More Care, More Workers? Gauging the Impact of Child Care Access on Labor Force Participation

Abstract

This study investigates the critical link between child care accessibility and local labor force participation, addressing a gap in current research that often lacks local spatial granularity. While over half of the U.S. population resides in child care deserts, disproportionately affecting rural, low-income, and minority communities, the economic implications for local labor markets remain underexplored. Leveraging Michigan child care license data and Census tract-level demographic and employment characteristics, this research employs a spatial econometric approach to estimate the impact of geographic distance to child care facilities on labor supply using descriptive data. Our findings consistently demonstrate that increased distance to child care is significantly associated with reduced labor force participation. While female labor force participation is lower in areas with constrained access to child care, we also found that households with two parents are also less likely to have full labor force participation when access to child care is constrained. The cost-effective framework used here can be replicated to identify specific communities most impacted by child care-related employment disruptions. The analytical findings can be instrumental in targeting and prioritizing child care policy interventions.

1. Introduction

American child care is in crisis. Over half (51%) of Americans live in “child care deserts,” or areas with three or more children having to compete for every licensed child care slot (Malik et al. 2018).1 During the COVID-19 pandemic, the U.S. federal government provided USD 24 billion in child care subsidies; three years later, the problem persisted country-wide (The White House 2024), with America’s child care deserts expected to have grown since the pandemic (Malik et al. 2020). The overall impact of America’s child care crisis has a profound effect on the nation’s economy, with Yellen (2020) estimating that low maternal employment reduces annual U.S. GDP by as much as five percent.

Like the rest of the U.S., Michigan has sharp constraints on child care availability—which have been exacerbated by the pandemic (Dunsford 2024). Before 2020, 44% of Michigan residents were estimated to live in a child care desert (Center for American Progress 2019). The persistence of these child care deserts directly shapes parental labor participation, as child care disruption in Michigan negatively impacts nearly a third of parents’ employment; this culminated in half of those parents (14%) leaving their job due to child care inaccessibility (U.S. Chamber of Commerce Foundation 2023). Lower-income households are also more likely to have child care issues impairing their labor participation (44%) than their high-income peers (27%) (U.S. Chamber of Commerce Foundation 2023). Absenteeism and turnover due to child care disruptions cost Michigan’s economy an estimated USD 2.88 billion annually, where much of the cost (USD 2.3 billion) is borne by employers due to parental workplace absenteeism and turnover, while the remainder (USD 576 million) is lost in state tax income (U.S. Chamber of Commerce Foundation 2023). And yet, because of a lack of geographic specificity in these and other analyses, it has proven difficult to determine which Michigan communities are most impacted by employment disruptions due to low child care access.

Policymakers have wide-ranging policy options for mitigating child care challenges. Malik et al. (2018) provides a comprehensive survey of policy recommendations for improving access to child care services. Most policy strategies are supply-enhancing, designed to make child care more accessible to households either physically or financially. Other supply-enhancing policies include providing financial incentives to businesses willing to establish child care services in targeted communities, like Michigan’s Caring for MI Future program (MiLEAP n.d.), or investing in the child care workforce through scholarships, stipends, and loan forgiveness programs. Often, promoting public–private partnerships is used to ensure access for disadvantaged populations.

For decades, policymakers have used child care as a tool to increase parental employment (Akgunduz and Plantenga 2018; Hotz and Wiswall 2019). A lack of reliable child care impedes parental employment opportunities, while greater availability of low-cost care is associated with greater maternal employment (Genç and Small 2024; Morrissey 2017). Both the financial (i.e., tuition) and temporal (i.e., transportation and waiting) costs of child care erode the returns to labor force participation of young and low-income households, especially those with single heads of household. Regardless of the form of child care policy, clear accounting of child care needs and resources is necessary.

Moreover, neighborhood demography is associated with the location and availability of child care providers. Suburban neighborhoods are less likely to be a child care desert (44%) than their rural (58%) and urban (55%) counterparts (Malik and Hamm 2017). White (49%) and Black (45%) families are the least likely to live in a neighborhood in a desert, while American Indian and Alaska Native (61%), Hispanic (57%), and Asian (51%) families are more likely to live in a child care desert (Malik and Hamm 2017). Low- and middle-income neighborhoods are the most likely to be a child care desert (Malik et al. 2020). This distribution of child care compounds economic inequity, as areas in greatest need of child care are unable to access it. Yet few studies recognize the role of disparities in geographic access to child care facilities and how distance to child care facilities affects parental employment on a state level.

Many U.S. studies on the challenges of child care access use the Census Bureau’s Survey of Income and Program Participation or the National Survey of Early Care and Education (NSECE 2014). These studies are national in scope but provide measurements down to the household or individual levels. However, spatial granularity restrictions of this data limit or prevent researchers’ exploration of physical proximity to child care facilities and outcomes, reducing the effectiveness of this approach for local policy considerations. While child care preferences vary by household, common considerations include quality/safety, price, children–provider interactions, and accessibility (i.e., proximity, hours of operation, and convenience) (Davidson et al. 2022; Herbst et al. 2020). Blumberg et al.’s 2024 analysis of California-licensed child care centers suggests that proximity to one’s home is a major factor. Although the weight given to location and convenience is difficult to quantify, it is an important consideration in parents’ choice of child care—hence the necessity of spatial analysis.

Policies that ensure broad access to quality child care represent an important community investment, creating a foundational support system that allows individuals, particularly parents, to more fully engage in both economic and civic life. Moreover, increased access to child care and parental employment has community-level benefits (Harbach 2019). As summarized by Harbach (2019), “…childcare generates significant social spillovers, including increased economic activity and development, a more sophisticated future workforce, an increased tax base, and cost savings on education, crime prevention, social services, and public assistance” (p. 466). Genç and Small (2024) suggest that access to quality child care also shapes the next generation’s workforce, as quality child care could provide greater educational and employment opportunities for low-income children. Other benefits include providing businesses with a larger pool of employees, as parental workplace absenteeism and turnover are reduced (Harbach 2019). Conversely, a lack of child care can negatively influence mothers’ employment and long-term career advancement, decreasing a community’s labor supply and businesses’ employment retention (Kimmel 2006).

Furthermore, child care facilities themselves can evolve into geographically central hubs for social capital formation (Oregon Early Childhood Inclusion 2024). They provide common meeting grounds for parents, caregivers, and staff, facilitating the exchange of information, mutual support, and collective action. This increased socialization is a direct catalyst for the formation of social capital, as it fosters the development of trust, shared norms, and reciprocal relationships among community members (Shan et al. 2014). Thus, investing in child care is not merely an expenditure on early education; it is also a strategic investment in the social infrastructure that underpins a thriving, connected, and resilient community.

This study sets out to explore the potential influence of child care access on local economic development through its effect on the size of the labor market. Our objective is to understand the association between access to licensed child care providers, as measured by geographic distance, and workforce participation of young households, with the implications that factors that reduce household labor participation reduce economic growth potential. The study uses tract-level data, which sets our approach apart from individual- and household-level analyses by providing an exploratory approach that any community or state can employ to identify needs and communicate potential economic losses associated with limited access to child care facilities. The measure of access used in this research is limited to geographic distance between households and the nearest licensed child care facility. The methods applied in this article can be replicated for other regions, as well as longitudinally, making this an accessible child care-workforce assessment framework for informing local and state-level public policies.

Our findings reflect those of other studies, namely, that limited access to child care is associated with reduced workforce participation. Our study is distinct in approaching the question of access by using travel distance. These findings contribute to an otherwise established literature around the benefits of public investment in child care services but go further to provide a cost-effective framework for communicating the potential economic losses due to limited access. While the study comes short of providing a causal assessment, it provides a baseline by which potential losses from reduced workforce participation can be measured and communicated.

2. Data

Our study leverages a combination of state child care license data and publicly available data across the state of Michigan. Our primary dataset, from the ECIC (Early Childhood Investment Corporation 2024), is a snapshot of all child care facilities across the state licensed by the Michigan Department of Licensing and Regulatory Affairs (LARA) (MI LARA 2024) for operation in July 2022. This data provides each facility’s physical address and geolocation, licensure type, stated ages of admittance, and number of seats approved under the child care license for that time. This dataset is the basis for determining our independent variable: physical proximity.

Our analysis focuses on how the local labor market is affected by child care access. Demographic and employment characteristics are derived from American Community Survey (ACS) data tabulated at the Census tract level. Therefore, our analysis is undertaken in the aggregate at the Census tract level, with the minimum distance between the tract centroid and child care providers used as our measure of access. The July 2022 child care license data showed significant variation in the counts of child care facilities by tract, with a mean count of 2.72, a maximum of 22, and a standard deviation of 2.44. The average distance from a tract centroid to the nearest child care provider is 1.13 miles, though this metric has a long right tail—in some areas of the state, parents would have to travel up to 20 miles to reach a child care provider. Figure 1 shows which areas of the state have the farthest travel requirements to a licensed child care facility; as providers tend to be located near population centers, the distribution of access across Michigan resembles how the population is distributed. With regard to provider characteristics, most child care providers are full-time (roughly 75% across the entire state) according to the MI LARA definition (MI LARA 2024). The proportion of providers that provide full-time care varies by area, though most local providers meet this standard.

The ECIC child care data is matched with data from the ACS (U.S. Census Bureau 2024b), consisting of 5-year population estimates at the Census tract level across all of Michigan’s 3017 unique tracts for 2018–2022. Around four percent of the Census tract data had invalid entries,2 resulting in 2889 complete Census tract records with information on (1) overall female labor force participation, (2) the proportion of children (under six years old) living in a household with two working parents, and (3) the proportion of children (under six years old) living in a single-parent household where the parent works (U.S. Census Bureau 2024b). Because the observations are aggregations at the Census tract level, the measures of interest are the labor force participation rates of female workers, the share of single-parent households where the parent works, and the share of two-parent households where both parents work.

Control Variables

We use a total of five control variables and three indices. The indices and three of the control variables are demographic to control for variation in ethnic, racial, and educational observable characteristics in the observed communities—as defined by Census tracts. These include the proportion of the population that is White, has a high school diploma or higher, and has a bachelor’s degree or higher, alongside three state-level rankings from the CDC’s Social Vulnerability Index (SVI) (Centers for Disease Control and Prevention et al. 2024) documentation, namely, the household characteristics, racial and ethnic minority status, and housing type and transportation indices. The remaining two control variables are related to the location itself: the proportion of the tract located in an urban area (as urban and rural communities likely have different relationships to formal child care) and the proportion of local employment found in shift-heavy industries.

Table 1. All summary statistics for the 2889 complete Census tracts. Sources: U.S. Census Bureau, CDC, MI LARA.

	Min	Mean	Std. Dev.	Max
Outcomes of interest
Female labor force participation	0.168	0.565	0.101	1.000
Children living in two-parent households/both work	0	0.576	0.300	1.000
Children living in single-parent households/parent works	0	0.729	0.355	1.000
Control variables
Proportion of residents that are White	0	0.732	0.279	1.000
Proportion of residents with a HS diploma or above	0.397	0.911	0.072	1.000
Proportion of residents with a bachelor’s degree or above	0	0.297	0.185	0.916
Proportion of jobs in tract that are “shift work”	0	0.368	0.214	0.966
Proportion of tract that is “urban”	0	0.656	0.436	1.000
SVI percentile: household characteristics	0	0.500	0.289	1.000
SVI percentile: racial and ethnic minority status	0	0.498	0.289	0.996
SVI percentile: housing type and transportation	0	0.500	0.289	1.000
Independent variables
Num. of child care centers	0	2.729	2.439	22.000
Minimum distance to nearest center	0.007	1.130	1.788	20.775
Minimum distance to nearest full-time center	0.011	1.254	2.012	28.911
Minimum distance to nearest part-time center	0.007	2.121	3.372	46.416
Proportion of full-time centers within 5 miles	0	0.726	0.195	1.000
Proportion of full-time centers within 10 miles	0	0.751	0.114	1.000

provides summary statistics of these variables.

Previous research, with information at the individual or family level, has used individual-level controls, including age, education, marital status, race, and previous income (Herbst 2010; Herbst and Tekin 2010). Where feasible, researchers also employed controls for location, such as the urban status of a family or region within the United States (Baum 2002; Connelly and Kimmel 2003). We use similar variables at a neighborhood level from the CDC’s Social Vulnerability Index (SVI) (Centers for Disease Control and Prevention et al. 2024) for 2022, which is derived from 2018–2022 Census data and reported at the Census tract level. The SVI ranks each tract within a state according to vulnerability along four dimensions: (1) socioeconomic status, (2) household characteristics, (3) racial and ethnic minority status, and (4) housing type and transportation. Because labor force participation implies socioeconomic status, only the latter three dimensions are used in the analysis for controlling for social vulnerability.

The “household characteristics” index uses information on the following categories, using prevalence in local population to rank each tract: aged 65 and older; aged 17 and younger; civilian with a disability; single-parent households; and English-language proficiency (Centers for Disease Control and Prevention et al. 2024). The “racial and ethnic minority status” index uses the proportion of local population that belongs to one or more minority groups (Centers for Disease Control and Prevention et al. 2024). Finally, the “housing type and transportation” index uses information on multi-unit structures, mobile homes, crowding, vehicle ownership, and group-quarters dwelling (Centers for Disease Control and Prevention et al. 2024).

Because some industries do not follow the traditional nine-to-five schedule, we additionally incorporate the 2021 sample of Longitudinal Employer-Household Dynamics (LEHD) (U.S. Census Bureau 2024a) and use the local place-of-employment concentration of those industries steeped in shift work. Shift work may provide households with flexibility in coordinating child care and participating in the labor force, especially for two-parent households (Presser 1986). However, for those engaged in shift work with varying schedules, like in retail, hospitality, and entertainment, varying schedules may hinder child care accessibility and therefore reduce labor force participation (Henly and Lambert 2014). In this analysis, employment in hospitality, retail, entertainment, or manufacturing is considered a proxy for shift work. The share of employment in shift work industries is calculated using Census ACS employment by industry.

With the exception of child care providers’ exact geographic coordinates within Michigan, all data used in this study is freely and publicly available in other regions and often in longitudinal form. However, providers’ latitudes and longitudes can be generated from their addresses, which are readily available on MI LARA, and are generally available from other child care-regulating authorities.

3. Methods

Our outcomes of interest—labor force participation for women and for families—are all measured by the number of participants relative to the working age population, aged 16 and over, of the respective Census tract populations (single-parent—male or female—or two-parent households). While this can be measured as a proportion, the more prudent modeling approach is to employ methods such as binary logistic regression to model the probability of success (labor force participation), which we do here. This allows us to use the number of people in and out of the labor force as counts in a binomial distribution rather than the Gaussian distribution assumed when using the ordinary least squares (OLS) approaches. As applied here, binary logistic regression is used to gauge neighborhood resources and families’ proximity (Chakrabarti and Joh 2019; Clark et al. 2018; Klein et al. 2017); most relevant is Blumenberg et al.’s (2024) examination of parental travel and child care availability.

A conceptual model can be specified in linear form as follows:

$$y=\gamma D+\beta \mathrm{X}+\delta +u ,$$

where $y$ is a vector of labor force participation rates by Census tract, $\beta \mathrm{X}$ is a linear combination correlating the set of control variables $\mathrm{X}$, and $D$ is a vector of distances from the centroid of Census tracts to the nearest child care facility measured in miles. The coefficient $\gamma$ is the primary coefficient of interest and represents the effect of the distance to a child care center on workforce participation. The term $\delta$ denotes a spatial process for recognizing spatial dependence. The model specification we prefer3 is a spatial lag model such that $\delta =\rho \mathrm{W}y,$ where $\mathrm{W}$ is a spatial k-nearest neighbor weight matrix. The $Wy$ term creates an average value of the k-nearest neighboring tracts’ labor force participation (expressed as a proportion) to account for local spatial autocorrelation. The term $\rho$ is then the spatial autocorrelation coefficient to be estimated within the logit model. The expected sign of the spatial autocorrelation cannot be determined a priori. In one context, the sign may be expected to be positive, as peer effects across nearby neighborhoods may contribute to similar household preferences for labor force participation. Alternatively, a minimum efficient scale may suggest that having child care facilities in one neighborhood would reduce the chance of having one in surrounding neighborhoods. Therefore, the presence or absence of spatial correlation should be tested to determine the appropriate model specification.

While we focus on the association of distance to child care and labor force participation, the extent of labor force participation by young-parent households may encourage child care business formation. Additionally, households with child care needs may choose to locate near child care centers. In the absence of any natural experiment affecting child care provision across the state, we cannot make causal claims. We instead seek to describe the existing relationship between access to care and labor force participation and provide suggestive evidence of how improving access would impact the labor force.

4. Results

The distance that a family must travel to reach a given child care provider can be thought of as a time cost, in addition to the financial cost that the household pays to have their children looked after. Because of this, we would expect that higher levels of access (here expressed as shorter distances) should result in higher levels of labor force participation, as the effective costs of child care utilization are lower. It is important to note that families typically include women, and so the effect on female labor force participation partially includes those predicted effects on families. The overall effect of a policy that improves access to child care facilities is its impact on female labor force participation. Family-level predictions illustrate the extent to which these increases affect parents’ employment as a portion of the overall effect.

Because families can commute across multiple tracts both for work and for potential child care provision, we may be concerned about the possibility of spatial autocorrelation in our model. Positive spatial autocorrelation seems most reasonable due to urbanization; neighborhoods in proximity are more likely to share similar characteristics, including employment characteristics. With this in mind, we first estimate a non-spatial model and use Moran’s I, a measure of spatial autocorrelation, to test the presence of spatial autocorrelation. Should such autocorrelation be found to exist, the preferable model will account for and correct this with a spatial term.

Table A1. Tests for spatial autocorrelation correction.

	No Spatial Term			Spatial Lag
	Moran’s I	Z-Score	Probability	Moran’s I	Z-Score	Probability
Female LBF	0.172 *	15.620	0.000	−0.095	−8.581	1.000
2-parent LBF	0.067 *	6.132	0.000	−0.017	−1.477	0.930
1-parent LBF	0.074 *	6.779	0.000	0.027 *	2.436	0.007

* p < 0.05. All probabilities are derived from a one-sided (positive autocorrelation) test. The spatial lag specification employs a k-nearest neighbors proximity matrix for a tract’s five nearest neighbors, with the average labor force participation of the five nearest Census tracts used as the spatially lagged term.

in Appendix A demonstrates the results of our Moran test, with and without applying a first-order spatial lag on k-nearest neighbors, demonstrating that the magnitude of the Moran’s I values is reduced for all measures of interest after introducing a spatial lag term. Thus, for the remainder of the paper, we control for spatially lagged values (according to a k-nearest neighbors proximity matrix, using each tract’s five nearest neighbors, and varying this as a robustness check) to account for this autocorrelation.

Our controls allow us to make appropriate comparisons between different tracts. High school completion appears to be associated with higher rates of labor force participation in all regressions, suggesting higher opportunity costs of not participating in the labor force. There is also a positive relationship with shift work, though the association is not as strong. Most importantly, in all cases, both our measure of child care access and our spatial lag term yield consistent results. Higher nearby labor force participation is positively correlated with a tract’s own participation rates, and the further away a Census tract is from the nearest child care provider, the less likely that women and parents of young children will join the labor force or be employed. As the coefficients reported in

Table 2. Binomial logistic (logit) regression of Census tract characteristics and the distance to the nearest child care provider.

	Female LBF (1)	2-Parent LBF (2)	1-Parent LBF (3)
Distance to nearest child care provider (mi.)	−0.027 *** (0.001)	−0.022 *** (0.003)	−0.017 *** (0.005)
Spatially lagged outcome variable	2.089 *** (0.018)	0.885 *** (0.024)	0.534 *** (0.034)
Proportion of tract population that is White	0.136 *** (0.008)	−0.226 *** (0.033)	−0.019 (0.036)
Proportion of tract population with HS degree or higher	0.999 *** (0.022)	3.820 *** (0.071)	2.143 *** (0.085)
Proportion of tract population with bachelor’s degree or higher	−0.170 *** (0.008)	0.030 (0.024)	0.359 *** (0.051)
Proportion of tract employment found in “shift-work” industries	0.033 *** (0.005)	0.176 *** (0.016)	0.020 (0.024)
Proportion of tract within an urban area	−0.021 *** (0.004)	−0.194 *** (0.011)	0.135 *** (0.022)
SVI theme ranking: Household Characteristics	−0.289 *** (0.004)	−0.108 *** (0.014)	−0.064 *** (0.023)
SVI theme ranking: Minority Status	0.365 *** (0.008)	0.124 *** (0.027)	−0.196 *** (0.047)
SVI theme ranking: Housing and Transportation	−0.073 *** (0.004)	−0.106 *** (0.013)	−0.030 (0.021)
Constant	−1.839 *** (0.022)	−3.326 *** (0.07)	−0.899 *** (0.095)
Observations	2889	2722	2592

Standard errors appear below the parameter estimates, in parentheses. *** p < 0.001.

represent how the log-odds ratio is affected, they also approximate percentage changes in labor force participation. A one-mile increase in the distance to the nearest child care provider is associated with a decrease in female labor force participation of about 2.7%, on average. Results are similar but smaller for both parents being in the labor force (2.2%) and single parents being in the labor force (1.7%).

Robustness Checks

Our results are robust to the number of neighbors used when determining how to construct the spatial lag term;

Table 3. The coefficient on distance to nearest child care provider for different potential specifications.

	Fem. LBF	2-Parent LBF	1-Parent LBF
Preferred	−0.027 ***	−0.022 ***	−0.017 ***
	(0.001)	(0.003)	(0.005)
Nearest 3 Neighbors	−0.029 ***	−0.022 ***	−0.017 ***
	(0.001)	(0.003)	(0.005)
Nearest 10 Neighbors	−0.026 ***	−0.025 ***	−0.016 ***
	(0.001)	(0.003)	(0.005)
Serves Kids Under 6	−0.027 ***	−0.022 ***	−0.015 ***
	(0.001)	(0.003)	(0.005)
Full-Time Only	−0.024 ***	−0.014 ***	−0.007
	(0.001)	(0.003)	(0.005)

Standard errors appear below the parameter estimates, in parentheses. *** p < 0.001. The first row features our preferred, spatially lagged specification, also detailed in Table 3. Rows 2 and 3 vary the number of nearest neighbors each tract is affected by, both lowering (to 3) and raising (to 10) the amount. Row 4 excludes the 2.3% of providers that do not serve children under six years of age (younger than school-age), while row 5 considers only providers that are reported to offer services “full time” when calculating distance. In both alternate specifications, all distances are weakly greater than those in our main specification, which includes all providers. Coefficients reflect changes in the log-odds ratio.

includes alternate specifications, in which we employ either the three or ten nearest neighbors to every tract, rather than five, as used in the main analysis. Figure 2 demonstrates this information visually, showing the results of regressions using a tract’s one-to-fifteen nearest neighbors. As the first three rows of Table 3 and Figure 2 demonstrate, our findings change very little based on the number of neighbors brought into the spatial lag (or error) term.4

We additionally conduct robustness checks regarding subsets of child care providers. First, we consider including only child care providers that explicitly serve children under 6 years old. In families with young children, while those children are too young to attend school, someone (whether a parent, family member, or professional) must look after them. While our main analysis uses all child care centers to calculate the minimum distance, about 2.3% of providers do not serve children under 6 years old. When using only providers that do serve children under 6, our results scarcely change—only decreasing somewhat for single-parent households (for a predicted effect of 1.5% rather than 1.7%).

Second, we examine child care providers that are open full-time as opposed to part-time. When considering how different types of providers might affect the local labor supply differently, it seems reasonable that full-time child care centers could be more important to family employment choice than part-time centers, as they could be more flexible for prospective working families. Using information on full-time centers only, while controlling for the mix of providers within 10 miles of the Census tract, all outcomes exhibit the same negative relationship as when considering all providers, though the size of our estimates shrinks somewhat. The relationship between provider distance and single-parent labor force participation becomes statistically insignificant, though this is logical—among all study groups, we would expect single-parent labor force participation to be the least elastic, which is borne out in all our results.

5. Policy Implications

Our findings support other research that shows that access to child care promotes individuals’ decisions to seek employment and that it does so at the community level, exhibiting the inverse association between child care availability and workforce participation. While the causal relation may not be clear, the association begs the question, what level of employment might we expect if such gaps in access did not exist? We can conjecture about this by simulating a change in access via the estimated binary logistics model. Accordingly, we estimate lost labor force participation under current gaps relative to a case where no household is more than eight miles and no more than five miles away from a child care facility. This predominantly reflects differences in urbanicity, as few urban areas require a five-mile or more commute to the nearest child care facility. As shown in Figure 3 and exhibited in

Table A2. Counties and tracts further than five miles from the nearest child care provider.

County: Census Tract	Iron: 1 *, 2 *, 5
Alcona: 1, 9705 *, 9706	Kalkaska: 9504 *, 9506.01, 9506.02 *
Alger: 1 *, 2, 3 *	Keweenaw: 1 *
Allegan: 310.02	Lake: 9601 *, 9611, 9613 *
Alpena: 1.02	Leelanau: 9701, 9704
Antrim: 9602	Lenawee: 620
Arenac: 9701	Luce: 9601, 9602 *
Baraga: 1 *, 2 *	Mackinac: 9502 *, 9503, 9504, 9505 *
Bay: 2862	Marquette: 12, 13, 23, 26
Berrien: 114, 115	Mecosta: 9603
Branch: 9510, 9511	Menominee: 9601, 9602, 9603
Cass: 10	Missaukee: 9601.01 *, 9601.02
Charlevoix: 1, 9	Montcalm: 9703
Cheboygan: 9604.02, 9605, 9606, 9607 *	Montmorency: 9101, 9102 *, 9103
Chippewa: 9706.02, 9707 *, 9710	Newaygo: 9701 *
Clare: 2 *, 3 *, 5, 7, 8	Oceana: 106
Crawford: 9601, 9604 *	Ogemaw: 9502, 9503
Delta: 9701, 9702	Ontonagon: 9701 *, 9702, 9703 *
Dickinson: 9501 *, 9502	Osceola: 9705.01
Emmet: 9701, 9703.01	Oscoda: 9702.01, 9703, 9704, 9705
Gladwin: 1.02 *, 2 *, 3 *	Otsego: 9501
Gogebic: 9502 *, 9508 *	Presque Isle: 9501 *, 9503 *
Grand Traverse: 5502	Roscommon: 9704.02, 9705, 9706, 9707, 9711 *
Houghton: 10 *, 7	Saginaw: 125
Huron: 9501, 9512	Sanilac: 9701, 9702, 9703 *
Ionia: 301.01	Schoolcraft: 1 *, 2 *
Iosco: 1.01 *, 7, 9	Wexford: 3803

This table consists of all 114 Census tracts throughout Michigan that are more than five miles away from the nearest licensed child care provider, sorted by county. The 39 tracts marked with an asterisk (*) are more than eight miles away from the nearest provider.

of Appendix A, much more of Michigan has restricted access to child care facilities within a distance of five miles than that with restricted access within eight miles. All dark areas are strictly non-urban areas for both.

It is important to note that these predictions should not be interpreted as forecasts of what would occur if Michigan were to mandate maximum child care travel distances. This paper provides descriptive data, not causal conclusions, as child care demand is shaped by a multitude of factors. We have included these simulations as simple derivations to illustrate the potential for job creation that is currently unfulfilled due to statewide limited access to formal child care facilities.

We use three specifications for these estimates. First, we employ an approach that does not account for spatial spillovers. While simulating workforce outcomes this way does not use the full predictive power of the model, the simulation calculations are much more manageable. We also use two spatial specifications, both employing a k-nearest neighbors approach using each tract’s five nearest neighbors (our preferred specification). Our first spatial specification incorporates but does not vary the spatial lag term. This underestimates the impact of any particular policy. In the “iterative” row, we use successive predictive iterations until the sum of squared differences from one iteration to the next is less than 0.001. Effectively, we iterate until the predicted outcomes converge to a steady level.

Table 4. Estimated impacts of setting a “ceiling” on the maximum distance a tract can be from its nearest child care provider.

	8-Mile Maximum Distance			5-Mile Maximum Distance
	Female LBF	2-Parent LBF	1-Parent LBF	Female LBF	2-Parent LBF	1-Parent LBF
Non-spatial	1220	45	11	3529	146	35
Spatial	740	39	11	2137	126	36
Iterative	3101	42	12	5643	820	1184
Eligible population	37,014	2875	1444	116,114	9874	4509

Table 4 outlines the estimated impacts of setting a “ceiling” on the maximum distance a tract can be from its nearest child care provider, at either eight or five miles. The first row uses a non-spatial model, the second uses a spatial model without varying the spatial lag term, and the third row uses iteration to allow the predicted effects to converge, including the positive “feedback” effects of such a policy across neighborhoods. The bottom row includes the population of individuals/households that are located in areas that would be affected by the policy.

shows policy simulation results. Using the non-spatial approach, if the furthest a parent had to travel was eight miles to access formal child care, we estimate a labor force participation increase of 1220 more women, statewide. If the furthest distance was reduced to five miles, we instead estimate that 3529 more women would join the labor force, about three times the size of the previous estimate. These results suggest that such a policy of setting a maximum distance required to locate a child care facility has the potential to strongly impact the employment opportunities in rural communities.

Using a spatial approach produces much larger estimated changes by incorporating positive spatial autocorrelation; in particular, a maximum distance of eight miles suggests that the female labor force participation would be 3101 higher, while that with a limit of five miles would be 5643. Similarly, the simulations suggest that two-earner households would likely be more common.

Relative to the non-spatial estimator, the spatial autocorrelation estimator diffuses the effect of distance to a child care center across surrounding tracts. Without updating labor force participation rates nearby, our naïve spatial estimate (with a smaller coefficient on distance to child care facilities) underestimates the magnitude of any policy change. However, once the simulations consider how improving access to child care in one tract has not only direct effects on families in that tract but also improves accessibility in neighboring tracts (see the iterative estimates in Table 4), the simulated effect on labor force participation is much larger.

While only about 0.9% of Michigan’s population lives in tracts more than eight miles from the nearest child care provider and about 2.9% lives in tracts more than five miles away (Figure 3), these estimates suggest that thousands of women either in the labor force or potentially contributing to the labor force may be constrained due to limitations in access to child care facilities. The implications suggest that improving access to child care facilities by placing more child care centers across the state may alleviate some constraints on labor force participation, specifically with younger households with children. The evidence suggests that both single-parent and two-parent household participation is inversely associated with distance to licensed child care centers. This exercise promotes communicating the economic challenges of child care constraints on local labor markets and provides a basis for gauging the severity of the effect on the workforce.

6. Limitations

The measure of child care access used in this study only considers the distance that parents would potentially need to travel to take their child to a given center. However, many other factors may influence whether families can use the closest child care facilities. For instance, the analysis does not consider the capacity of individual providers to take on additional children. Additional factors may constrain access to identified facilities, including mismatch between service provision and users’ needs (hours, service age, learning philosophies, etc.). While the data we use does contain information about providers’ licensed capacities, facilities have varying levels of true capacities that do not reflect their licensed capacities, which may vary due to staff shortages, age limitations, the current mix of children, and other issues common in this sector. It is likewise difficult to outline the market that a particular child care provider occupies and estimate how many children and families might feasibly use the provider’s services.

The model used in this analysis is likewise agnostic on the effect of pricing on household access to child care. The expected cost of child care is likely to vary by market and by type of facilities (full- or part-time, off-hours, etc.). Consistent with economic theory, it is likely that the cost of child care is higher in markets with limited geographic access to child care facilities. However, that relationship is not tested in this study. Along similar lines, households in high-cost markets or with multiple children will likely face disproportionate financial and logistical constraints in access to commercial child care services that may not be captured by the minimum distance measurement used here for gauging child care accessibility. Beyond limiting the interpretation of the measured effect here, approaches that effectively control for these factors will likely improve the statistical precision but not necessarily the magnitudes of the estimates reported here.

Finally, this analysis does not assess how parents ultimately choose where to send their children, where parents may have choices beyond licensed day care. Such choices may include informal care, such as that supplied by extended family members and grandparents, though it is difficult to collect data on such arrangements. More robust data-collection strategies from regulatory agencies or through relatively more expensive surveys of households would provide more and better data to understand such decisions, though it would need to be supplemented through individual- or household-level analyses. Such work is beyond the scope of our paper but is critical to better understand this policy environment in the future.

7. Conclusions

While Michigan’s child care deserts—as well as America’s larger child care crisis—deserve long-term governmental interventions, focusing on areas with limited spatial child care accessibility may help reintroduce parents back into the workforce while decreasing child care-related workplace absenteeism and turnover. Considering the extent of post-COVID-19 labor-supply shortages, especially in rural communities, improving access to child care may be of critical importance to jumpstarting local economic and community development. The literature shows that access to child care is critical for families of young children and can play an important role in the labor force participation decisions of these families. While previous research has been carried out to investigate the effects of such policies on children’s socio-educational outcomes, less has been done to link child care availability to local economic growth at the local level.

Across the state of Michigan, we found that proximity to child care providers is significantly associated with lower labor force participation for female workers and parents of young children. In addition to expanding existing locations, our paper suggests that policy levers leading to new locations could have strong effects in reversing some of the workforce participation constraints observed throughout the U.S., especially in the wake of the pandemic.

Furthermore, this study has significantly highlighted the inverse relationship between proximity to child care facilities and labor force participation in Michigan, particularly for women and parents of young children. The findings confirm the existing literature on the importance of child care access for families and expand upon it by demonstrating the community-level effects on labor market size and economic growth. This analysis clearly indicates that child care constraints are particularly relevant for rural areas, such as Michigan’s Upper Peninsula.

Using commonly accessible local data, this study shows the spatial intersection between child care availability and parental employment and goes on to provide a measure of the contribution of greater access to area employment and workforce participation. Especially in but not limited to rural areas, access to child care facilities may be a limiting factor for workforce participation, and anticipating outcomes of policy changes is the first step in implementing progressive policy directives. This research offers a practical framework for policymakers to assess potential economic growth constraints due to limited child care access and provides a basis for future progressive policy directives aimed at addressing these constraints, at all levels of the government.

The use of Census tract-level data in this analysis provides spatial granularity often missing in other studies, making this approach more effective for local policy considerations. The approach demonstrated here is a cost-effective way of introducing a locally relevant descriptive analysis of the locus of access to child care and employment that any community or state can implement to pinpoint which communities are most impacted by employment disruptions due to low child care access. Such details are often missing from other analyses but can be instrumental in targeting and prioritizing policy interventions.

Given these insights, future research should focus on a deeper exploration of the causal mechanisms and a more nuanced understanding of the multifaceted barriers to child care access. Specifically, future research could investigate the combined effects of child care costs and travel distance, which would offer a more comprehensive measure of accessibility barriers. Furthermore, it is crucial to account for the actual availability of open slots at child care facilities, as relying solely on licensed capacity might misrepresent the effective geographic distance households need to travel. Developing methodologies to assess these causal associations at the local level will significantly enhance the policy implications and provide more direct responses for improving workforce participation among young households. Such advancements will move beyond descriptive analysis to offer actionable insights for fostering a more resilient and engaged workforce.