Policy Research Working Paper               10239




        The Effects of Childcare on Women
                   and Children
  Evidence from a Randomized Evaluation in Burkina Faso

                        Kehinde F. Ajayi
                            Aziz Dao
                        Estelle Koussoubé




Africa Region
Gender Innovation Lab
November 2022
Policy Research Working Paper 10239


  Abstract
 This paper studies whether providing affordable childcare                         childcare center usage for children aged 0 to 6 years.
 improves women’s economic empowerment and child                                   Women’s employment and financial outcomes improved.
 development, using data from a sample of 1,990 women                              Additionally, child development scores increased. However,
 participating in a public works program in Burkina Faso.                          the analysis finds no significant effects on women’s deci-
 Of 36 urban work sites, 18 were randomly selected to                              sion-making autonomy, gender attitudes, or intrahousehold
 receive community-based childcare centers. One in four                            dynamics, suggesting the importance of considering multi-
 women who were offered the centers used them, tripling                            ple dimensions of childcare impacts.




 This paper is a product of the Gender Innovation Lab, Africa Region. It is part of a larger effort by the World Bank to
 provide open access to its research and make a contribution to development policy discussions around the world. Policy
 Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted
 at mkoussoube@worldbank.org.




        The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
        issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
        names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
        of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
        its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                      Produced by the Research Support Team
The Effects of Childcare on Women and Children: Evidence from a
            Randomized Evaluation in Burkina Faso∗
                    Kehinde F. Ajayi†               Aziz Dao‡                            e§
                                                                         Estelle Koussoub´




          Keywords : gender, labor, welfare, childcare, early childhood development
          JEL codes : J16, J13, I38, O15




   ∗
     This paper is a product of the World Bank Africa Gender Innovation Lab, Office of the Chief Economist, Africa
Region. We thank David Evans, Markus Goldstein, Rachael Pierotti, Christophe Rockmore, colleagues from the
Africa Gender Innovation Lab, and seminar participants from the Center for Global Development and the Center
for the Study of African Economies 2022 Conference for thoughtful comments. We are grateful to Rebekka Grun,
Gilberte Kedote, Florence Kantiono, Roland Berehoudougou, Marie Bernadette Kabre, L’Institut de Formation et
de Sp´ecialisation des Personnels de la Petite Enfance (IFSPPE), and the PTR-HIMO team for their collaboration on
this project. This work was funded by the Early Learning Partnership and the Umbrella Facility for Gender Equality
multi-donor trust funds. The findings, interpretations, and conclusions expressed in this paper are entirely those of
the authors and do not necessarily represent the views of the World Bank, its Executive Directors, or the governments
of the countries they represent. This study was registered on the AEA RCT Registry (AEARCTR-0006811). The
Institutional Review Boards at the Health Media Lab, Washington DC and Comit´      e d’Ethique Institutionnel pour la
Recherche en Sciences de la Sant´ e (CEIRES), Burkina Faso approved this project.
   †
     Center for Global Development, 2055 L Street NW, Washington, DC 20036. Email: kajayi@cgdev.org.
   ‡
     The World Bank, 1818 H Street NW, Washington, DC 20433. Email: adao1@worldbank.org.
   §
     The World Bank, 1818 H Street NW, Washington, DC 20433. Email: mkoussoube@worldbank.org.
1    Introduction

Across the world, women are more likely than men to be children’s primary caregivers and there

is increasing awareness that these childcare responsibilities limit women’s economic opportunities

(World Bank, 2012; Delecourt and Fitzpatrick, 2021). At the same time, there is broad consensus

that investing in early childhood development has the potential to reduce poverty and improve key

socio-economic life outcomes (e.g,. Garces, Thomas and Currie, 2002; Gertler et al., 2014; Heckman

and Karapakula, 2019; Devercelli and Beaton-Day, 2020). Experts also agree that children need

a nurturing environment to thrive (World Health Organization, United Nations Children’s Fund

and World Bank Group, 2018). Policymakers and researchers have often considered the objectives

of supporting women’s employment and enhancing child development in isolation, instead of fully

recognizing their interdependence. In principle, providing affordable childcare could address both

of these issues by increasing women’s economic empowerment and stimulating early childhood

development. Yet, there are potential trade-offs that emerge. Childcare is often delivered without

an explicit focus on early childhood education (ECE), thus positive impacts on child development

are not guaranteed. In contrast, ECE interventions are often prohibitively expensive or time-

intensive for caregivers and may therefore impede women’s ability to pursue economic activities

(Mateo Diaz and Rodriguez-Chamussy, 2016). Understanding how to simultaneously achieve these

two objectives remains an understudied issue.

    This paper analyzes a community-based model of integrating childcare centers into an urban

public works program designed to support youth employment in Burkina Faso. We evaluate the

impacts of this accessible childcare provision on women’s employment and other empowerment

outcomes as well as on child development. We use a randomized controlled trial to estimate the

causal effects of childcare, analyzing data from a sample of 1,990 women participating in the six-

month long public works program and their eligible children. The program’s implementation team

identified 36 urban public works sites with the potential to host a childcare center. We then

randomly selected 18 sites to receive the community-based childcare centers and the remaining

18 sites had no additional childcare provision. With this sample size, the study is powered to

detect reasonable effect sizes (Section 3 provides detailed discussion). We conducted a baseline

survey in the early stages of childcare implementation and a follow up survey 14 months later. The



                                                2
public works project focused on enrolling participants from economically and socially disadvantaged

backgrounds, so the childcare intervention could potentially promote equitable childcare access by

benefiting disadvantaged children and their families.

       We find that 25% of eligible women take up the opportunity to use the childcare centers. This

translates to tripling the use of any childcare centers for children aged 0 to 6 over our evaluation

period. These effects are persistent and extend beyond use during the public works program, with

higher rates of childcare usage over the past 24 hours during the follow up survey 14 months later,

which suggests that initial exposure to the childcare centers generates lasting demand. For women,

we find improved employment outcomes, which are most robust for women with children aged 0

to 2 and concentrated in increases in salaried work. Additionally, we find positive effects on self-

reported psychological well-being (although, this result is less robust to attrition correction) and

financial resilience and savings. However, we find no significant changes in women’s participation in

decision-making, their gender attitudes, nor the intrahousehold division of childcare and household

tasks. For children, we find improved development scores, driven by improvements in gross and

fine motor skills. We do not find any improvements, nor deterioration, in children’s language skills.

Our estimates likely constitute a lower bound for the effects of providing access to childcare centers,

as the COVID-19 pandemic reduced households’ exposure to the intervention.1

       Our work makes three contributions to the literature. First, we present rigorous evidence on

the effects of a community-based childcare model in a developing economy. Although substantial

evidence demonstrates that childcare responsibilities limit women’s earnings, how to deliver ac-

cessible care for young children while ensuring that they receive sufficient stimulation remains an

open question. We estimate the impacts of an innovative model of providing affordable access to

childcare by trained providers.2 Several studies have found null or negative impacts of center-based
   1
      In addition to disrupting operation of the childcare centers, the pandemic also reduced employment opportunities.
Analyses from the first round of a nationally representative High Frequency Phone Survey (HFPS) of households
conducted in June 2020 by the Burkina Faso National Institute of Statistics and Demography with technical and
financial assistance of the World Bank found that around 10% of respondents who used to work before the COVID-
19 outbreak were not working at the time of the survey, and that the COVID-19 related economic slowdown has
translated in a reduction of income for most employees, non-farm business owners as well as farmers (Nkengne et al.,
2020).
    2
      Similar mobile childcare interventions exist in other countries, notably the Mobile Creches NGO in India has
been operating mobile childcare centers alongside construction sites in urban areas for the past 50 years, but the
best existing quantitative evidence on effects of access to these facilities comes from a comparison of children who
attended childcare centers for less than a month to those who attended for a full six months, in the same public works
site (Creches, 2012). Without an explicitly-designed comparison group, it is difficult to know whether differences
in observed outcomes truly reflect the causal effects of access to childcare centers or partly result from preexisting



                                                          3
care on children’s development outcomes in other developing economies (for example, Rosero and

Oosterbeek (2011) in Ecuador; Bernal et al. (2019) in Colombia; and Blimpo et al. (Forthcoming)

in Gambia). Other studies find significant positive effects (Martinez, Naudeau and Pereira (2017)

                           opez B´
in Mozambique; Hojman and L´     oo (2019) in Nicaragua; Dean and Jayachandran (2020) in

India; Attanasio et al. (2022) in Brazil; and Bjorvatn et al. (2022) in Uganda). The range in effects

largely depends on the quality of available alternatives and researchers generally find more positive

effects for children from underprivileged households.

    Second, we simultaneously evaluate impacts on women’s empowerment. Although still thin,

there is a growing literature documenting the impacts of childcare access on women’s employment

                                                                              ınez A. and
outcomes in low and middle-income countries (Du and Dong (2013) in China, Mart´

        a (2017) in Chile, Martinez, Naudeau and Pereira (2017) in Mozambique, Hojman and
Perticar´

 opez B´
L´     oo (2019) in Nicaragua, Clark et al. (2019) in Kenya, Halim, Johnson and Perova (forth-

coming) in Indonesia, Attanasio et al. (2022) in Brazil, and Bjorvatn et al. (2022) in Uganda).3

Nevertheless, there is limited evidence of the effects of childcare interventions on women’s autonomy

and psychological well-being, and only few studies examine both child development and women em-

powerment outcomes at the same time (Evans, Jakiela and Knauer, 2021). In related work, Clark

et al. (2019) evaluate the effects of subsidized childcare in Kenya. They similarly find significant

positive effects on women’s employment but no change in women’s participation in household de-

cisions, except for increased involvement in decisions on children’s healthcare. We build on their

study by further extending our focus to include child-level outcomes. To the best of our knowledge,

this paper is the first to simultaneously evaluate impacts on women’s employment, broader em-

powerment, and children’s development. The paper also reports the impacts of childcare provision

on the time use of other caregivers, expanding the understanding of the effects of childcare provi-

sion in low income countries. This is important as it addresses a key question for policymakers,

namely can childcare provision benefit both children and other caregivers, women specifically. In

this paper, we demonstrate the multidimensional effects of this community-based childcare model

and highlight the limitations of childcare provision in enhancing women’s decision-making agency

and transforming intrahousehold gender dynamics.
differences in the types of families who took up the opportunity to use these services versus those that did not.
   3
     Earlier studies have found mixed effects of public preschool and kindergarten on women’s employment in the
United States, including Gelbach (2002); Cascio (2009); Fitzpatrick (2010, 2012).



                                                       4
    Finally, our setting allows us to separately examine two channels through which childcare provi-

sion can impact women’s empowerment – i) directly, by training and employing women to provide

childcare services; and ii) indirectly, by relaxing labor constraints for women who are able to use

the centers to care for their children at a subsidized cost while they pursue economic opportunities

elsewhere. Existing work typically focuses on effects through this second channel without incorpo-

rating impacts through the first channel. We find in our context that both mechanisms appear to be

relevant. Employment effects are strongest for women who work in the childcare centers. However,

significant positive results remain when we exclude these women from our analysis sample. Our

analysis therefore indicates the importance of accounting for both direct and indirect employment

effects when considering the potential impacts of an intervention to provide childcare.



2     Context and Intervention

2.1   Childcare Context

Burkina Faso has three types of early childhood development centers: (i) public preschools; (ii)

private facilities mostly concentrated in urban areas; and (iii) community-based facilities funded by

the community with technical support from the government and/or development partners (World

Bank, 2014). Early childhood education centers for children 0 to 3 years old are rare and con-

centrated in large urban areas, where they are delivered by private actors. Preschool facilities for

children 3 to 5 years old also remain mostly concentrated in urban areas and delivered by private

actors. In 2020, 71% of preschool facilities were private facilities. Community-based facilities rep-

resented around 17% of these facilities, and public facilities around 12%. Enrollment in preschool

facilities for children 3 to 5 years, although still low, has seen a large increase since 2012. In 2020,

5.6% of children 3 to 5 years of age were enrolled in preschool facilities compared to 3.5% of children

                                 ere de l’Education Nationale, 2020).
in this age range in 2012 (Minist`

    Since 2007, Burkina Faso has adopted a strategy for integrated early childhood development

across the Education, Health, Nutrition, Social Protection, Child Protection, Water, Sanitation

and Hygiene sectors, followed in 2012 by a national program of development of basic education,

including preschool education, and a national parental education program in 2013. Although these

regulations and frameworks have allowed some coordination of activities and service delivery across


                                                   5
sectors, a lot remains to be done especially in terms of funding mobilization. For instance, in 2013,

                                                                              eres en charge de
only 0.6% of public expenditures was allocated for preschool education (Minist`

  education et de la Formation, 2017).
l’´


2.2   Youth Employment Context

In 2018, an estimated 37% of young women and 22% of young men aged 16 to 35 in Burkina

Faso were not in school or employed outside the home (authors’ calculations using the Harmonized

Survey on Households Living Standard 2018/19 data). Additionally, 33% of young women and 38%

of young men who were working were still living in poverty as defined by earning less than USD

1.90 a day (International Labour Organization, 2020). Against this backdrop, the Government of

Burkina Faso through its Youth Employment and Skills Project (PEJDC) recruited 10,255 young

women (85%) and men (15%) in 49 communes to implement labor-intensive public works (“PTR-

HIMO”, in French) in all regions of the country from mid-July 2019. The eligibility conditions

were: i) being a Burkina Faso citizen; ii) out of school or never enrolled in school; iii) not a

former beneficiary of the project; and iv) aged 16 to 35. PTR-HIMO participants were selected

through a call for applications for the public work scheme in each commune by the government.

The public work included the construction of bridges and rural roads, the maintenance of urban

roads and administrative spaces, and reforestation. The number of desired workers in each site

varied depending on the size of the site and the type of work, with a minimum requirement of

30% women at each site. Sites were defined as a commune (municipality) for smaller urban centers

or an arrondissement (neighborhood) in the two largest cities, Ouagadougou and Bobo-Dioulasso.

Two rounds of selection procedures were followed: i) registration of interested applicants; and ii)

participation in a public lottery – all applicants drew a piece of paper from a box containing papers

with either yes or no, where drawing yes implied being selected to participate to the program and

no implied not being selected.

   Selected participants (“brigadiers”) for the project began their public works assignment in July

2019. They worked from 8am to 2pm Monday to Saturday and received 37,000 FCFA a month

(approximately 63 USD at the time of the study, roughly the minimum wage in Burkina Faso), for

six months of project participation between July 2019 to February 2020, including a one month




                                                 6
break in August.4


2.3     Childcare Intervention

The childcare intervention was designed to address the constraint that childcare responsibilities

impose on women’s time. Prior to the intervention, female public works participants typically

organized themselves by identifying some among them to look after their children while the others

were conducting the public works. The PEJDC project team developed the “mobile creches” as

an innovative childcare intervention to formalize childcare responsibilities by establishing a quality

source of childcare that would allow women to focus on productive activities. This development

process involved close collaboration with the Ministry of National Education and the Ministry of

Women. The intervention was introduced as a pilot in Manga, one of the communes of a preceding

World Bank-financed public works project.

      The mobile creches childcare intervention was integrated into the public works component of the

PEJDC to follow participants as they move from work site to work site. The intervention includes:

(i) availability during PTR-HIMO working hours of full coverage tents or existing safe spaces that

offer an environment designed specifically for children aged 0 to 6 years, with added protection from

the sun, dust, inclement weather and potential accidents; (ii) two nutritious meals per day;5 (iii)

low-cost toys and learning materials; and (iv) information for parents on childcare and nutrition

based on contents from Burkina’s national program of parental education. The maximum capacity

for each childcare center was 50 children. Satellite creches were also introduced close to the public

works site to allow breastfeeding infants to remain close to their mothers.

      Each childcare center was operated by 7 to 10 public works participants who received a 3-

day training to become “brigadieres assistantes maternelles” (BAMs) and would then attend to

children in the centers. Interested participants at each public works site were asked to volunteer.

Volunteers were screened based on a set of selection criteria, after which 10 volunteers from each

site were selected to receive a 3-day training before starting to operate the childcare centers. BAMs

received the same compensation as other brigadiers and worked the same schedule of hours but at

the childcare centers instead of working in the labor-intensive public works assigned to their peers.
  4
    The average exchange rate in 2019 was around 1 US dollar for 586 West African CFA franc (XOF).
  5
    At least one care provider per creche was trained to cook balanced meals relying mainly on locally grown and
seasonal foods, based on the established meal calendar.



                                                       7
BAMs continued to receive the same 37,000 FCFA monthly wage for each month they worked at the

childcare centers after the public works ended. Centers were supported by supervision visits from

government education and social workers. Parents were asked to provide a nominal contribution of

approximately USD 6 per month for snacks, although this was on a voluntary basis and payment

was not enforced.

   During the public works, the childcare centers officially operated during the same hours as the

public works, i.e., 8am to 2pm. In practice, the centers were open as early as 6am to welcome the

children.


2.4   Theoretical Framework

In the remainder of this section, we briefly outline the theory of change underlying the potential

effects of childcare on our key outcomes of interest: women’s employment, women’s decision-making

autonomy and gender attitudes, and child development.


Effects of childcare on women’s employment. Childcare centers could potentially affect

women’s employment through two main channels. First, by directly creating employment for women

who are hired as childcare center attendants. Second, by relaxing labor constraints for women who

are not directly employed by the centers but are able to use the centers to care for their children

at a subsidized cost while they pursue earning opportunities.

   Standard economic models predict that subsidized childcare provision would impact labor supply

decisions through this second channel by lowering women’s reservation wages (Connelly, 1992). For

women initially working less than the hours of childcare provided, we would expect an increase in

hours worked and earnings. For women already working more than the hours of childcare provided,

the childcare centers could provide an income subsidy and may decrease incentives to work. Given

that the public works program focused on recruiting unemployed youth, we assume that most

women in this population were initially underemployed. The main impacts of access to childcare

centers would therefore be to allow women to increase the amount of time they spend on economic

activities. We also anticipate increased productivity (or quality of time spent on income-generating

activities) because women would be able to focus on income-generating work, while children are

in the childcare centers. Finally, we anticipate that women would have improved psychological



                                                 8
well-being due to reduced stress about multitasking while caring for children and more mental

bandwidth. Thus, we predict increases in time spent on income generating activities and increased

earnings as primary outcomes. In addition, we anticipate improvements in self-reported well-being

and financial resilience as secondary outcomes due to improvements in women’s employment and

earnings.


Effects of childcare on women’s decision-making autonomy, gender attitudes, and in-

trahousehold dynamics. Our anticipated effects on these dimensions of women’s empowerment

are theoretically ambiguous. Under a collective model of household decision-making, affordable

childcare could increase women’s employment and relative earnings, thereby increasing women’s

                                                                e, 2021). Alternatively, the
intrahousehold bargaining power and empowerment (Hiller and Tour´

theory of gender deviance neutralization predicts that women may feel compelled to perform stereo-

typical gender roles within the home in order to neutralize the act of ceding some of their childcare

responsibilities to a childcare center (Atkinson and Boles, 1984; Greenstein, 2000). Thus, any po-

tential improvements in women’s empowerment resulting from increases in bargaining power might

be offset by women engaging in behaviors to emphasize their traditional female roles and conform

to prevailing gender norms. The overall effect of childcare on women’s decision-making autonomy,

gender attitudes, and intrahousehold dynamics in this context is therefore an empirical question.6


Effects of childcare on child development. We hypothesize that children of public works

participants in sites with the childcare intervention will spend more time in childcare centers. The

resulting impact on child development depends on the relative quality of the alternative childcare

options. Given that caregivers in these community-based childcare centers are trained and super-

vised, we anticipate that there could be improved early childhood development for enrolled children

due to increased stimulation by trained childcare providers and increased access to age-appropriate

toys and learning materials. Conversely, if children would instead have stayed at home with an

engaged caregiver, then it is possible that receiving individualized attention at home would be more

beneficial than the care received in a group setting.
   6
    In contrast to predominant American expectations on maternal practice, anthropological research on motherhood
ideals among urban Asante market traders in neighboring Ghana (Clark, 1999) and an ongoing complementary
qualitative study with a sample of urban mothers in Burkina Faso suggest that parenting norms in this context
emphasize meeting children’s resource needs, rather than necessitating mothers spend time with them.



                                                       9
3        Methodology

3.1       Research Design

Our impact evaluation focuses on 36 urban public works sites in 17 communes of the PEJDC

intervention area in 9 regions of Burkina Faso (Boucle du Mouhoun, Centre, Centre-East, Centre-

North, Centre-West, Centre-South, Hauts Bassins, North, Plateau Central). The 36 sites were

selected based on their potential to support a childcare center. The number of participants at each

site ranged from 100 to 500 and the proportion of women ranged from to 58% to 96%.

        We used a computer program to randomly assign the 36 urban public works sites in the research

sample to a treatment group (18 sites) with implementation of community-based childcare centers

and a control group (18 sites) with no childcare centers. To increase the similarity of the 2 groups,

the randomization was stratified across 10 blocks based on the geographical location and number

of brigadiers assigned to each site. The sites were defined by communes (municipalities) in smaller

urban areas and by arrondissements (neighborhoods) in the two larger cities of Ouagadougou and

Bobo-Dioulasso. Figure 1 illustrates the spatial distribution of our study sites.


3.2       Sampling

Our sample comes from the register of public works participants in the 36 study sites. We began

with a screening survey to restrict our sampling frame to women who were primary caregivers for a

child aged 0 to 6 (because the centers were open to children below the age of 7). Among participants

meeting these eligibility criteria, we randomly selected 2,160 individuals. In each of the 36 sites,

the size of the sample was proportional to the number of eligible participants in that site. The

study sample therefore included 2,160 households, with data collected on the 2,160 female public

works participants, their spouses or partners if co-residing, and any children aged 0-6 in their care,

for a total of up to 4,320 individual interviews and a minimum of 2,160 child assessments (Figure

2 presents our sampling strategy).7
    7
    With this sample of 36 sites (18 per treatment group), an average of 60 participants per site, power of 0.8,
significance level of 0.05, and an estimated intracluster correlation of 0.1, the study was designed to detect a minimum
effect of 0.2 standard deviations.




                                                          10
3.3    Data

In December 2019, we conducted a baseline survey of female respondents, any cohabiting partners,

and children aged 0 to 6, through face-to-face interviews. The baseline questionnaire asked women

participants about their household composition, demographic background, economic activities of

the respondent and any co-resident partner, consumption, dwelling and assets, knowledge and

practices on child health and nutrition, input into productive decisions of the respondent and

any co-resident partner, freedom of movement, acceptance towards domestic violence, sharing of

housework, happiness, mental health,8 and time use. We directly interviewed the woman and

cohabiting partner. We also asked about the primary caregiver for each child for each hour over

the past 24 hours. We measured child development using the Development Milestone Checklist

(DMC III), which was adapted and validated in Burkina Faso for children aged 0 to 8 (Prado et al.,

2014). The DMC III includes questions on gross motor, fine motor, and language skills. Due to

the time demands of administering the DMC, we randomly selected one child aged 0 to 2 and one

child aged 3 to 5 from each household to include in the sample, instead of administering it to all

children in each household.9 The baseline sample includes 2,150 households and 3,126 children.

The baseline survey occurred approximately one month after the first childcare centers opened, we

therefore control for baseline outcomes in our econometric specifications.

    In February 2021, we conducted an endline survey by phone, due to COVID-19 social distancing

requirements. We interviewed each female respondent about herself, her partner, and her children.

We were able to contact 1,990 respondents from our baseline sample (a 92.6% tracking rate). We

have some differential attrition by treatment status with a slightly higher attrition rate in the

treatment group (8.6%) compared to the control group (6.4%). We address this in our analysis.

Figure 3 indicates the timeline of implementation and data collection activities. All 18 sites that

were randomized to receive a childcare center ended up opening a childcare center, so we have

perfect compliance with assignment to open a childcare center.
    8
      We measure depressive symptoms using the Center for Epidemiological Studies Depression Scale Revised Short
Form (CESD-R-10) (Miller, Anton and Townson, 2008; Radloff, 1977). The CESD, a screening test for depression
and depressive disorder, is one of the most used instrument to measure depression in ECD studies in low and middle
income countries (Evans et al., 2022).
    9
      We selected children up to age 5, even though children up to age 6 could attend the childcare centers, so that
we could ensure that these children would still be eligible to attend the childcare centers at the time of the follow up
survey. At baseline, two children were selected regardless of the age categories; i.e. two children aged 3 to 5 could be
selected if there were no children aged 0 to 2. At endline, we collected data on children sampled at baseline, for only
one randomly selected child in each age category.


                                                          11
   Our study period was interrupted by the onset of the COVID-19 pandemic. The childcare

centers began operating in November 2019 and were intended to run continuously for a year,

but instead closed in March 2020 due to COVID-19. They reopened in October 2020, with the

support of the Burkina Faso Ministry of Women and Ministry of Education, the mairies (municipal

governments), and the World Bank, which each provided funding, in-kind resources, or technical

assistance to operate the childcare centers after the closing of the PEJDC project in February 2020.

Participating households therefore had a maximum of eight months of potential exposure to the

childcare centers (four months before the COVID-19 closure and then four months following the

reopening and before our follow-up phone survey). Importantly, the second phase of operation

occurred after program participants had completed their public works assignment, which ended

in February 2020. This timing allows us to observe usage of the childcare centers and impacts

on economic outcomes in a period during which program participants were no longer guaranteed

employment through the public works program. Additionally, the childcare centers operated mainly

from 8am to 2pm to coincide with the public works schedule during the first phase of operation,

but they were optionally open to parents for a full workday of care when they reopened in October

2020.


3.4     Empirical Strategy

We use the following analysis of covariance (ANCOVA) specification to estimate Intent to Treat

(ITT) effects:


                    Yist = α + βT reatments + γYist−1 + X is + W s + εist                        (1)


where Yist is the outcome for individual i from public works site s at time t. T reatments is an

indicator for whether site s was one in which a childcare center was established under the project.

Yist−1 is the baseline value of the outcome measure for individual i. X is is a vector of control

variables. Household-level controls include household size, number of children aged 6 or under,

number of children aged 7 to 15, number of economically active household members, head is female,

head is ethnically Mossi, head age, head is in a monogamous marriage (to capture differences in




                                                12
intra-household dynamics and related differences in women’s and children’s welfare outcomes)10 ,

head is not formally educated, head is working. Brigadiere controls include age and indicators

for being in a monogamous marriage and not formally educated. Child-level controls include age,

indicators for being female, having a biological mother in the household, and being the child of the

household head. W s is a vector of binary variables representing the stratification bin from which

site s is drawn. Finally, εis is an idiosyncratic error, assumed to be independent across sites but

allowed to be correlated within a site (i.e., we cluster standard errors at the public work site level).

To account for outliers, we winsorize continuous variables at the 99th percentile.

      We also estimate effects of the treatment-on-the-treated (TOT) for key outcomes.


                         Yist = α + β Childcares + γYist−1 + X is + W s + εist                      (2)


where we use a two-stage least squares procedure, with a child ever having used a childcare center

as the treatment indicator in the first stage.

      To address the possibility of spurious results due to multiple hypothesis testing, and to maxi-

mize the power of our statistical tests, we combine outcome measures covering similar domains into

a summary index (Kling, Liebman and Katz, 2007). We focus on five domains– women’s employ-

ment, women’s decision-making and gender attitudes, women’s psychological well-being, women’s

financial resilience and savings, and children’s development. For each index, we normalize each

outcome measure relative to the control group mean and standard deviation. We convert the sign

of individual measures where necessary so that higher scores indicate improved outcomes and then

calculate an equally weighted average of the normalized components of each domain. For obser-

vations with missing values for a given indicator, we take the mean of the other variables in that

domain. For observations missing data on all measures in a given domain, we set the summary

index to missing. As an alternative way to deal with missing observations when constructing the

summary indices, we also excluded observations with any missing values for individual indicators

from our analysis.

      In addition to our main specifications, we conduct two sets of robustness checks. First, to

address differential attrition between the treatment and control groups, we estimate Lee bounds
 10
      See for instance Akresh, Chen and Moore (2016); Barr et al. (2019); Rossi (2018).




                                                         13
(Lee, 2002). Second, to account for the small number of clusters in each of our treatment groups,

we estimate alternative p-values using randomization inference (Athey and Imbens, 2017).

         We estimate separate effects for our full sample and for the subsample of households with a

child aged 0 to 2 as well as estimating separate child-level effects for children within this age-range

to examine whether childcare centers have differential impacts for women with young children.11



4         Results

4.1        Baseline Characteristics

Table 1 presents balance on baseline women’s and household characteristics for the full sample.

Women in our sample are aged 31 on average and two thirds are in a monogamous marriage.

Approximately 46% of respondents have no formal education. In comparison, 64.5% of women in

Burkina Faso and 48.7% of women living in urban areas have no formal education according to

the latest Harmonized Survey on Household Living Conditions, conducted in 2018/2019. Thus,

education levels in our urban sample are similar to those for women in the urban population in

general. Almost 98% of women report having worked for at least one hour in the last 30 days, which

is consistent with respondents’ participation in the public work activities. On average, women in

our sample worked 33 hours per week in the last month. Time use data reveal that in the last

24 hours, women in our sample spent on average of 6.7 hours caring for their children, 2.5 hours

on domestic chores and 3.4 hours doing paid work. Women in the control group spent more time

working per week (37 hours for women in the control group, compared to 28 hours on average

for women in the treatment group, p <0.05), and more time on domestic chores (2.6 hours in the

control group, compared to 2.4 hours in the treatment group, p <0.05).

         Women earn an average monthly total income of approximately 19,000 FCFA at baseline. This

is lower than the 37,000 FCFA monthly payment from the public works, which reflects program

implementation challenges. Public works programs aim to provide temporary employment opportu-

nities for the most vulnerable, but by the time of the baseline survey (four months into the program)

public works participants were yet to receive their allowances because of persistent payment delays

during the study period.
    11
   We do not have enough statistical power to detect heterogeneous treatment effects by age of the youngest child.
We therefore present these results as suggestive evidence.


                                                       14
       Table 2 presents balance on baseline child-level characteristics for the full sample. About 49%

of the children in our sample are female, 94% are living with their biological mothers and 86%

are the children of the heads of households. On average, children are 3 years old, with children

in the control group being slightly older than children in the treatment group (2.97 years old vs

3.02, p <0.05). The average child development score in our sample is 85 points out of a total of 152

points, with an average gross motor score of 36 points (ranging from a minimum of 0 to a maximum

of 52), an average fine motor score of 20 points (ranging from a minimum of 0 to a maximum of

54), and an average language score of 29 points (ranging from a minimum of 0 to a maximum of

46).

       Although our tracked sample, used in the analysis, mainly appears to be balanced on baseline

characteristics, we find statistically significant differences at the 5% level in 4 of the 22 baseline

household characteristics and 5 of the 18 baseline child characteristics we test, including some key

employment and time use variables.12 To create a scale-free measure of overlap in the distribution

of covariates in the two treatment groups, we estimate normalized differences following Imbens and

Rubin (2015). The only variable with a normalized difference above their suggested cutoff of 0.25

standard deviations is total hours worked per week, with a normalized difference of 0.27. Given

that the public work sites were randomly assigned to receive a childcare center, we view these

differences as resulting from chance. We address this imbalance by using an ANCOVA model as

our preferred specification in our regression analysis.


4.2      Use of Childcare Centers

At baseline, only a very small proportion of households (2%) reported having any child in the

household attend a childcare center in the last 24 hours (Table 1). At the follow-up survey, 24% of

respondents reported having at least one child attend a childcare center in the last 12 months (12%

of respondents in the control group versus 37% of respondents in the treatment group). Asked

about the types of childcare centers attended, 12% of respondents reported having at least one

child attend a mobile childcare center (less than 1% of respondents in the control group versus 25%

of respondents in the treatment group).
  12
    We also test for balance in the subsample of households with children 0-2 years old and for children 0-2 years old,
using the same baseline variables. Appendix Tables A1 and A2 present baseline balance for the sample of households
with a child aged 0 to 2 and for children within this age-range. We find similar baseline differences.



                                                          15
      Table 3 shows the effects of being assigned to receive a childcare center on the use of childcare

for the full sample and for the subsample of households with children under two years of age.

Participants in the treatment group are 11.7 percentage points more likely to have a child in the

household who attended a childcare center in the past 24 hours compared to a control group

rate of 6 percent. This translates to a tripling in current use of childcare centers at the time of

the follow up survey, 14 months after the baseline. We see similar effects when we look at the

likelihood of any childcare center use over the evaluation period. Participants in the treatment

group are 24 percentage points more likely to have had at least one child attend a mobile creche

in the past 12 months. These results demonstrate persistent increases in childcare center take

up that extend beyond use during the public works program, suggesting that initial exposure to

the childcare centers generated lasting demand once the centers reopened following the 8-month

COVID-19 related closures.


4.3     Effects on Summary Measures

Table 4 and Figure 4 present treatment effects on our summary measures of outcomes for the full

sample and for the subsample of households with children under two years of age. The first row of

Column 1 indicates a positive and significant impact on our summary employment measure.13 The

ITT estimates show an increase in our women’s employment measure of 0.08 SD and for the full

sample and 0.09 SD for the subsample of respondents in households with children under two years

of age respectively. We also see significant and relatively larger positive impacts of being assigned to

receive a childcare center on our summary measures of women’s psychological well-being (column

3),14 women’s financial resilience and savings (column 4),15 and on child development (column

5),16 with an increase of 0.175 SD. However, our results indicate no significant impacts on women’s

involvement in decision-making or gender attitudes (column 2).17 None of the estimated effects for

the subsample of women with children under age 2 are significantly different from the results we

find for the full sample, suggesting that childcare had broadly similar impacts for all women in the

sample at an aggregate level. However, we do not have enough statistical power to definitively rule
 13
    Aggregate   index   of   the employment outcomes listed in Tables 5, 6 and 7.
 14
    Aggregate   index   of   outcomes listed in Table 11.
 15
    Aggregate   index   of   outcomes listed in Table 12.
 16
    Aggregate   index   of   outcomes listed in Table 13.
 17
    Aggregate   index   of   outcomes listed in Tables 8 and 9.



                                                             16
out the possibility of there being heterogeneous effects.

   The subsequent rows in each panel of Table 4 present results from our robustness checks.

The Lee bounds on the employment index remain significantly positive and the p -values from

randomization inference indicate significant positive effects at the 10% level for households with

children under age 2, but not for the full sample. The Lee lower bound estimates of mental health

impacts are statistically insignificant and the randomization inference p -values also rise above the

10 percent level. Both the estimated positive impacts on women’s financial outcomes and child

development scores persist across all robustness checks, with the exception of a randomization

inference p -value of 0.12 for women’s financial outcomes in the full sample. Altogether, these results

provide strong evidence that the childcare intervention had positive impacts on child development,

along with indications of positive effects on women’s employment, and insignificant impacts on

women’s decision-making autonomy and gender attitudes. We find similar results in our alternative

approach which excludes observations with any missing data for outcomes included in our summary

indices (these results are reported in Appendix Table A7).


4.4   Effects on Specific Outcomes

In the remaining sections, we discuss effects on specific outcomes. The estimated magnitudes

of impacts on individual outcomes are more straightforward to interpret and provide additional

insights on the impacts of childcare provision, with the caveat that we do not correct these individual

estimates for multiple hypothesis testing. We therefore view these results as offering a suggestive

indication of patterns and take our effects on summary measures as our preferred estimates of

aggregate impacts.


Women’s employment. Tables 5, 6 and 7 present the results on women’s employment at follow-

up for the full sample and the subsample of households with children under two years of age. While

we do not find any statically significant impact on the likelihood of working in our full sample,

Table 5 shows a statistically significant impact at the 5% level of the assignment to receive a

childcare center on income from salaried work. Moreover, women caring for children under two

years of age spend more time in salaried employment (significant at the 10% level) and have a

larger increase in income from salaried work. We do not find any significant impacts on other types



                                                  17
of employment, i.e., agricultural employment and self-employment in non-agricultural activities.18

To further examine impacts on employment outcomes, we plot the distribution of the women’s

employment index and the distribution of salaried income for each treatment group in Figures 5

and 6. These figures clearly indicate a rightward shift in the distribution of women’s employment

outcomes.


Women’s decision-making autonomy, gender attitudes, and intrahousehold dynamics.

We do not find any statistically significant impact of the intervention on women’s decision-making

in the household, sharing of domestic work, freedom of movement, or gender attitudes (Table 8).

For women caring for children under two years of age, there is a marginally significant increase in

the likelihood that the respondent’s partner sometimes takes care of children (Table 9). However,

we find no significant changes in time use, with the exception of a 20 percent increase in time spent

doing paid work over the past 24 hours, compared to a control group mean of 3.4 hours (Table 10).


Women’s psychological well-being. Turning to women’s psychological well-being, we find a

small beneficial impact of the intervention, reducing the likelihood of experiencing depression and

self-reported unhappiness (significant at the 5% level) for the full sample (Table 11). We find similar

effects for women in households with children under two years of age. These results are consistent

with work from Bossuroy et al. (2022), finding that a multi-faceted poverty alleviation program

in Niger (neighboring Burkina Faso) improved mental health by 0.13 to 0.23 standard deviations,

particularly when it included a psychosocial component.


Women’s financial outcomes.                We use an OLS specification to estimate impacts on financial

outcomes since we only collected this information in the endline survey. Almost 59% of women in

the control group had saved any money during the past 12 months and 40% were able to pay FCFA

20,000 (approximately USD 34) in the case of an emergency (Table 12). We find a 13% percent

increase in women’s likelihood of having saved in the last 12 months (significant at the 5% level)

and a 25% increase in women’s reported capacity to mobilize financial resources in an emergency.
  18
     We collected data on extensive margin labor market participation for all non-agricultural non-salaried activity.
However, due to a mistake in the skip patterns in the programmed version of the survey questionnaire, we only
collected information on hours worked in the last month and total monthly income for non-agricultural activity in a
household enterprise, therefore excluding non-agricultural self-employment income from non-household enterprises.
Thus, while our results in column 1 of Table 7 are unaffected, results in columns 2, 3, and 4 unfortunately exclude
self-employment activities and could therefore understate any treatment effects.


                                                         18
Child development. Table 13 reports the results on children development outcomes. We find

that the intervention led to an improvement in both gross and fine motor scores for children in our

full sample and in the sample of children under two years of age. However, we do not find any

impact of the intervention on children’s language scores, although the coefficients are all positive.

Altogether, this indicates that the childcare centers stimulated the development of children’s motor

skills without compromising language development.


4.5   Robustness Checks

Table A3 presents results from robustness checks to address differential attrition as well as adjusted

p-values from randomization inference and wild bootstrap clustered standard errors to address the

small number of cluster (public work sites) in the sample. The strongest results are the increase

in women’s time in hours doing paid work during the last 24 hours and total monthly income for

women with children aged 0 to 2, as well as improvements in child development outcomes for the

full sample and for children aged 0 to 2. The remaining results survive the Lee bounds but lose

significance with the corrections for the small number of clusters.


4.6   Analysis of Mechanisms

Turning to understanding the potential mechanisms, we find no increases in partners’ involvement

in childcare and negative but statistically insignificant effects on the duration of time for which

the mother was the primary caregiver for a child over the past 24 hours (Table 14). Our ITT

estimates indicate an average increase of approximately 20 additional minutes in a childcare center

over the past 24 hours and a decrease of 15 minutes spent by mothers as primary caregivers. These

averages likely mask substantial heterogeneity, with some heavier childcare users in the sample.

Altogether, our results suggest that the main mechanism behind the positive employment impacts

we find could be improvements in the quality of time spent on economic activities and reductions

in childcare-related distractions.

   Our sample size limits our ability to precisely decompose the employment effects into those

coming from the direct employment of childcare center attendants (BAMs) and indirect employment

resulting from relaxation of labor supply constraints for women who use the childcare centers but

work elsewhere. As suggestive evidence on this, we estimated an additional set of results excluding


                                                 19
the 69 BAMs (7% of the treatment group). We present these results in the Appendix (Tables A8

and A9). The magnitudes of our estimates are smaller but the effects on our summary outcome

measures remain statistically significant. Appendix Figure A1 indicates that monthly incomes for

the BAMs were higher than for other public works participants in treatment sites, peaking in the

range of 30,000 CFA, which is close to the 37,000 CFA monthly payment for BAMs working in the

childcare centers. Figure A2 similarly indicates a rightward shift in the employment index. Finally,

Figures A3 and A4 indicate that the income distribution and employment index in the treatment

group substantially converge towards those of the control group once we remove the BAMs from

our analysis sample, suggesting that the direct employment channel was important in this context,

albeit not the only mechanism at work.



5    Discussion

There are certain limitations of the research design worth noting. First, we cannot measure the

effects of access to childcare on the decision to participate in the public works program because the

childcare intervention was introduced several months after participants had already been recruited

and commenced their work assignments. Second, the World Bank project team provided increased

monitoring and safeguards to the public works program in sites with childcare centers, so we cannot

perfectly isolate the effects of access to childcare from the effects of the additional accompanying

measures provided to treatment sites. Third, public works participants in both the treatment and

control sites received parental education training from social workers, which could have improved

their ability to stimulate their children and could reduce the differences in early childhood develop-

ment outcomes observed across the two groups. Finally, the COVID-19 pandemic reduced childcare

exposure to eight months, instead of one year as planned. There is evidence from nationally rep-

resentative data collected at the early stages of the COVID-19 pandemic in Burkina Faso that the

pandemic has reduced economic opportunities, which may limit the employment and earning effects

of the intervention. Our observed impacts are therefore suggestive of potentially larger impacts of

sustained childcare provision in a more favorable economic climate. Conversely, follow-up child as-

sessments were conducted by phone (not direct observation) due to social distancing protocols, so

there is some potential for bias from self-reports. Altogether, these results provide strong indication



                                                  20
of a promising model of childcare provision, with scope for future work to offer further validation.

       The monthly cost of operating the childcare centers was USD 16.6 per child if used at the full

capacity of 50 children per center. This is lower than our estimated treatment on the treated (TOT)

effect on monthly earnings, with increases of FCFA 13,253.96 (approximately USD 23) for the full

sample and FCFA 14,447.78 (approximately USD 25) for the sample of women with a child under

the age of two. In practice, the childcare centers typically operated below maximum capacity and

had 33 children on average, so the operating costs per child were higher (at USD 25.2). Nonetheless,

the intervention potentially had broader impacts beyond the monetary benefits of income increases

if the training provided to the childcare center attendants generated positive development outcomes

for other children in their communities. Although operating costs are comparable to the earnings

increases observed for program participants within our treated sample, these are higher than costs

                                    ınez A. and Perticar´
in other settings. For example, Mart´                   a (2017) report that community-based

preschools in Mozambique cost USD 3.09 per child per month and increased child development

scores by 0.33SD (implying they were three times as cost effective as the Burkina Faso childcare

intervention, which had an estimated child development TOT effect of 0.88SD with a cost of USD

25.2 per child). The key differences are higher provider to child ratios, with the Mozambique

preschools and other community-based models typically having a ratio of 1:15, instead of 1:5 ratio

of the mobile childcare centers. Relatedly, the Burkina Faso childcare centers catered to children

aged 0 to 6 whereas the Mozambique preschools and other similar models typically focus on children

aged 3 to 5. Almost 60% of women in our sample had at least one child aged between 0 to 2. Given

that the take-up rates and treatment effects for children in this age group were similar to those

for the older age-group, the added costs of caring for younger children appear to be driving the

decreased cost-effectiveness. Thus, the program’s expansive scope of serving the youngest children

(typically excluded from childcare programs) comes with the trade-off of increasing costs, which is

an important factor to consider in the prioritization of program objectives.19
  19
     This cost-effectiveness comparison comes with the caveat that attendance at the mobile childcare centers was
irregular for many children. We unfortunately cannot link individual attendance records to survey data for children
to adjust our cost-effectiveness estimates.




                                                        21
6    Conclusion

This study provides new evidence on the impacts of childcare centers on women’s economic em-

powerment and children’s development in a low-income setting. We demonstrate the potential for

a community-based intervention to improve both women’s and children’s well-being. Although

the childcare centers were envisioned to operate for a full year, the intervention exposure was re-

duced to eight months due to the onset of the COVID-19 pandemic. Nonetheless, we find strong

indications of positive impacts on women’s employment (both through direct employment of child-

care providers as well as through an indirect channel of relaxing constraints on women’s time).

Furthermore, we find positive effects on women’s financial outcomes and improvements in child

development. Despite these positive impacts, we find no increases in decision-making autonomy

and gender attitudes, or changes in the intrahousehold division of childcare. While it is possible

that impacts on women’s agency accrue over time and might be visible in a longer run follow up,

these intermediate results suggest the potential value of a complementary gender-transformative

intervention to address intrahousehold dynamics and social norms, in order to fully enhance the

multiple dimensions of women’s empowerment.




                                                22
References
Akresh, Richard, Joyce J. Chen, and Charity T. Moore. 2016. “Altruism, Cooperation,
 and Efficiency: Agricultural Production in Polygynous Households.” Economic Development and
 Cultural Change, 64(4): 661–696.
Athey, Susan, and Guido W. Imbens. 2017. “The econometrics of randomized experiments.”
 In Handbook of Field Experiments. Vol. 1 of Handbook of Economic Field Experiments, , ed.
 Abhijit Vinayak Banerjee and Esther Duflo, 73–140. North-Holland.
Atkinson, Maxine P., and Jacqueline Boles. 1984. “WASP (Wives as Senior Partners).”
 Journal of Marriage and Family, 46(4): 861–870.
Attanasio, Orazio, Ricardo Paes de Barros, Pedro Carneiro, K. David Evans, Lycia
 Lima, Pedro Olinto, and Norbert Schady. 2022. “Public Childcare, Labor Market Outcomes
 of Caregivers, and Child Development: Experimental Evidence from Brazil.” unpublished.
Barr, Abigail, Marleen Dekker, Wendy Janssens, Bereket Kebede, and Berber
 Kramer. 2019. “Cooperation in Polygynous Households.” American Economic Journal: Ap-
 plied Economics, 11(2): 266–83.
Bernal, Raquel, Orazio Attanasio, Ximena Pea, and Marcos Vera-Hernndez. 2019. “The
 effects of the transition from home-based childcare to childcare centers on children’s health and
 development in Colombia.” Early Childhood Research Quarterly, 47: 418–431.
Bjorvatn, Kjetil, Denise Ferris, Selim Gulesci, Arne Nasgowitz, Vincent Somville, and
  Lore Vandewalle. 2022. “Childcare, labor supply, and business development: Experimental
  evidence from Uganda.” C.E.P.R. Discussion Papers CEPR Discussion Papers 17243.
Blimpo, Moussa Pouguinimpo, Pedro Manuel Amaro Da Costa Luz Carneiro, Pamela
  Jervis, and Todd Pugatch. Forthcoming. “Improving access and quality in early childhood
  development programs: Experimental evidence from the Gambia.” Economic Development and
  Cultural Change.
Bossuroy, Thomas, Markus Goldstein, Bassirou Karimou, Dean Karlan, Harounan
 Kazianga, William Parient’e, Patrick Premand, Catherine C. Thomas, Christopher
 Udry, Julia Vaillant, and Kelsey A. Wright. 2022. “Tackling psychosocial and capital
 constraints to alleviate poverty.” Nature, 605(7909): 291–297.
Cascio, Elizabeth U. 2009. “Maternal Labor Supply and the Introduction of Kindergartens into
 American Public Schools.” Journal of Human Resources, 44(1): 140–170.
Clark, Gracia. 1999. “Mothering, Work, and Gender in Urban Asante Ideology and Practice.”
  American Anthropologist, 101(4): 717–729.
Clark, Shelley, Caroline W. Kabiru, Sonia Laszlo, and Stella Muthuri. 2019. “The impact
  of childcare on poor urban women’s economic empowerment in Africa.” Demography, 56(4): 1247–
  1272.
Connelly, Rachel. 1992. “The Effect of Child Care Costs on Married Women’s Labor Force
 Participation.” The Review of Economics and Statistics, 74(1): 83–90.
Creches, Mobile. 2012. Accelerating Learning report - An Evaluation of the Balwadi Programme
 of Mobile Creches.

                                               23
Dean, Joshua T, and Seema Jayachandran. 2020. “Attending kindergarten improves cognitive
 development in India, but all kindergartens are not equal.”

              ene, and Anne Fitzpatrick. 2021. “Childcare matters: Female business owners
Delecourt, Sol`
 and the baby-profit gap.” Management Science, 67(7): 4455–4474.

Devercelli, Amanda E, and Frances Beaton-Day. 2020. “Better Jobs and Brighter Futures:
 Investing in Childcare to Build Human Capital.” Washington, DC: World Bank.

Du, Fenglian, and Xiao-yuan Dong. 2013. “Women’s Employment and Child Care Choices
 in Urban China during the Economic Transition.” Economic Development and Cultural Change,
 62(1): 131–155.

Evans, David K., Pamela Jakiela, and Heather A. Knauer. 2021. “The impact of early
 childhood interventions on mothers.” Science, 372(6544): 794–796.

Evans, David K., Pamela Jakiela, Heather A. Knauer, and Amina Mendez Acosta.
 2022. “Tools to measure the impact of early childhood development interventions on maternal
 mental health in low- and middle-income countries.” SSM - Mental Health, 2: 100127.

Fitzpatrick, Maria Donovan. 2010. “Preschoolers Enrolled and Mothers at Work? The Effects
  of Universal Prekindergarten.” Journal of Labor Economics, 28(1): 51–85.

Fitzpatrick, Maria Donovan. 2012. “Revising Our Thinking About the Relationship Between
  Maternal Labor Supply and Preschool.” Journal of Human Resources, 47(3): 583–612.

Garces, Eliana, Duncan Thomas, and Janet Currie. 2002. “Longer-Term Effects of Head
 Start.” American Economic Review, 92(4): 999–1012.

Gelbach, Jonah B. 2002. “Public Schooling for Young Children and Maternal Labor Supply.”
 American Economic Review, 92(1): 307–322.

Gertler, Paul, James Heckman, Rodrigo Pinto, Arianna Zanolini, Christel Vermeersch,
 Susan Walker, Susan M. Chang, and Sally Grantham-McGregor. 2014. “Labor market
 returns to an early childhood stimulation intervention in Jamaica.” Science, 344(6187): 998–1001.

Greenstein, Theodore N. 2000. “Economic dependence, gender, and the division of labor in the
 home: A replication and extension.” Journal of Marriage and family, 62(2): 322–335.

Halim, Daniel, Hillary C. Johnson, and Elizaveta Perova. forthcoming. “Preschool Avail-
 ability and Women’s Employment: Evidence from Indonesia.” Economic Development and Cul-
 tural Change.

Heckman, James J, and Ganesh Karapakula. 2019. “Intergenerational and Intragenerational
 Externalities of the Perry Preschool Project.” National Bureau of Economic Research Working
 Paper 25889.

Hiller, Victor, and Nouhoum Tour´  e. 2021. “Endogenous gender power: The two facets of
 empowerment.” Journal of Development Economics, 149: 102596.

Hojman, Andr´  es, and Florencia L´ opez B´oo. 2019. “Cost-effective public daycare in a low-
 income economy benefits children and mothers.” Inter-American Development Bank IDB Publi-
 cations (Working Papers) 9786.


                                               24
Imbens, Guido W., and Donald B. Rubin. 2015. Causal Inference for Statistics, Social, and
  Biomedical Sciences: An Introduction. Cambridge University Press.

International Labour Organization. 2020. “ILOSTAT database.” https: // ilostat. ilo.
  org/ data/ , Accessed: 2022-04-22.

Kling, Jeffrey R, Jeffrey B Liebman, and Lawrence F Katz. 2007. “Experimental Analysis
 of Neighborhood Effects.” Econometrica, 75(1): 83–119.

Lee, David S. 2002. “Trimming for Bounds on Treatment Effects with Missing Outcomes.” Na-
  tional Bureau of Economic Research Working Paper 277.

    ınez A., Claudia, and Marcela Perticar´
Mart´                                          a. 2017. “Childcare effects on maternal employ-
 ment: Evidence from Chile.” Journal of Development Economics, 126: 127–137.

Martinez, Sebastian, Sophie Naudeau, and Vitor Pereira. 2017. “Preschool and child devel-
 opment under extreme poverty: Evidence from a randomized experiment in rural Mozambique.”
 The World Bank Policy Research Working Paper Series 8290.

Mateo Diaz, Mercedes, and Lourdes Rodriguez-Chamussy. 2016. Cashing in on Educa-
 tion: Women, Childcare and Prosperity in Latin America and the Caribbean. Latin American
 Development Forum.

Miller, William, Hubert Anton, and Andrea Townson. 2008. “Measurement properties of
 the CESD scale among individuals with spinal cord injury.” Spinal cord, 46: 287–92.

      ere de l’Education Nationale, de l’Alphab´
Minist`                                               etisation et de la Promotion des
 Langues nationales. 2020. “Annuaire Statistique de l’Education Prescolaire 2019/2020.”

      eres en charge de l’´
Minist`                     education et de la Formation, UNICEF, UNESCO Pˆ               ole de
 Dakar de IIPE. 2017. Burkina Faso: rapport d’´                 eme ´
                                                    etat du syst`   educatif national du Burkina
                                                        eforme du continuum d’´
 Faso: pour une politique nouvelle dans le cadre de la r´                      education de base.

Nkengne, Clarence Tsimpo, Marco Tiberti, Nga Thi Viet Nguyen, Prospere R.
 Backiny-Yetna, Zakaria Koncobo, Lucien Belemkoabga, and Adama Tiendrebeogo.
 2020. “COVID-19 Impact Monitoring at the Household Level: Burkina Faso, Brief No. 1.”

Prado, Elizabeth L, Amina A Abubakar, Souheila Abbeddou, Elizabeth Y Jimenez,
 Jrme W Som, and Jean-Bosco Oudraogo. 2014. “Extending the Developmental Milestones
 Checklist for use in a different context in Sub-Saharan Africa.” Acta Paediatrica, 103(4): 447–454.

Radloff, Lenore S. 1977. “The CES-D Scale: A self-report depression scale for research in the
 general population.” Applied psychological measurement, 1(3).

Rosero, Jose, and Hessel Oosterbeek. 2011. “Trade-offs between different early childhood in-
 terventions: Evidence from Ecuador.” Tinbergen Institute Tinbergen Institute Discussion Papers
 11-102/3.

Rossi, Pauline. 2018. “Strategic Choices in Polygamous Households: Theory and Evidence from
 Senegal.” The Review of Economic Studies, 86(3): 1332–1370.

World Bank. 2012. World Development Report 2012. Washington, DC: World Bank.

World Bank. 2014. “Burkina Faso Early Childhood Development: SABER Country Report 2014.”

                                                25
World Health Organization, United Nations Children’s Fund, and World Bank Group.
 2018. “Nurturing care for early childhood development: a framework for helping children survive
 and thrive to transform health and human potential.” Geneva: World Health Organization.




                                              26
Figures and Tables

                     Figure 1: Impact Evaluation Map




                                   27
Figure 2: Sample Selection




           28
Figure 3: Study Timeline




          29
Figure 4: Summary outcomes




           30
Figure 5: Employment index distribution by treatment site




                           31
Figure 6: Total monthly income distribution by treatment site




                             32
                                    Table 1: Baseline Balance Household level (Full sample)

                                                                         (1)                       (2)                       (3)              T-test
                                                                     Treatment                  Control                     Total            Difference
Variable                                                      N/[Clusters]   Mean/SE    N/[Clusters]   Mean/SE    N/[Clusters]    Mean/SE     (1)-(2)

Household level

Household size                                                   928         6.015         1062        6.085         1990         6.052       -0.070
                                                                 [18]       (0.219)        [18]       (0.171)        [36]        (0.135)
Number of children aged 6 or less                                928         1.547         1062        1.603         1990         1.577       -0.055
                                                                 [18]       (0.063)        [18]       (0.045)        [36]        (0.038)
Number of children aged 7 to 15 in the hh                        928         1.561         1062        1.612         1990         1.588       -0.051
                                                                 [18]       (0.076)        [18]       (0.074)        [36]        (0.053)
Number of economically active members in the hh                  928         1.787         1062        1.881         1990         1.837       -0.095
                                                                 [18]       (0.093)        [18]       (0.081)        [36]        (0.062)
Household head is a female                                       919         0.104         1055        0.088         1974         0.096      0.016**
                                                                 [18]       (0.013)        [18]       (0.013)        [36]        (0.009)
Household head age                                               919        40.260         1055       40.462         1974        40.368       -0.202
                                                                 [18]       (0.522)        [18]       (0.374)        [36]        (0.313)
Household head is a mossi                                        916         0.693         1053        0.741         1969         0.719       -0.048
                                                                 [18]       (0.085)        [18]       (0.072)        [36]        (0.054)
Household head is in monogamous marriage                         918         0.702         1054        0.683         1972         0.692        0.018
                                                                 [18]       (0.033)        [18]       (0.034)        [36]        (0.024)
Household head is non educated                                   919         0.456         1055        0.430         1974         0.442        0.026
                                                                 [18]       (0.034)        [18]       (0.031)        [36]        (0.022)
Household head is working                                        919         0.768         1054        0.787         1973         0.778       -0.018
                                                                 [18]       (0.024)        [18]       (0.027)        [36]        (0.018)
Age of brigadiere                                                928        30.764         1062       30.981         1990        30.880       -0.217
                                                                 [18]       (0.305)        [18]       (0.367)        [36]        (0.242)
Brigadiere is in monogamous marriage                             928         0.683         1062        0.668         1990         0.675        0.016
                                                                 [18]       (0.033)        [18]       (0.032)        [36]        (0.023)
Brigadiere is non educated                                       928         0.444         1062        0.478         1990         0.462       -0.034
                                                                 [18]       (0.030)        [18]       (0.033)        [36]        (0.023)
Brigadiere total monthly income                                  928       16559.433       1062      21178.503       1990       19024.484    -4619.070
                                                                 [18]      (2495.596)      [18]      (4513.668)      [36]       (2754.983)
Brigadiere total work duration in hours per week                 928        27.683         1062       37.314         1990        32.822      -9.631**
                                                                 [18]       (3.545)        [18]       (5.207)        [36]        (3.580)
Brigadiere has worked for 1 h last 30 days                       928         0.985         1062        0.972         1990         0.978       0.013*
                                                                 [18]       (0.005)        [18]       (0.010)        [36]        (0.006)
Brigadiere has worked for 1 h last 6 months                      928         0.996         1062        0.991         1990         0.993        0.005
                                                                 [18]       (0.002)        [18]       (0.005)        [36]        (0.003)
Time in hours taking care of children during last 24h            928         5.760         1062        6.045         1990         5.912       -0.285
                                                                 [18]       (0.365)        [18]       (0.198)        [36]        (0.195)
Time in hours cooking/washing/housework during last 24h          928         2.417         1062        2.598         1990         2.513      -0.181**
                                                                 [18]       (0.123)        [18]       (0.097)        [36]        (0.079)
Time in hours spent at school during last 24h                    928         0.049         1062        0.050         1990         0.050       -0.001
                                                                 [18]       (0.023)        [18]       (0.019)        [36]        (0.015)
Time in hours doing a paid job during last 24h                   928         3.504         1062        3.420         1990         3.459        0.084
                                                                 [18]       (0.238)        [18]       (0.401)        [36]        (0.239)
Household had any children in a creche in the past 24 hours      928         0.028         1062        0.016         1990         0.022        0.012
                                                                 [18]       (0.008)        [18]       (0.004)        [36]        (0.004)

  Notes: The value displayed for t-tests are the differences in the means across the groups. Standard errors are clustered at
  variable commune. Fixed effects using variable id sample are included in all estimation regressions. ***, **, and * indicate
  significance at the 1, 5, and 10 percent critical level.




                                                                            33
                                           Table 2: Baseline Balance Child level (Full sample)

                                                                                (1)                      (2)                      (3)            T-test
                                                                           Treatment                  Control                   Total           Difference
Variable                                                             N/[Clusters]   Mean/SE   N/[Clusters]   Mean/SE   N/[Clusters]   Mean/SE    (1)-(2)

Child level

Child is female                                                         1436         0.485       1702        0.497        3138        0.492      -0.012
                                                                        [18]        (0.014)      [18]       (0.009)       [36]       (0.008)
Child age                                                               1436         2.969       1702        3.069        3138        3.023     -0.100**
                                                                        [18]        (0.043)      [18]       (0.051)       [36]       (0.034)
Biological mother is in the household                                   1432         0.929       1698        0.945        3130        0.938      -0.015*
                                                                        [18]        (0.010)      [18]       (0.007)       [36]       (0.006)
Child is the household head son/daughter                                1434         0.849       1699        0.868        3133        0.860      -0.019
                                                                        [18]        (0.031)      [18]       (0.009)       [36]       (0.015)
Child (age=6) goes to school                                            131          0.985       155         0.994        286         0.990      -0.009
                                                                        [18]        (0.010)      [18]       (0.006)       [36]       (0.006)
Gross motor score                                                       1003        35.952       1132       36.348        2135       36.162      -0.396
                                                                        [18]        (0.563)      [18]       (0.432)       [36]       (0.349)
Fine motor score                                                        1003        19.921       1132       19.765        2135       19.838       0.156
                                                                        [18]        (0.463)      [18]       (0.306)       [36]       (0.266)
Language score                                                          1003        29.303       1132       28.799        2135       29.036       0.504
                                                                        [18]        (0.806)      [18]       (0.339)       [36]       (0.411)
Total child development score                                           1003        85.176       1132       84.913        2135       85.037       0.264
                                                                        [18]        (1.740)      [18]       (0.933)       [36]       (0.940)
The duration in hours where mother keeps the child                      1432         3.154       1698        3.336        3130        3.252      -0.182
                                                                        [18]        (0.154)      [18]       (0.174)       [36]       (0.118)
The duration in hours where father keeps the child                      1432         0.225       1698        0.234        3130        0.230      -0.009
                                                                        [18]        (0.024)      [18]       (0.018)       [36]       (0.014)
The duration in hours where sister keeps the child                      1432         0.114       1698        0.173        3130        0.146     -0.058***
                                                                        [18]        (0.015)      [18]       (0.024)       [36]       (0.017)
The duration in hours where brother keeps the child                     1432         0.044       1698        0.053        3130        0.049      -0.009
                                                                        [18]        (0.011)      [18]       (0.014)       [36]       (0.009)
The duration in hours where another familly member keeps the child      1432         0.800       1698        0.882        3130        0.844      -0.082
                                                                        [18]        (0.089)      [18]       (0.068)       [36]       (0.054)
The duration in hours where non familly member keeps the child          1432         0.214       1698        0.145        3130        0.176      0.069*
                                                                        [18]        (0.044)      [18]       (0.022)       [36]       (0.025)
The duration in hours of the child in the caring center                 1432         0.048       1698        0.025        3130        0.036      0.023*
                                                                        [18]        (0.014)      [18]       (0.006)       [36]       (0.008)
The duration in hours of the child with no care                         1432        14.017       1698       13.607        3130       13.795       0.410
                                                                        [18]        (0.480)      [18]       (0.437)       [36]       (0.321)
Child used a creche in the past 24 hours                                1432         0.018       1698        0.010        3130        0.014       0.008
                                                                        [18]        (0.006)      [18]       (0.003)       [36]       (0.003)



  Notes: The value displayed for t-tests are the differences in the means across the groups. Standard errors are clustered at
  variable commune. Fixed effects using variable id sample are included in all estimation regressions. ***, **, and * indicate
  significance at the 1, 5, and 10 percent critical level.




                                                                               34
                               Table 3: Effects on the use of childcare centers

                                  (1)                    (2)                        (3)                           (4)
                      Household had any children At least one child         At least one child       Maximum number of months
                            in a creche in       attended a creche in attended a public works creche  any child in the household
                          the past 24 hours      the past 12 months       in the past 12 months          attended any creche
Full sample
ITT estimate                    0.119***                0.241***                  0.235***                      0.720***
                               (0.026)                 (0.026)                   (0.027)                       (0.145)
Observations                    1961                    1965                      1965                          1965
R2                              0.051                   0.127                     0.181                         0.081
Control mean                    0.100                   0.120                     0.003                         0.571

Have child aged 0-2
ITT estimate                    0.112***                0.268***                  0.253***                      0.756***
                               (0.028)                 (0.037)                   (0.033)                       (0.170)
Observations                    1124                    1126                      1126                          1126
R2                              0.053                   0.155                     0.202                         0.086
Control mean                    0.094                   0.091                     0.003                         0.472


                 Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.




                                                                35
                             Table 4: Effects on summary outcome measures (indices)

                                    (1)                 (2)                    (3)                (4)              (5)
                            Women’s employment Women’s decision-making Women’s mental health Women’s finance Child development
  Full sample
  ITT estimate                     0.083**               0.002                 0.162**            0.174***         0.203***
                                  (0.040)               (0.045)               (0.069)            (0.058)          (0.058)
  Lee lower bound ITT              0.069*                -0.026                0.123*             0.153**          0.190***
                                  (0.040)               (0.046)               (0.068)            (0.057)          (0.057)
  Lee upper bound ITT              0.116***              0.024                 0.202***           0.201***         0.222***
                                  (0.036)               (0.043)               (0.068)            (0.056)          (0.055)
  Rand. inference pvalues          0.180                0.920                  0.200              0.120           0.020
  TOT estimate                     0.351**               0.007                 0.686**            0.742***         0.882***
                                  (0.168)               (0.190)               (0.281)            (0.258)          (0.282)
  Observations                     1967                  1967                   1957              1967            1618
  R2                              0.081                 0.108                  0.117             0.083            0.608
  Control mean                    -0.000                -0.008                 -0.000            -0.000           0.000

  Have child aged 0-2
  ITT estimate                     0.094***              -0.032                0.120**            0.175**          0.186***
                                  (0.033)               (0.052)               (0.057)            (0.069)          (0.060)
  Lee lower bound ITT              0.074**               -0.063                0.075              0.151**          0.166**
                                  (0.033)               (0.053)               (0.056)            (0.068)          (0.061)
  Lee upper bound ITT              0.130***              -0.004                0.166***           0.207***         0.220***
                                  (0.031)               (0.048)               (0.058)            (0.067)          (0.059)
  Rand. inference pvalues          0.100                0.660                  0.180              0.080           0.020
  TOT estimate                     0.367***              -0.129                0.472**            0.692**          0.701***
                                  (0.130)               (0.195)               (0.221)            (0.273)          (0.230)
  Observations                     1127                  1127                   1121              1127              717
  R2                              0.074                 0.118                  0.120             0.096             0.601
  Control mean                    -0.000                -0.006                 -0.000            -0.001           -0.000


Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10. Each cell represents one estimate of the
mean effect for a family of outcomes. Intent to Treat (ITT) effects estimated using the ANCOVA specification in equation 1
and TOT effects estimated using the specification in equation 2. Employment index in column 1: An indicator for working
at least 1 hour on activity in the last 30 days, number of hours spent on activity in the past 30 days, total monthly income
and IHS of total monthly income for salaried, agricultural, and non-agricultural activities, and time in hours doing paid
work in the past 24 hours (outcomes in Tables 5, 6, 7, and column 4 of Table 10). Empowerment index in column 2:
input into household decisions (on own income, household savings, major household purchases, childbearing, and children’s
education) autonomy over household decisions, (-) prevented by husband or household member from visiting relatives or
working outside the home, (-) believes that a husband is justified to beats his wife if she burns food or neglects children,
partner contributes to food preparation, housekeeping, laundry, and childcare (outcomes in Table 8, 9, and columns 1 and
2 of Table 10). Mental health index in column 3: (outcomes in Table 11). Finance index in column 4: saved money during
the past 12 months, saved money in a formal institution, saved money in an informal institution, able to pay 20,000 FCFA
in case of an emergency (outcomes in Table 12). Child development index in column 5: gross motor score, fine motor score,
and language score (outcomes in Table 13).




                                                                  36
                          Table 5: Effects on women’s salaried activity outcomes

                                  (1)                           (2)                       (3)                       (4)
                      Brigadiere has done salaried   Brigadiere salaried work  Brigadiere total monthly Brigadiere total monthly
                          work 1h last 30 days     duration in hours per month      salary income          salary income (ihs)
Full sample
ITT estimate                      0.005                        11.613                     2993.193**                  1.102**
                                 (0.057)                       (9.091)                   (1167.416)                  (0.494)
Observations                      1966                          1967                        1967                      1967
R2                                0.106                         0.221                       0.212                     0.139
Control mean                      0.329                        91.317                     6494.465                    3.480

Have child aged 0-2
ITT estimate                      0.031                        17.549*                    3267.278***                 1.343***
                                 (0.062)                       (9.386)                   (1093.277)                  (0.489)
Observations                      1126                          1127                        1127                      1127
R2                                0.096                         0.237                       0.219                     0.139
Control mean                      0.328                        96.645                     6452.145                    3.460


                  Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.




                       Table 6: Effects on women’s agricultural activity outcomes

                                   (1)                            (2)                          (3)                      (4)
                      Brigadiere has worked on agr Brigadiere agr act work duration Brigadiere total monthly Brigadiere total monthly
                           act 1h last 30 days           in hours per month              agr act income        agr act income (ihs)
Full sample
ITT estimate                     -0.023                         9.165                       106.979                    0.266
                                (0.044)                        (7.298)                     (321.352)                  (0.194)
Observations                     1966                            1967                         1967                    1967
R2                               0.106                           0.367                       0.433                    0.144
Control mean                     0.260                          77.202                      2581.367                  1.201

Have child aged 0-2
ITT estimate                     -0.036                         8.954                       59.204                     0.228
                                (0.045)                        (8.056)                     (292.838)                  (0.178)
Observations                     1126                            1127                         1127                    1127
R2                               0.134                           0.349                       0.433                    0.143
Control mean                     0.272                          79.889                      2733.261                  1.282


                  Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.




                                                                37
                           Table 7: Effects on women’s non-agricultural activity outcomes

                                            (1)                            (2)                       (3)                      (4)
                             Brigadiere has worked on non-agr Brigadiere non-agr act work Brigadiere total monthly Brigadiere total monthly
                                    act 1h last 30 days       duration in hours per month   non-agr act income     non-agr act income (ihs)
Full sample
ITT estimate                                    0.065                                 0.681                               721.356                        -0.005
                                               (0.052)                               (8.951)                             (927.379)                      (0.564)
Observations                                    1966                                  1967                                 1967                          1967
R2                                              0.060                                 0.189                                0.331                         0.123
Control mean                                    0.548                                36.259                              3230.932                        2.244

Have child aged 0-2
ITT estimate                                    0.062                                 0.564                               875.980                        0.050
                                               (0.047)                               (8.637)                             (797.004)                      (0.496)
Observations                                    1126                                  1127                                 1127                          1127
R2                                              0.083                                 0.156                                0.262                         0.129
Control mean                                    0.543                                34.192                              2919.580                        2.130


                         Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.



                               Table 8: Effects on women’s other empowerment outcomes
                               (1)                  (2)                 (3)                  (4)                    (5)                   (6)                   (7)
                      Index of brigadieres Index of brigadieres Index of brigadieres Index of brigadieres Husband or household Justified that a husband    Work division
                        voice on decision    voice on decision     own decision         own decision      member restrict wife     beats his wife if       and husband
                         making in the        making in the        making in the        making in the      visiting or working    she burns food or    participation in work
                       household (total) household (average) household (total) household (average)          in last 6 months       neglects children       index (total)
Full sample
ITT estimate                 -0.209               -0.048              -0.075              -0.023                0.000                   -0.005                 0.004
                            (0.185)              (0.037)             (0.344)             (0.070)               (0.013)                 (0.045)                (0.222)
Observations                1967                  1943                1967                1954                  1967                    1967                     1967
R2                          0.123                 0.141               0.083               0.090                 0.027                   0.034                    0.170
Control mean                4.439                 0.908               3.290               0.675                 0.083                   0.214                    4.313

Have child aged 0-2
ITT estimate                 -0.300               -0.060              -0.144              -0.031                0.012                   -0.002                 -0.060
                            (0.187)              (0.038)             (0.356)             (0.073)               (0.016)                 (0.044)                (0.220)
Observations                1127                  1117                1127                1120                  1127                    1127                     1127
R2                          0.186                 0.201               0.108               0.110                 0.044                   0.062                    0.146
Control mean                4.427                 0.901               3.274               0.668                 0.087                   0.222                    4.361


                         Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.



                                           Table 9: Effects on intra-household dynamics

                                       (1)                     (2)                            (3)                              (4)
                             Brigadiere takes care Brigadiere often takes care Brigadiere takes care of children Brigadieres partner often takes
                               of children alone           of children              as well as her partner              care of children
Full sample
ITT estimate                           -0.031                         -0.035                                0.063                                    -0.001
                                      (0.048)                        (0.060)                               (0.041)                                  (0.026)
Observations                           1788                           1788                                  1788                                     1788
R2                                     0.088                          0.055                                 0.072                                    0.092
Control mean                           0.307                          0.505                                 0.153                                    0.034

Have child aged 0-2
ITT estimate                           -0.055                         -0.020                                0.077*                                   -0.003
                                      (0.043)                        (0.058)                               (0.038)                                  (0.027)
Observations                           1035                           1035                                  1035                                     1035
R2                                     0.100                          0.079                                 0.102                                    0.093
Control mean                           0.325                          0.495                                 0.145                                    0.034


                         Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.




                                                                                   38
                                     Table 10: Effects on women’s time use

                                           (1)                      (2)                    (3)               (4)
                                       Taking care           Cooking/washing/           At school           Doing
                                       of children              housework                                 paid work
      Full sample
      ITT estimate                        -0.280                   -0.130                 0.018              0.613***
                                         (0.396)                  (0.148)                (0.100)            (0.200)
      Observations                        1961                     1964                   1963               1963
      R2                                  0.249                    0.184                  0.527              0.135
      Control mean                        4.800                    2.810                  0.402              3.461

      Have child aged 0-2
      ITT estimate                        -0.069                   -0.006                 0.044              0.650**
                                         (0.387)                  (0.153)                (0.094)            (0.253)
      Observations                        1122                     1125                   1124               1124
      R2                                  0.266                    0.170                  0.581              0.125
      Control mean                        4.765                    2.723                  0.418              3.266


Notes: Each column reports the number of hours spent on a given activity in the last 24 hours. Robust standard errors in
parentheses *** p<0.01, ** p<0.05, * p<0.10.




                            Table 11: Effects on women’s psychological well-being

                                                (1)                           (2)                          (3)
                                          Depression score            Depressed (score>10)             Unhappiness
        Full sample
        ITT estimate                           -0.631                         -0.075*                      -0.143**
                                              (0.476)                        (0.040)                      (0.056)
        Observations                            1957                          1957                        1957
        R2                                     0.114                          0.089                       0.077
        Control mean                           10.920                         0.604                       2.315

        Have child aged 0-2
        ITT estimate                           -0.350                         -0.054                       -0.127**
                                              (0.412)                        (0.035)                      (0.055)
        Observations                            1121                          1121                        1121
        R2                                     0.111                          0.087                       0.102
        Control mean                           10.701                         0.592                       2.288


                     Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.




                                                             39
                         Table 12: Effects on women’s financial outcomes

                              (1)               (2)                 (3)                       (4)
                      Saved money during Saved money via      Saved money via    Could pay 20000 CFA in case
                        past 12 months   formal institution informal institution       of urgent matter
Full sample
ITT estimate                 0.081**             0.079*               0.070*                     0.108***
                            (0.038)             (0.042)              (0.040)                    (0.033)
Observations                 1963                1967                1967                        1967
R2                           0.060               0.074               0.041                       0.052
Control mean                 0.588               0.317               0.405                       0.404

Have child aged 0-2
ITT estimate                 0.079*              0.088*               0.071                      0.100***
                            (0.044)             (0.045)              (0.047)                    (0.033)
Observations                 1124                1127                1127                        1127
R2                           0.072               0.093               0.057                       0.043
Control mean                 0.583               0.306               0.412                       0.387


               Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.




                               Table 13: Effects on child development

                              (1)              (2)             (3)                     (4)
                       Gross motor score Fine motor score Language score Total child development score
Full sample
ITT estimate                  1.545***            2.748***           0.160                      4.391***
                             (0.498)             (0.640)            (0.528)                    (1.343)
Observations                  1618                1618               1618                       1618
R2                           0.481                0.325              0.561                      0.636
Control mean                 43.524              20.689             34.816                     99.029
Max score                      52                  54                 46                         152

Have child aged 0-2
ITT estimate                  1.939***            1.318**            0.832                      3.895***
                             (0.496)             (0.499)            (0.714)                    (1.312)
Observations                  717                  717                717                        717
R2                           0.534                0.257              0.538                      0.636
Control mean                 39.139              16.959             26.584                     82.682
Max score                      52                   41                 46                        139


               Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.




                                                     40
           Table 14: Effects on main primary caregivers (duration in hours where)

                            (1)           (2)           (3)           (4)             (5)           (6)
                       mother cares father cares sister cares brother cares        child in     child with
                       for the child for the child for the child for the child childcare center  no care
Full sample
ITT estimate               -0.258         -0.072         -0.050*        -0.022           0.352***    0.710
                          (0.886)        (0.186)        (0.029)        (0.017)          (0.121)     (0.982)
Observations                2968          2968            2968          2968             2968       2968
R2                          0.107         0.248           0.546         0.524            0.326      0.161
Control mean               15.982         1.570           0.242         0.158            0.407      1.615

Have child aged 0-2
ITT estimate               -0.204         -0.185         -0.034         0.006            0.398***    0.884
                          (0.887)        (0.198)        (0.033)        (0.020)          (0.079)     (1.081)
Observations                1206          1206            1206          1206             1206       1206
R2                          0.101         0.232           0.488         0.483            0.317      0.184
Control mean               17.032         1.602           0.256         0.148            0.184      1.254


               Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.




                                                     41
Appendix

In the figures and tables that follow, we provide additional results with baseline balance tests for the

subsample of households with children aged 0 to 2, robustness checks, and tests of heterogeneity.

                             Table A1: Baseline Balance Household level (Have a child aged 0-2)

                                                                                    (1)                           (2)                           (3)                  T-test
                                                                                Treatment                      Control                         Total                Difference
Variable                                                                 N/[Clusters]   Mean/SE        N/[Clusters]   Mean/SE        N/[Clusters]    Mean/SE         (1)-(2)
Household level

Household size                                                               538            6.314          599            6.314          1137            6.314        0.000
                                                                             [18]          (0.302)         [18]          (0.211)         [36]           (0.178)
Number of children aged 6 or less                                            538            1.784          599            1.828          1137            1.807       -0.044
                                                                             [18]          (0.089)         [18]          (0.054)         [36]           (0.050)
Number of children aged 7 to 15 in the hh                                    538            1.496          599            1.516          1137            1.507       -0.020
                                                                             [18]          (0.102)         [18]          (0.102)         [36]           (0.071)
Number of economically active members in the hh                              538            1.877          599            1.907          1137            1.893       -0.029
                                                                             [18]          (0.106)         [18]          (0.085)         [36]           (0.067)
DV =1 if household head is a female                                          534            0.092          598            0.075          1132            0.083        0.017
                                                                             [18]          (0.013)         [18]          (0.015)         [36]           (0.009)
Househould head age                                                          534           39.234          598           39.403          1132           39.323       -0.169
                                                                             [18]          (0.518)         [18]          (0.429)         [36]           (0.330)
DV=1 if the household head is a mossi                                        532            0.709          596            0.735          1128            0.723       -0.026
                                                                             [18]          (0.094)         [18]          (0.074)         [36]           (0.058)
DV=1 if hh head is in monogamous marriage                                    534            0.697          597            0.675          1131            0.685        0.022
                                                                             [18]          (0.038)         [18]          (0.043)         [36]           (0.029)
DV=1 if the household head is non educated                                   534            0.446          598            0.421          1132            0.433        0.024
                                                                             [18]          (0.038)         [18]          (0.031)         [36]           (0.024)
DV=1 if household head is working                                            534            0.794          597            0.774          1131            0.783        0.020
                                                                             [18]          (0.021)         [18]          (0.030)         [36]           (0.018)
Age of brigadiere                                                            538           29.322          599           29.728          1137           29.536       -0.406
                                                                             [18]          (0.353)         [18]          (0.309)         [36]           (0.234)
Brigadiere is in monogamous marriage                                         538            0.675          599            0.664          1137            0.669        0.010
                                                                             [18]          (0.035)         [18]          (0.044)         [36]           (0.028)
Brigadiere is non educated                                                   538            0.387          599            0.464          1137            0.427      -0.077**
                                                                             [18]          (0.029)         [18]          (0.035)         [36]           (0.024)
Brigadiere total monthly income                                              538         17381.325         599         19767.549         1137          18638.448    -2386.224
                                                                             [18]        (3120.340)        [18]        (4465.018)        [36]          (2781.537)
Brigadiere total work duration in hours per week                             538           27.965          599           36.081          1137           32.241      -8.115**
                                                                             [18]          (4.068)         [18]          (4.927)         [36]           (3.510)
Brigadiere has worked for 1 h last 30 days                                   538            0.980          599            0.970          1137            0.974        0.010
                                                                             [18]          (0.008)         [18]          (0.013)         [36]           (0.008)
Brigadiere has worked for 1 h last 6 months                                  538            0.993          599            0.987          1137            0.989        0.006
                                                                             [18]          (0.004)         [18]          (0.007)         [36]           (0.004)
Time in hours taking care of children during last 24h                        538            6.505          599            6.824          1137            6.673       -0.319
                                                                             [18]          (0.457)         [18]          (0.194)         [36]           (0.230)
Time in hours taking for cooking/washing/housework during last 24h           538            2.426          599            2.649          1137            2.543      -0.222**
                                                                             [18]          (0.110)         [18]          (0.093)         [36]           (0.077)
Time in hours spending at school during last 24h                             538            0.082          599            0.038          1137            0.059        0.043
                                                                             [18]          (0.039)         [18]          (0.023)         [36]           (0.022)
Time in hours doing a paid job during last 24h                               538            3.413          599            3.329          1137            3.369        0.083
                                                                             [18]          (0.251)         [18]          (0.389)         [36]           (0.235)
Household had any children in a creche in the past 24 hours                  538            0.022          599            0.008          1137            0.015       0.014*
                                                                             [18]          (0.009)         [18]          (0.003)         [36]           (0.004)
Notes : The value displayed for t-tests are the differences in the means across the groups. Standard errors are clustered at variable commune. Fixed
effects using variable id sample are included in all estimation regressions. ***, **, and * indicate significance at the 1, 5, and 10 percent critical
level.




                                                                                      42
                                 Table A2: Baseline Balance Child level (Is a child aged 0-2)

                                                                                    (1)                          (2)                          (3)                 T-test
                                                                               Treatment                      Control                       Total                Difference
Variable                                                                 N/[Clusters]   Mean/SE       N/[Clusters]   Mean/SE       N/[Clusters]   Mean/SE         (1)-(2)
Child level

Child is female                                                              587           0.491          653           0.487          1240             0.489      0.004
                                                                             [18]         (0.014)         [18]         (0.013)         [36]            (0.009)
Child age                                                                    587           1.136          653           1.240          1240             1.191    -0.104***
                                                                             [18]         (0.044)         [18]         (0.046)         [36]            (0.031)
Biological mother is in the household                                        584           0.971          650           0.972          1234             0.972     -0.001
                                                                             [18]         (0.008)         [18]         (0.009)         [36]            (0.006)
Child is the household head son/daughter                                     586           0.860          651           0.877          1237             0.869     -0.017
                                                                             [18]         (0.030)         [18]         (0.010)         [36]            (0.015)
Gross motor score                                                            493          23.968          544          24.449          1037            24.220     -0.481
                                                                             [18]         (0.639)         [18]         (0.667)         [36]            (0.448)
Fine motor score                                                             493          14.422          544          14.645          1037            14.539     -0.223*
                                                                             [18]         (0.412)         [18]         (0.335)         [36]            (0.256)
Language score                                                               493          14.389          544          14.550          1037            14.473     -0.160
                                                                             [18]         (0.573)         [18]         (0.517)         [36]            (0.376)
Total child development score                                                493          52.779          544          53.643          1037            53.232     -0.864
                                                                             [18]         (1.581)         [18]         (1.463)         [36]            (1.044)
The duration in hours where mother keeps the child                           584           4.125          650           4.274          1234             4.204     -0.149
                                                                             [18]         (0.217)         [18]         (0.229)         [36]            (0.157)
The duration in hours where father keeps the child                           584           0.189          650           0.182          1234             0.185      0.007
                                                                             [18]         (0.038)         [18]         (0.032)         [36]            (0.024)
The duration in hours where sister keeps the child                           584           0.100          650           0.161          1234             0.132    -0.062**
                                                                             [18]         (0.021)         [18]         (0.031)         [36]            (0.021)
The duration in hours where brother keeps the child                          584           0.037          650           0.048          1234             0.043     -0.011
                                                                             [18]         (0.011)         [18]         (0.014)         [36]            (0.009)
The duration in hours where another familly member keeps the child           584           0.659          650           0.695          1234             0.678     -0.037
                                                                             [18]         (0.067)         [18]         (0.057)         [36]            (0.043)
The duration in hours where non familly member keeps the child               584           0.146          650           0.112          1234             0.128      0.034
                                                                             [18]         (0.034)         [18]         (0.027)         [36]            (0.021)
The duration in hours of the child in the caring center                      584           0.033          650           0.000          1234             0.015    0.033***
                                                                             [18]         (0.013)         [18]         (0.000)         [36]            (0.006)
The duration in hours of the child with no care                              584          12.682          650          12.402          1234            12.534      0.280
                                                                             [18]         (0.492)         [18]         (0.506)         [36]            (0.350)
Child used a creche in the past 24 hours                                     584           0.012          650           0.000          1234             0.006    0.012***
                                                                             [18]         (0.005)         [18]         (0.000)         [36]            (0.002)
Notes : The value displayed for t-tests are the differences in the means across the groups. Standard errors are clustered at variable commune. Fixed
effects using variable id sample are included in all estimation regressions. ***, **, and * indicate significance at the 1, 5, and 10 percent critical
level.




                                                                                    43
                            Table A3: Robustness checks for primary outcome indicators

                                  (1)            (2)                (3)              (4)               (5)                 (6)
                            Brigadier has   Brigadier total   Brigadiere time   Brigadier total  Brigadier total       Total child
                            worked for 1 h work duration in in hours doing paid monthly income monthly income (ihs) development score
                             last 30 days  hours per month work during last 24h
  Full sample
  ITT estimate                  -0.023         19.591             0.550***         3531.518            0.698*             4.302***
                               (0.038)        (14.645)           (0.198)          (2176.089)          (0.405)            (1.350)
  Lee lower bound ITT           -0.038         15.026             0.475**          3265.859            0.575              3.963***
                               (0.037)        (14.910)           (0.197)          (2211.038)          (0.398)            (1.331)
  Lee upper bound ITT           -0.017         31.031**           0.768***         5482.407**          0.849**            4.690***
                               (0.038)        (13.437)           (0.197)          (2040.422)          (0.400)            (1.293)
  TOT estimate                  -0.097         82.841             2.332***        14963.922*           2.964*            18.706***
                               (0.159)        (62.977)           (0.885)          (8637.942)          (1.682)            (6.479)
  Observations                  1966           1967               1963              1967               1967                1618
  R2                            0.082          0.100              0.068             0.051              0.070              0.633
  Control mean                  0.723         206.864             3.461           12936.754            5.856              99.029
  Rand. inference pvalues       0.580          0.380              0.200             0.180              0.340               0.020

  Have child aged 0-2
  ITT estimate                  0.009          24.535*            0.671**          3285.891*           1.022**            3.895***
                               (0.040)        (12.802)           (0.252)          (1856.626)          (0.404)            (1.312)
  Lee lower bound ITT           -0.013         19.382             0.575**          2981.112            0.881**            3.366**
                               (0.038)        (12.681)           (0.256)          (1849.512)          (0.395)            (1.331)
  Lee upper bound ITT           0.019          36.492***          0.913***         5564.845***         1.201***           4.609***
                               (0.039)        (11.927)           (0.259)          (1554.561)          (0.403)            (1.313)
  TOT estimate                  0.035          96.029*            2.644**         12886.256**          4.007***          14.619***
                               (0.153)        (51.218)           (1.054)          (6493.315)          (1.552)            (4.956)
  Observations                  1126           1127               1124              1127               1127                 717
  R2                            0.088          0.102              0.086             0.048              0.062              0.636
  Control mean                  0.706         212.704             3.266           13012.128            5.728              82.682
  Rand. inference pvalues       0.940          0.360              0.060             0.180              0.060               0.040


Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.Each cell represents one estimate of the
mean effect for a family of outcomes. Intent to Treat (ITT) effects estimated using the ANCOVA specification in equation
1 and TOT effects estimated using the specification in equation 2.




                                           Table A4: Joint test of orthogonality

                                                           Prob > F          Joint test at 5% level


                               Household level
                               Full sample                   0.0039                 Rejected
                               Have child 0-2                0.0000                 Rejected




                               Child level
                               Full sample                   0.0008                 Rejected
                               His child 0-2                 0.0000                 Rejected




                                                                   44
               Table A5: Depression score variables


Score        Variables




             She   was bothered by details of daily life more than usual
             She   had trouble concentrating on what you were doing
             She   felt sad
             She   felt that everything she did took all her energy
             She   felt nervous, tense or worried
Depression   She   had trouble sleeping peacefully
             She   felt alone
             She   was so tired that she couldn’t do anything
             She   was confident in the future
             She   was happy




                                  45
                            Table A6: Composition of summary indices


Index                   Variables




                        Brigadiere has done a salary work for 1h last 30 days
                        Brigadiere has worked 1h on non-agr act last 30 days
                        Brigadiere has worked 1h on agr act last 30 days
                        Brigadier total monthly salary income win at 99%
                        Brigadier total monthly non agricole activity income win at 99%
                        Brigadier total monthly agr activity income win at 99%
                        Brigadier total working hours in a month on agr act win at 99%
                        Brigadier total working hours in a month on non agr act win at 99%
Employment              Brigadier total working hours in a month on salary act win at 99%
                        Time in hours doing a paid job during last 24h win at 99%
                        IHS of Brigadier total monthly salary income win at 99%
                        IHS Brigadier total monthly non agricole activity income win at 99% ihs
                        IHS Brigadier total monthly agr activ income win at 99%




                        Index of brigadiere’s voice on decision making in the household (total)
                        Index of brigadiere’s voice on decision making in the household (average)
                        Index of brigadiere’s own decision making in the household (total)
                        Index of brigadiere’s own decision making in the household (average)
                        (Reverse) Husband or a household member restrict wife to visit or to work, last 6 months
Decision and attitude   (Reverse) Justified that a husband beats his wife if she burns food or neglects children
                        (Reverse) Brigadiere takes care of children alone
                        (Reverse) Brigadiere often takes care of children
                        Brigadiere takes care of children as well as her partner
                        Brigadiere’s partner takes often care of children
                        Work division and husband participation to work index (total)
                        (Reverse) Time in hours taking care of children during last 24h win at 99%
                        (Reverse) Time in hours taking for cooking/washing/housework during last 24h win at 99%




                        (Reverse) Unhappiness
Mental health           (Reverse) Depression score
                        (Reverse) DV=1 if depressed (score>10)




                        Brigadiere has saved money during past 12 months
Financial               DV=1 if brigadiere has saved money via formal institution
                        Brigadiere has saved money via informal institution
                        Brigadiere could pay 20000 FC in case of urgent mater




                        Gross motor score win at 99%
Child development       Fine motor score win at 99%
                        Language score win at 99%




                                                       46
       Table A7: Effects on summary outcome measures (excluding missing observations)

                                  (1)                 (2)                    (3)                (4)              (5)
                          Women’s employment Women’s decision-making Women’s mental health Women’s finance Child development
Full sample
ITT estimate                     0.084**                  -0.014                   0.162**                 0.175***        0.175***
                                (0.040)                  (0.047)                  (0.069)                 (0.058)         (0.052)
Lee lower bound ITT              0.069*                   -0.040                   0.123*                  0.156**         0.162***
                                (0.040)                  (0.048)                  (0.068)                 (0.058)         (0.051)
Lee upper bound ITT              0.117***                 0.009                    0.202***                0.202***        0.190***
                                (0.036)                  (0.045)                  (0.068)                 (0.056)         (0.049)
Rand. inference pvalues          0.180                   0.800                     0.200                  0.120            0.020
TOT estimate                     0.354**                  -0.061                   0.687**                 0.743***        0.759***
                                (0.168)                  (0.196)                  (0.282)                 (0.256)         (0.251)
Observations                     1963                    1785                      1957                    1963             1618
R2                               0.080                   0.116                     0.117                  0.083             0.620
Control mean                     0.001                   0.017                     0.000                  -0.001           -0.098


                      Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.


                   Table A8: Effects on the use of childcare centers (excluding BAM)

                                           (1)                    (2)                        (3)                        (4)
                               Household had any children At least one child         At least one child      Number of months child
                                   in a creche in the     attended a creche in attended public works creche   attended public works
                                     past 24 hours        the past 12 months       in the past 12 months    creche(average at hh level)
Full sample
ITT estimate                                0.119***                0.241***                   0.235***                 0.530***
                                           (0.026)                 (0.026)                    (0.027)                  (0.095)
Observations                                1961                     1965                     1965                     1965
R2                                          0.051                    0.127                    0.181                    0.103
Control mean                                0.100                    0.120                    0.003                    0.344

Full sample without BAM
ITT estimate                                0.096***                0.205***                   0.204***                 0.391***
                                           (0.027)                 (0.022)                    (0.023)                  (0.084)
Observations                                1894                     1898                     1898                     1898
R2                                          0.040                    0.100                    0.153                    0.079
Control mean                                0.100                    0.120                    0.003                    0.344


                    Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.


                   Table A9: Effect on summary outcome measures (excluding BAM)

                                      (1)                 (2)                    (3)                (4)              (5)
                              Women’s employment Women’s decision-making Women’s mental health Women’s finance Child development
Full sample
ITT estimate                         0.083**                 0.002                   0.162**                0.174***        0.203***
                                    (0.040)                 (0.045)                 (0.069)                (0.058)         (0.058)
Observations                          1967                   1967                     1957                  1967            1618
R2                                   0.081                  0.108                     0.117                0.083            0.608
Control mean                         -0.000                 -0.008                   -0.000                -0.000           0.000

Full sample without BAM
ITT estimate                         0.068*                  0.004                   0.153**                0.164***        0.196***
                                    (0.040)                 (0.045)                 (0.069)                (0.057)         (0.058)
Observations                          1900                   1900                     1890                  1900            1557
R2                                   0.085                  0.112                     0.112                0.080            0.605
Control mean                         -0.000                 -0.008                   -0.000                -0.000           0.000


                    Notes: Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10.



                                                                   47
                Figure A1: Total monthly income distribution in treated sites




Notes: This figure displays the distribution of monthly income for public works participants in
the 18 treatment sites, distinguishing participants who worked as childcare center attendants
       (BAM) from those who did not work as childcare center attendants (non-BAM).




                                             48
                  Figure A2: Employment index distribution in treated sites




Notes: This figure displays the distribution of employment index for public works participants in
 the 18 treatment sites, distinguishing participants who worked as childcare center attendants
        (BAM) from those who did not work as childcare center attendants (non-BAM).




                                              49
Figure A3: Total monthly income distribution by treatment site (without BAM)




                                    50
Figure A4: Employment index distribution by treatment site (without BAM)




                                  51