor supply.

• There was a significant effect on hours worked during the height of the pandemic.