In 2021, half of Americans reported delaying or foregoing health care due cost reasons (Kearney et al., 2021). Significant efforts have been devoted to addressing cost-related barriers to care, including improving price transparency, subsidizing care, mandating free preventive care, insurance reform, alternative payment models, billing reform, and antitrust efforts. These policies mainly aim to reduce the price of health care; however, an additional policy option to improve affordability is to relax consumer budget constraints. Providing income is a policy option that is gaining popularity as a potential solution to a variety of societal problems (Hoynes and Rothstein, 2019). In the context of health care, households without liquid assets to pay for deductibles, co-payments, and other out-of-pocket costs may be especially likely to delay or forego care.

In this paper, we study the effect of a large cash distribution on health care consumption in a sample of individuals with employer-sponsored health insurance. Cost-related barriers are a problem for this group as 47% of commercially-insured individuals report difficulty affording medical care (Kearney et al., 2021). We study the effect of this cash distribution on a specific type of health care use – prescription medications. Prescription medication is an essential component of medical care, used in cases of immediate need for treating acute conditions as well as for long-term management of chronic diseases. Compared to other forms of medical care, patients generally have more control over their use of medication. Moreover, consumption of medication may be constrained as one in four Americans report difficulty affording their prescription medication (Kearney et al., 2021).

We exploit the arrival of the Alaska Permanent Fund Dividend (PFD), which is an annual cash infusion received by all Alaska residents. The magnitude of the payment is large at approximately $1,500 per household member and comprises 6% of the average family's annual income. It is both universal (i.e., given to nearly everyone)1 and unconditional (i.e., no restrictions on how the funds are used). Our data is the 2013 to 2019 MarketScan Research Outpatient Prescription Drug Database containing prescriptions for Alaska residents with employer-based health insurance. The sample contains 742,711 days of prescription fills by 50,866 individuals. The sample is large, comprising nearly 5% of Alaskans who have employer-sponsored health insurance in 2013 (Table 1). Our identification strategy leverages the fact that the dividend is disbursed on the first Thursday of October to create a withinAlaska comparison group. We compare changes in prescriptions in the 14-day window on either side of the October PFD distribution to changes in prescriptions in the 14-day window on either side of the first Thursday of other months of the year in a difference-in-differences design. Our main analysis relies on within Alaska variation, which cannot rule out an October effect.2 Therefore, we expand our sample to include other states and use a synthetic control analysis to find a convex set of states that act as an Alaska counterfactual.

We find no effects of the PFD on the total number of prescriptions filled and can rule out effects larger than 0.5 % in the week of the Alaska PFD and 1.4% in the week after. Individuals who face higher out-of-pocket costs for medication may be more sensitive to changes in liquidity, so we estimate effects separately for branded versus generic prescriptions and for groups that face different degrees of cost sharing. Individuals with high deductible health plans reduce prescriptions in the two weeks prior to the PFD then increase prescriptions in the week after the PFD; however, the latter result is not statistically significant and may be due to noise in the data so we encourage caution in interpreting this result. We find no changes in generic or branded prescriptions, for individuals not enrolled in a high deductible health plan, or for individuals with varying levels of out-of-pocket costs. Receipt of the PFD has been found to increase alcohol and drug related mortality (Evans and Moore, 2011) and arrests (Watson et al., 2019), suggesting that behavioral changes could cause consumption of acute condition medications or first-time prescriptions to increase. Likewise, if patients are delaying drugs required to manage an ongoing health concern, evidence would be found through changes in refills and prescriptions for chronic conditions. We find no evidence of changes in consumption of acute, chronic, first prescriptions, or refills around the date of the PFD. Lastly, we conduct a synthetic control analysis using a control group of individuals with employer sponsored health insurance from outside Alaska. Results of the synthetic control analysis are consistent with the findings of our main analysis, showing no effect of the PFD on prescription use.

We contribute to the relatively small literature on the effects of cash transfers on health care consumption. Past research finds that liquidity increases are associated with increased visits to medical providers (Holmes et al., 2018), hospitalizations (Dobkin and Puller, 2007;

Forget, 2013; Gross and Tobacman, 2014; Holmes et al., 2018), and prescriptions (Lyngse, 2020; Gross et al., 2022); however, increases in liquidity are not always associated with changes in health care consumption. Using a randomized-control trial design, Jaroszewicz et al. (2022) find no impact of payments to individuals in poverty on health expenditures. To our knowledge, we contribute the first estimates of a cash distribution on prescription use among individuals with employer-sponsored insurance. Two previous studies of prescription use have focused on targeted transfers to elderly and low income groups (Gross et al., 2022; Lyngse, 2020). In a population of Medicare Part D enrollees, Gross et al. (2022) find that on the day that Social Security checks are received, prescriptions increase 6% to 12% among beneficiaries subjected to the most cost sharing. However, dually eligible Medicare-Medicaid beneficiaries subject to no cost sharing do not increase prescriptions and the authors can rule out increases larger than 0.4%. Using data on Danish welfare recipients, Lyngse (2020) finds that the propensity to fill a prescription increases by 52% on transfer income paydays. In contrast, we can rule out relatively small effects of the PFD in the first two weeks after the distribution.

Our study highlights the fact that takeaways from means-tested programs that typically go to the most disadvantaged may not apply to the broader population. Many factors play a role in liquidity sensitivity. Ex ante, it is unclear whether responsiveness to liquidity would be higher or lower in the employer-insured population than in the populations studied in the emerging literature by Gross et al. (2022) and Lyngse (2020). These populations are different with respect to insurance coverage, income, age, and health. Our results are similar to the no copay group in Gross et al. (2022), suggesting that both insurance and income may be important modifiers of the effect. Even within a particular cash transfer, significant heterogeneity in responses across groups with different characteristics are observed. The marginal propensity to consume from the PFD is U-shaped, with low-liquidity, low income households and and high liquidity, high income households consuming the largest fraction of the distribution (Kueng, 2018). Similarly, a recent paper by Bayer et al. (2023) shows that the stimulus checks during the COVID pandemic that were based on unemployment status had a much larger multiplier effect -1.5- than the ones that were unconditional -0.25making the conditional transfer's effect six times as large. This difference is driven partially by the differences in marginal propensity to consume by the two groups as the unemployed are much more liquidity constrained.

In addition to the differences in population characteristics between this paper and Gross et al. (2022) and Lyngse (2020) described above, the nature of the distribution (i.e., annual and not through a social welfare program) may impact how it is spent via mental accounting, in which consumers categorize income and expenses into different mental accounts in order to track their finances (Thaler, 1999). For example, Kooreman (2000) finds that the elasticity of children's clothing expenditures is more sensitive to changes in the designated childhood allowance than other income sources and Hastings and Shapiro (2018) find that the marginal propensity to consume food is much higher from Supplemental Nutrition Assistance Program (SNAP) benefits than from cash, even for households for whom SNAP benefits are economically equivalent to cash. Likewise, monthly transfers may elicit a different behavioral response than an annual distribution. In an experiment, Chambers and Spencer (2008) find that experiment participants elected to spend more of tax refunds delivered monthly than annually. In comparing spending responses to changes in payroll taxes withheld to lump-sum payments, Coronado et al. (2005) finds similar propensity to spend while Sahm et al. (2012) find more propensity to spend from a lump-sum payment. Notably, Sahm et al. (2012) show that a majority of households did not notice the change in withholding that they study, potentially blunting any effects the monthly income change would have on spending.

If consumers view funds received from the PFD differently than monthly government assistance, then the PFD is perhaps most comparable to a tax refund. Recent evidence on the relationship between tax refunds and health expenditures comes from Holmes et al.(2018), who uses transaction-level data from 1.2 million accounts receiving a tax refund.

After receipt of the refund, spending on health averages $30 per account, of which less than 1% ($0.30) goes towards goods that can be stockpiled such as medications. Our findings are in line with this result.

In addition to contributing to the literature of cash transfers on prescription use, we also add to the literature on the effects of the Alaska PFD, of which there is limited evidence of on health-related outcomes. Evans and Moore (2011) find that the Alaska PFD temporarily increases then decreases mortality. Using police report data, Watson et al. (2019) find increases in requested medical assists and substance abuse crime incidents. As the longest running universal cash transfer in the United States, the PFD is a useful natural experiment for learning about the potential effects of future cash transfer programs.

The paper proceeds as follows. Section 2 provides background on related literature and the PFD, Section 3 includes a description of the data, Section 4 contains the regression specifications, Section 5 presents results, Section 6 presents a synthetic control analysis, and Section 7 concludes.