�鶹������ý

The child care desert measure has been around for nearly a decade and has become a standard tool in policy conversations. What made you want to examine whether it was actually doing what it’s supposed to do?

I have often wondered whether child care deserts are truly identifying areas where it is the most difficult for parents to find care. Anecdotally, we often see stories about long waitlists in urban areas, but deserts are overrepresented in rural areas.

Deserts are defined as areas where there are three or more children per licensed slot. Having a low number of slots per capita could signal insufficient care, but it could also signal differences in parents’ preferences for staying home with young children or using informal care. So if parents in cities are more likely to want formal care than parents in rural areas, for example, using the same threshold everywhere when defining insufficient access could lead to misidentification.

I finally decided to tackle this project after receiving a call from a reporter about how Utah is the worst state in the country for child care access because it has the most child care deserts. I explained that parents in Utah may have different preferences for nonparental care compared to parents in other parts of the country, and so it’s not clear that Utah actually is the worst off. But I wanted to be able to offer a better answer in the future based on empirical evidence, not just speculation, so that’s when I got to work!

While the ideal way to measure child care access would take both supply and demand into account, you note that demand for child care is quite difficult to measure. Why is that?

The first challenge here is that we do not have great data on families’ child care choices or providers’ enrollment, so, in general, we do not have the necessary data. But there is also a deeper challenge, which is that while we could potentially observe parents’ child care choices, we cannot directly observe their preferences. What we would like to know is: if I were to drop a new child care center into a county, how many children would enroll? In other words – is the reason they were making a different child care choice because they could not find an available slot or was it because, given current prices, they preferred their selected care option? It is very difficult to answer that question.

In the simplest economic model, if a family’s most preferred care type at current prices is center-based care, then their child would be enrolled in center-based care (and thus this whole discussion of child care deserts would be moot, actually!). But we know that there are waitlists in the child care market, so that may not be true. For example, when we first moved to South Carolina, I was looking for center-based care for my two children. However, the centers we toured were full, so we ended up hiring a nanny. All that a researcher would be able to observe is that we hired a nanny – they wouldn’t know that we would have chosen center-based care if we had found a slot!

What did you find when you tested whether desert status actually predicts whether parents in those areas are having difficulty finding an open slot?

For the most part, I find that child care desert status does not predict whether providers have open slots. Specifically, I test whether child care desert status predicts center-based provider vacancy rates from the 2019 National Survey of Early Care and Education (NSECE), which includes a nationally representative survey of providers. If care is more difficult to find in child care deserts, then I would expect that providers in these areas would be less likely to have an open slot and would have lower vacancy rates compared to providers in non-desert areas.

But with one modest exception, I find that vacancy rates are not related to child care desert status. In other words, areas designated as “child care deserts” do not seem to have more pent-up demand for care than non-deserts. The one exception is for one-year-old classrooms: providers in child care deserts have one-year-old vacancy rates that are about four percentage points lower than providers not in child care deserts. So while on average, providers have a 13 percent vacancy rate for one-year-olds, in deserts, it’s nine percent. But there is no consistent relationship across the other four one-year age bands (infant, twos, threes, fours).

You tested two alternative measures that adjust for local demand rather than applying a uniform national threshold, including the, which compares the number of licensed slots to the number of children with all parents working. Could you describe both approaches and how they performed compared to the standard desert measure?

The goal with both approaches is to allow the threshold for “low access” to vary based on an estimate of local parents’ demand for center-based care. The short answer is that I find that these alternative measures actually are associated with lower vacancy rates!

For the first measure, I approximate the Buffett Institute’s Child Care Gap by comparing licensed slots in a county to the number of young children with all parents in the labor force. The idea is that a key piece of demand is coming from families with working parents, so this approach may provide a better approximation for the size of potential demand than including all young children. I take the 20 percent worst-performing counties by this metric and label them as having a Child Care Gap (CCG).

For the second alternative measure, I create a prediction for each county of the expected number of slots per child under five based on supply in other counties with similar demographic characteristics. The idea is that if individuals with similar demographics have similar demand, then the number of slots in demographically similar areas may provide a good approximation for how many slots we might expect. One advantage of this metric is that it may be able to take into account not only preferences for work but also preferences for formal care specifically. I compare the predicted slots to the actual slots and designate the 20 percent with the largest gaps as having demographic-adjusted low supply (DALS).

Both CCG and DALS are strongly predictive of low vacancy rates for infants and toddlers. For example, providers in CCG or DALS counties are over 40 percent less likely to have an infant vacancy than providers in other areas. The results are similar, though smaller in magnitude, for one-year-olds and two-year-olds. These relationships are generally slightly larger for DALS compared to CCG, but they are in the same ballpark and highly statistically significant. Overall, these findings suggest that there are areas where parents face more difficulty finding an available slot but identifying them requires adjusting for proxies of demand.

One of the more striking findings is that while none of the measures were particularly strong at identifying access problems for older children, the two alternative measures were quite good at identifying access problems for infants and toddlers, while the standard desert measure was not. Why do you think the alternative measures are better suited to capturing access problems for the youngest children?

This is a great question! I think there are a couple of likely explanations, though I cannot test them directly. First, it is easier for the market to respond if there is a shortage of three- or four-year-old care. Elementary schools can add preschool classrooms, significantly lowering start-up costs. On the other hand, infant and toddler care is much more costly to provide and less easily incorporated into elementary schools. Center-based providers may also be reluctant to add infant and toddler classrooms because, as I show in, those classrooms have the thinnest margins and may even lose money. These barriers to entry may be what allows for persistently low vacancy rates.

Parents also have more care options for three- and four-year-olds, including within public or private elementary schools. By the beginning of the school year, most parents may have already found a slot and may be reluctant to move an older child from an established preschool class mid-year. Providers may thus be able to fill infant and toddler slots mid-year more easily as these families may be more willing to switch arrangements. So vacancy rates for older classrooms may be driven more by idiosyncrasies in the timing of (hard-to-fill) departures, while vacancies in younger classrooms may be due to shorter waitlists (i.e., lower demand). Thus, the vacancy rates in the younger classrooms may be more closely tied to demand than the vacancy rates in the older classrooms.

Can you talk a bit more about why the question of accurately measuring child care access has important policy implications? For example, in the paper you note that the child care desert metric was used to target child care investments when allocating American Rescue Plan Act grants.

Given scarce public resources, policymakers interested in making supply-side investments in child care (i.e., start-up or expansion grants) want to target them to areas most in need. And so, to invest in the right places, we need a good measure of need! The measure that has been available to this point is child care deserts, and so several states, such as,,, and, have or had grant programs specifically targeted to these areas.

But if our measure of child care access is not accurate, funds may not go to areas most in need. We could even end up in a situation where funds are used in an area with very low demand and a new child care center ends up underenrolled. This may be seen as a policy failure and could pose an existential threat to supply-side programs, even while there may be another place where those same dollars could have built a center that would be bursting at the seams. So measurement matters!

What research do you think needs to happen next? Are there gaps in our understanding of child care access that your findings point to, and what would it take to fill them?

Data is always a challenge in this area! More data on provider vacancy rates, in particular, could help pinpoint areas where parents have more difficulty finding care. In addition, my paper uses simple proxies for parent demand. Economists have other tools in their toolkit that allow them to build models of demand to estimate how households would respond if a center were placed in a particular location. Such models could be used to help with targeting resources, though they would be more accurate with additional data, particularly household-level information on families’ child care provider choices.

But my paper provides evidence that it is, in fact, possible to identify areas where it is especially difficult for parents to access child care, even when working with the imperfect data that is available. That there was any relationship between any of these metrics and vacancy rates may actually be the most surprising finding in the paper! We might have expected markets to adjust (i.e., prices or supply to rise) so that vacancies did not systematically vary. Digging deeper to understand what market forces are leading to this disparity could provide policy ideas for how to best address possible supply shortages.