Responsible researcher: Bruno Benevit
Original title: Health Insurance and Mortality: Experimental Evidence from Taxpayer Outreach
Authors: Jacob Goldin, Ithai Zvi Lurie and Janet McCubbin
Intervention Location: United States
Sample Size: 8.9 million taxpayers
Sector: Healthcare
Variable of Main Interest: Health insurance coverage, Mortality
Type of Intervention: Experiment, Dissemination
Methodology: OLS, IV, Logit
Summary
Despite several studies on the topic, the relationship between health insurance and mortality is still a widely discussed topic in the field of health economics. Sizing the impact of health insurance coverage is fundamental to improving the effectiveness of health policies. This study aimed to evaluate the impact of randomly sending informational letters from the IRS to 3.9 million households that paid a tax penalty for not having Affordable Care Act on health insurance coverage and mortality. . The results indicate that the intervention increased coverage during the following two years and reduced mortality among middle-aged adults during the same period.
The relationship between health insurance and mortality is a constantly debated topic in health economics literature. According to data from Lurie and Pearce (2019), approximately 9% to 13% of Americans under the age of 65 do not have year-round health insurance coverage. The share of Americans who do not have coverage for at least one month during the year ranges between 21% and 26%. With the aim of expanding health insurance coverage, the federal Affordable Care Act (ACA) was enacted in 2010, making health insurance coverage mandatory for most taxpayers. In this sense, measuring how health insurance coverage impacts mortality is essential for policymakers.
People's health insurance coverage in the United States is mostly provided by employers. Government programs such as Medicaid , Medicare , or the Veterans Administration represent the second largest source of coverage. U.S. citizens without these types of coverage have the option of enrolling in an Exchange purchased through their state's health insurance marketplace or a non- Exchange .
Unlike programs like Medicaid , individuals can only enroll in health insurance coverage via the Exchange during the open enrollment period for the corresponding year. For example, the enrollment period for the year 2017 covered the period from November 1, 2016 to January 31, 2017. Additionally, individuals are required to apply for Exchange coverage by the 15th day of the month preceding the month in which coverage should begin. Most private plans offered by employers also have a similar enrollment period, typically ending near the end of the calendar year.
In early 2017, the Internal Revenue Service (IRS) sent informational letters to taxpayers who had previously paid an income tax charge for lack of health insurance coverage under the so-called individual mandate provision of the ACA. Thus, this intervention allows the assessment of the causal relationship between coverage and taxpayer mortality.
From 2014 to 2018, the federal ACA law made health insurance coverage mandatory for most Americans, resulting in the payment of a fee for noncompliance. Individuals who did not have qualified health insurance coverage for themselves or a dependent for one or more months during the year were required to report and pay the penalty on their annual income tax return unless an exemption applied.
The taxpayer letter intervention was funded by the Department of Health and Human Services (HHS) and implemented by the IRS. The letter informed recipients that they had paid a fee in 2015, provided information about the penalty and plan costs for 2017, and provided instructions regarding whether to purchase coverage through the Exchange and Medicaid .
Recipients who met the criteria for receiving the letter were randomly selected, corresponding to a group that underwent the intervention (86%) and a control group that did not receive the letter (14%). Randomization was stratified by age and sex of primary filer, marital status, number of dependents, income, presence of self-employment income, 2014 penalty/fee status, and whether the taxpayer's state expanded the Medicaid and/or participated in the Federal Exchange during 2017.
The sample used was constructed from IRS income tax returns submitted for the year 2015 showing the existence of positive penalties relating to the ACA. The authors adopted several exclusion criteria for the sample, disregarding recipients with an address outside the United States and/or enrolled in Exchange in 2015 or 2016. Data on health insurance coverage were extracted from forms submitted annually to the IRS by private insurers and public, self-insured employers, and health insurance marketplaces (Form 1095 A/B/C), providing monthly coverage information at the individual level. The forms provide information about monthly coverage and the type of coverage the individual was enrolled in (for example, whether the coverage was Medicaid ). Mortality data is obtained from the Social Security Death File , which records the date of all deaths in the United States. The final sample consists of 4.5 million declarations, corresponding to 8.9 million individuals (main taxpayer, spouse, and up to 4 dependents).
The analyzes in the study considered characteristics of the individuals, the household, the region/state, the IRS penalty situation and coverage of 2014 and 2015. The period covers the years 2015 to 2018, considering the period 2017 to 2018 as the period of treatment. No significant differences were identified in the means of the variables used between the treated and control groups.
The first analysis verifies how the intervention affected taxpayers' coverage of the months under health insurance coverage and the probability of being covered in at least one month during the period from 2017 to 2018. Regressions were estimated by dividing the analysis considering the full sample, only for those who were insured during the entire first 11 months of 2016, and only for those who were not insured during the entire first 11 months of 2016. Additionally, regressions were conducted under the same division criteria and considering only adults between 45 and 64 years old at the end of 2017.
The second analysis seeks to identify the effects of the intervention over the months in terms of coverage between the treated and control groups and the difference between the groups. The third analysis presents the effect of the intervention for each type of health insurance coverage, dividing the regressions between the full sample and adults between 45 and 64 years old.
The following analyzes seek to identify how exogenous variation in health insurance coverage caused by the intervention explains the relationship between coverage and mortality. To this end, using the OLS and IV methods, the effects of the intervention on mortality were estimated, checking whether possible effects were due to increases in health insurance coverage and the effects of each month of coverage induced. Additionally, several robustness tests were conducted with different sample specifications in relation to age, placebo tests and analyzes regarding the heterogeneity of the coverage effect caused by the intervention over the months. These analyzes considered the sample of adults between 45 and 64 years old, the age group with the highest mortality rate.
The results of the first analyzes demonstrated that people without insurance in any month throughout 2016 who underwent the intervention were 1.1 percentage points more likely to take out some health insurance coverage in the two years following the intervention compared to the control group, representing a relative increase of 1.9%. On average, the intervention resulted in an average of 0.23 additional months of coverage during this period, or one additional year of coverage for every 52 individuals treated.
Regarding the effects by type of coverage contracted, the effect occurred mainly through adherence to Exchange and, to a lesser extent, through the Medicaid . The coverage rate was higher among the treatment group, with a reduction in the difference between the treated and control groups over the two years following the intervention.
The results of the effect of additional coverage induced by the intervention showed a reduction in the mortality rate for the group that requested some coverage after the intervention. The death rate among previously uninsured people ages 45 to 64 was lower in the treatment group than in the control group by approximately 0.06 percentage points. This result corresponded to one fewer death for every 1,587 individuals treated in the two years following the intervention. No effects were identified for groups beyond the age range of 45 to 64 years.
The instrumental analysis indicates that the average effect per month of induced coverage on mortality was approximately -0.18 percentage points over the two years after the intervention. However, the authors point out that the most reliable magnitude of the effect on mortality is probably approximately -0.04 percentage points. The results also show that the annual effect on mortality is less than 12 times the estimated effect per month due to the concavity in the relationship between coverage and mortality. In other words, the monthly effects reduce over the 2-year period of the intervention.
This article provided important evidence related to the use of disclosure policy strategies to increase health insurance coverage. The results demonstrated that the groups that showed the greatest reductions in mortality were those for whom the benefits of health coverage tend to be small, especially when the disclosure refers to financial penalties for remaining uninsured. According to the authors, these results suggest that lack of attention, shaping how tax incentives interact with adverse selection in health insurance markets, and causing behavioral frictions that reduce coverage uptake may be especially concentrated among those who would benefit from the sign up. Given this scenario, disclosure policies of this type are effective in inducing health insurance coverage.
References
LURIE, I.Z.; PEARCE, J. Health Insurance Coverage from Administrative Tax Data. Office of Tax Analysis Working Paper 117 , 2019.