ByEmily A. Gallagher
The most financially stressed families plan to save or pay down debt with a larger share of their tax refund once they become eligible for medical coverage under Medicaid. In contrast, eligibility seems to make no difference in savings rates for the average low-income household that does not meet these stress criteria, as outlined in the previous issue of In the Balance.Throughout our analysis, we focus on Medicaid eligibility (sometimes referred to as “access”) rather than enrollment because enrollment is not particularly relevant when households are implicitly insured. Eligible households are covered retroactively for up to three months prior to the month of application, allowing Medical providers to enroll individuals retroactively and bill Medicaid for the care. See https://www.medicaid.gov/medicaid/eligibility/index.html
Several competing forces may account for the difference in the savings responses. We identified three possible forces: the use of bankruptcy as a way to manage medical bills; a desire on the part of uninsured households to save for future health events; and the impression that one’s Medicaid access is often transient. We measured the impact of those three forces.
First, Medicaid access gives hope to harried families who otherwise despair of ever keeping up with their bills. Such households anticipate that anything saved to cover a future medical bill would be lost in a bankruptcy. Medicaid offers them a reason to save: Households rest assured that they can hold onto their savings even if they are hit with a health problem because the medical bills won’t push them into bankruptcy.
This explanation relies on the fact that households can avoid health care bills by declaring personal bankruptcy. In a sense, bankruptcy acts for some households as a high deductible health plan:For a discussion of this hypothesis, see: Mahoney, Neale. “Bankruptcy as Implicit Health Insurance.” American Economic Review, 2015, Vol. 105, No. 2, pages 710–46. The deductible is proportional to how much the household has saved at the moment of bankruptcy.
In that role, bankruptcy might reduce household savings rates. Among financially distressed households, the increase in savings under Medicaid may reflect a reduced necessity to resort to bankruptcy.
Consistent with this explanation, we documented that in states where households tend to be forced to hand over more of their savings in a bankruptcy, Medicaid has more than twice the positive effect on the propensity to save from tax refunds as in other states.
As alluded to in the previous article, a second possible force has to do with incentives to save as a form of self-insurance against large medical expenses. Knowing that future medical bills will be covered frees up those doing comparatively better to repair their cars or buy school clothes for the kids. Households that are not in financial distress may, therefore, save less when gaining access to Medicaid.
At first glance, however, we do not find evidence in support of this second force. On average, households that are not in financial distress do not reduce their savings when they become eligible for Medicaid.
This result is puzzling. Does the average low-income household not save for rainy day health expenses? Could that explain why we do not observe a lower savings rate for these households when they get Medicaid?
We think that a more likely explanation has to do with the transient nature of Medicaid access for some households. Fluctuations in monthly income combined with frequent income verification by program administrators cause some households to qualify for Medicaid in some months but not in others. By some estimates, more than 30 percent of households lose eligibility within six months of enrollment. This phenomenon, dubbed “Medicaid churn,” is well-documented.Sommers, Benjamin D.; and Rosenbaum, Sara. “Issues in Health Reform: How Changes in Eligibility May Move Millions Back and Forth between Medicaid and Insurance Exchanges.” Health Affairs, February 2011. Vol. 30, No. 2, pages 228–36. Swartz, Katherine; Short, Patricia Farley; Graefe, Deborah Roempke; and Uberoi, Namrata. “Evaluating State Options for Reducing Medicaid Churning.” Health Affairs, July 2015, Vol. 34, No. 7, pp. 1180–7.
Under our hypothesis, households that view their Medicaid access as temporary—available only as long as their income remains below a certain threshold—may not reduce their savings when they get Medicaid access. In fact, they may attempt to help themselves for a future when they no longer have Medicaid by saving a bit more while their medical expenses are covered.
On the other hand, households that view their new Medicaid eligibility as enduring over the long term may have a more negative savings response to gaining access.
Our tests offer some support for this hypothesis. The figure plots the predicted savings response at different likelihoods of being Medicaid eligible from a regression model. Households that are and are not in financial distress (“high distress” and “low distress,” respectively) are plotted separately. The probability that a household will have Medicaid access is calculated using its state Medicaid rules and its demographic information.
This figure shows the predicted percentage point change in household savings rates at different probabilities of Medicaid eligibility. Savings rates are measured as the percentage of the tax refund that a household intends not to consume for at least six months. Medicaid eligibility probabilities are estimated from differences in state Medicaid income thresholds and differences in the income distributions across demographic combinations of parent status, education, age, gender and race. By assumption, Medicaid access is more temporary at lower probabilities, and vice versa. The sample is split at the median according to an index of financial distress. The associated regression model, variables and coefficient estimates are detailed in Gallagher, et al. (2018).Sabat, Jorge; Gallagher, Emily; Gopalan, Radhakrishnan; and Grinstein-Weiss, Michal. “Medicaid and Household Savings Behavior: New Evidence from Tax Refunds.” Dec. 4, 2017. https://ssrn.com/abstract=3052026 or http://dx.doi.org/10.2139/ssrn.3052026
Medicaid access is likely to be perceived as transient by households that have a low probability of eligibility to begin with, such as households in states with a very low income-eligibility limit or members of demographic groups that have a relatively high average income. Even a small change in either rules or income might be sufficient to end their access.
Consistent with this expectation, among households in “low distress,” having a low probability of Medicaid access is associated with modestly higher savings. In particular, a household with a 20 percent probability of Medicaid access saves about 2 percentage points more of its refund than a similar household with no chance at Medicaid.
In contrast, “low distress” households with a high probability of Medicaid eligibility have a substantial negative savings response to gaining access. For example, at an 80 percent simulated probability of Medicaid eligibility, the savings rate is 5 percentage points below that of a similar household without Medicaid access.
To summarize, when a household is doing OK financially and gains enduring access to Medicaid, the household feels free to lower its savings rate. When that same household gains Medicaid in a more temporary form instead, it lifts its savings rate. When all “low distress” households are taken together, these two opposing responses wash out on average.
For households that are financially distressed (“high distress”), the predicted response in the figure is always positive and increases with the likelihood of Medicaid access, consistent with Medicaid obviating the role of bankruptcy as de facto health plan.
One of the most challenging problems facing most families is determining how much, if anything, to save in the face of unknown health expenditure risk. Our data suggest that low-income households are thoughtfully approaching and responding to this complex problem and to changes in their Medicaid access.
Our work helps inform the debate over the effect of Medicaid on a family’s savings. Do families that get such medical coverage save less money, knowing that they won’t need it to pay a big medical bill? Or, do they save more, knowing that their savings won’t be gobbled up by doctors or bankruptcy courts?
The truth is nuanced. But the savings response to Medicaid depends markedly on the degree of financial distress the household is facing. An extra $100 in the bank may make little difference to many readers of this article. To a struggling family, it might mean being able to replace the flat tire on the car that gets a breadwinner to work in the morning. To them, Medicaid is money in the bank.
The effect of Medicaid eligibility on savings behavior is addressed in a working paperSee Sabat, et al. from the Center for Social Development and Olin Business School at Washington University in St. Louis and the Center for Household Financial Stability at the Federal Reserve Bank of St. Louis.
We took advantage of a natural experiment created when some states opted to expand Medicaid eligibility under Affordable Care Act, while others did not, which allowed us to contrast savings behaviors.In particular, states that expanded Medicaid allow able-bodied adults earning up to 138 percent of the federal poverty line to enroll. States that did not expand Medicaid typically have very low income limits. For example, as of 2015, Virginia had an income limit of 45 percent of the poverty line for parents and no access for childless adults.
Our research draws on a database of survey responses from 67,000 low-income households that used a free online tax-preparation software, which is part of the IRS Free File Alliance, when filing their taxes from 2013 to 2017.Statistical compilations disclosed in this document relate directly to the bona fide research of and public policy discussions concerning savings behavior as it relates to tax compliance. Compilations are anonymous and do not disclose information containing data from fewer than 10 tax returns or reflect taxpayer-level data with the prior explicit consent from taxpayers. Compilations follow the tax preparer’s protocols to help ensure the privacy and confidentiality of customer tax data. The households in this data set consented to the use of their anonymized responses for research.
The data include information on tax refund size and how much of it households planned to save, to use to pay down debt or to spend. We used this information to measure their propensity to save or pay down debt rather than consume. The data also allowed us to distinguish which families are eligible for Medicaid. Finally, the data contain rich demographic information as well as a number of measures of financial distress, which allowed us to determine which families are struggling the most to pay their bills.
We studied the share of the tax refund each household intended to allocate to savings or debt reduction rather than to consumption. To the extent that uninsured health shocks reduce the net worth of households that are not eligible for Medicaid, net worth may mechanically be lower for such households. While still being highly correlated with actual savings behavior, our intention-based savings measure does not suffer as much from this drawback as does measuring savings as changes in a household’s assets.