How a Policy Rule Became a Natural Experiment
To know whether higher CACFP reimbursement changed what Illinois home daycare providers offered, it is not enough to compare providers with generous payments to providers with smaller ones. The subject here is the policy transition that moved some initially low-reimbursement providers into school mode after the 2010 Healthy, Hunger-Free Kids Act (HHFK). Following that transition makes the causal question concrete: which providers changed reimbursement status, which remained comparable controls, and what must be assumed before the difference can be interpreted as an effect?
Before HHFK, the reimbursement state was tied to eligibility
The historical CACFP system gave home providers two broad reimbursement levels. Any licensed daycare home could participate at the lower level for qualifying meals and snacks. A provider meeting additional criteria could receive the higher level for all qualifying meals and snacks served to children in care. Other arrangements allowed higher reimbursement for particular children whose households met the criteria. In the language of the study, these provider and child routes produced different participation modes, including school, census, categorical, all-high, mixed, and all-low.
The distinction between those states matters because eligibility was not merely an administrative label. A provider's location, income, or participation in another means-tested program could be related to the families they served, the local demand for care, or the provider's own resources. Those same characteristics could affect whether the provider offered evening care or longer hours even without a reimbursement change. Comparing all high-reimbursement homes with all low-reimbursement homes would therefore mix the policy effect with pre-existing differences.
The reform changed one particularly visible route: school-mode eligibility. Before HHFK was implemented, a daycare home could qualify through school mode if it lived within the attendance boundary of a public elementary school where at least 50% of students were eligible for free or reduced-price lunch. Since implementation in 2011, the relevant boundary could be that of a public elementary, middle, or high school with the same 50% threshold.
That expansion changed the geography of eligibility. A provider did not need to move, and the families in the provider's care did not need to change, for the provider's status to change. A home that sat near a high-poverty school but outside the older elementary-school boundary could become eligible when the set of relevant schools broadened. The study describes this as allowing providers in areas of less-dense poverty, but close to areas of denser poverty, to reach Tier I through school mode.
A simple example shows why this is a useful policy lever. Imagine a home that was all-low in 2009 and was located near several schools. Under the earlier rule, only the public elementary-school boundary counted. After HHFK, a qualifying middle- or high-school boundary could count as well. If the provider was reassigned to school mode, the higher reimbursement schedule changed the provider's economic conditions without requiring the provider to apply for a new kind of childcare license or to choose evening care first. The policy rule came between the provider's baseline operation and the later service decision.
Redetermination changed status as well as the rule
HHFK also required every CACFP-participating daycare home to have its Tier I or Tier II status redetermined. This administrative step was important because the paper reports that many Tier II homes had been misclassified and were moved into Tier I. The problem often involved failure to identify a provider's school-mode status. In other words, the post-reform change was not only a newly permissive boundary rule. It was also a systematic review that corrected earlier classifications.
The study cites a USDA analysis concluding that redetermination was the more important source of the overall increase in Tier I providers than the school-mode rule change itself. That point changes how the reform should be understood. The treatment is not a clean story in which Congress simply redraws one map and every subsequent change follows from the map. It is a policy package: broader school-mode eligibility plus a mandated effort to determine which providers actually belonged in the higher tier.
The package is still useful for the study because it created a sharp change in reimbursement status for many providers. But the mechanism must stay attached to the status transition. The relevant question is not whether HHFK was good or bad in the abstract. It is whether providers who moved from the low schedule to the high schedule subsequently changed their offered meals, schedules, prices, or quality relative to providers whose reimbursement remained low.
This is also why child-level Tier II arrangements are not the preferred treatment for the causal comparison. A provider receiving higher payment only for certain children may have a different mix of claims and a different reason for qualifying. Categorical and all-high modes likewise reflect choices or circumstances that can be endogenous to the provider or the families they serve. School-mode reassignment offers a more defensible source of variation because it is tied to the HHFK rule and redetermination rather than to a provider's decision to enter another safety-net program or to document individual household eligibility.
Two paths from the same starting state
The study narrows the comparison further. It begins with providers who were in all-low mode in 2009. By 2011, some of them had switched into school mode and the high reimbursement schedule; these are the treated providers. Others remained in all-low mode and form the control group. Both groups start from the same observed reimbursement state, but only one group experiences the policy-induced move to the higher schedule.
This is the structure that makes a difference-in-differences comparison possible. Rather than asking whether treated providers are different from control providers at one point in time, the researchers ask whether the change in outcomes from before to after HHFK is larger for the treated group. If dinner offerings rise for both groups by the same amount, the common change is not attributed to the reimbursement. If dinner rises more among providers whose reimbursement moved from low to high, the difference in those changes is the treatment estimate.
The approach gains credibility from comparing each provider with their own earlier operation. A home provider may have a distinctive location, owner, or clientele that is difficult to measure. Provider fixed effects absorb stable differences of that kind. Year effects absorb changes shared by providers in the survey period. The remaining comparison is not perfectly isolated from every possible influence, but it is more focused than a simple cross-sectional comparison of high- and low-tier homes.
The crucial idea is a counterfactual. For a provider who moved to the high schedule, we can observe what happened after the move, but we cannot observe what that same provider would have done at the same time while remaining low-reimbursement. The providers who stayed all-low supply an estimate of that missing path, provided their outcomes would have changed similarly in the absence of the treatment. The control group is thus not a perfect twin for each treated home. It is a benchmark for the change that would otherwise be invisible.
Starting both groups in all-low mode makes that benchmark more defensible. It removes one large source of baseline contrast: the treated provider is not selected from homes that were already receiving the high schedule. Even so, the groups can differ in location, demand, or other circumstances. The design does not claim that the HHFK rule erased those differences. It asks whether those differences were stable or already producing different trends before the reimbursement transition, and then uses the post-reform divergence as the relevant evidence.
The policy package also helps separate a trigger from a provider's ordinary business choice. A provider may make many decisions about menus, hours, and clients, but the change studied here arrived through an eligibility rule and an administrative redetermination. The provider's residence relative to newly relevant school boundaries and the correction of its program classification could move the reimbursement state before the provider decided whether evening care was worth offering. That ordering is what gives the food-to-time mechanism an observable policy input.
The ordering is not the same as saying that the provider had no agency. Providers still chose whether to claim qualifying meals, continue operating, or offer a particular shift. The study estimates the average response among those who remained active and whose reimbursement status followed the relevant path. That is a more precise question than “what does CACFP do?” It is closer to: when an existing participating home is moved from the low to high schedule by this policy transition, how does its offered operation change relative to an otherwise defined low-schedule path?
The pre-reform checks are therefore not decorative statistics. If treated and control homes were already separating in dinner offerings or operating hours before HHFK, a post-2011 gap could simply continue that earlier pattern. Parallel movement before the reform makes the comparison more credible for those outcomes. Balance tests using predetermined outcomes ask a related question: could the later treatment status already be predicted by how the providers looked before the policy? A result that does not reject balance cannot prove perfect equivalence, but it reduces one concrete reason to attribute the later change to selection.
What would change the conclusion? Evidence of sharply different pre-reform trends, a policy affecting only treated providers at the same time, or a systematic reason that school-mode reassignment tracked an unmeasured change in demand would weaken the causal interpretation. Conversely, consistent pre-reform movement and a response concentrated after the reimbursement change would strengthen it. The study reports supportive parallel-trend and balance evidence, but the logic remains conditional rather than absolute.
The treatment path can be described as a sequence of states:
The diagram is deliberately narrower than the full CACFP system. Other eligibility modes existed, and some providers moved among them, but the study's clean comparison keeps the baseline at all-low and focuses on the low-to-high school-mode path. The control is not “a provider who never participates in CACFP.” It is a provider who continues to participate while remaining at the lowest schedule after the reform.
Natural experiment does not mean random assignment
Calling this a natural experiment means that a real-world policy created treatment variation that researchers did not assign themselves. It does not mean providers were placed into groups by a lottery. Providers who gained school-mode status may still have differed from providers who remained all-low in ways the available data do not fully capture. The study therefore maintains an as-good-as-random assumption within the selected all-low group: after accounting for the design, the HHFK-driven reassignment should not be systematically connected to the potential change in a provider's service outcomes.
That assumption is plausible for a stronger reason than the simple timing of the law. The researchers examine whether the treatment and control groups were already moving differently before HHFK, and they test predetermined outcomes from before the reform. Those checks belong to the evidence about the design, not to the legal text of the reform. They cannot prove that every unobserved difference is absent, but they can reveal whether the groups display obvious pre-existing divergence in the outcomes that later appear to respond.
There are also limits built into the historical records. Complete Illinois school-catchment maps for the HHFK era were unavailable, so the researchers could not construct an alternative control group based on each provider's counterfactual school-mode eligibility. They also could not identify every incorrect mode assignment before HHFK. The study uses school districts as the available geographic level and clusters its standard errors there, even though the policy variation is more closely tied to school boundaries.
These gaps do not erase the value of the design, but they determine the language it can support. The policy transition is strong enough to make a causal estimate more credible than a before-and-after anecdote. It is not strong enough to justify saying that every provider was randomly assigned, that no other policy change mattered, or that the estimate automatically applies to providers who entered, exited, or never participated in CACFP.
The reform therefore serves two roles at once. Institutionally, it changed who could receive the high food reimbursement. Analytically, it created two paths from a shared low-reimbursement starting point. The economic mechanism explains why that change might affect dinner and the clock; the comparison design determines whether the subsequent difference is plausibly attributable to the reimbursement rather than to ordinary provider choice. The next question is what the researchers could actually observe along those two paths.