« Back to Results
Using Matched Data to Address Policy Questions
Friday, Jan. 7, 2022
12:15 PM - 2:15 PM (EST)
Society of Government Economists
Using Linked Survey and Administrative Data to Simulate Effects of Proposed Policies: The Case of the Child Tax Credit Expansion
We use linked survey and administrative data to simulate the effect of a proposed expansion of the Child Tax Credit (CTC) that would increase benefit amounts and make it fully refundable. While policymakers frequently rely on simulations of the effects of proposed policies to inform their decisions, existing efforts typically rely on survey data, which suffer from income misreporting and insufficient detail to model proposed policies precisely. These limitations bias our understanding of the targeting of proposed policies and their effects on economic well-being. We show how the Comprehensive Income Dataset (CID)—which links household surveys with an unprecedented set of administrative tax and program data—can be used to overcome these limitations in the case of the proposed CTC expansion. Specifically, the CID provides a more accurate income baseline including the receipt of specific government programs and current CTC benefits, as well as the ability to more precisely model the CTC expansion itself through the ability to observe tax filing information and children’s birth dates. We use the CID to simulate the effect of the expanded CTC on the distribution of income and poverty (along with deep, near and twice poverty) rates for specific demographic groups. We also assess targeting of the expanded CTC benefits, evaluating the extent to which benefits target individuals with lower incomes, lower levels of material well-being, and who lack receipt of other government benefits. We compare the CID-based estimates with purely survey-based estimates, allowing us to quantify the importance of linked survey and administrative data for simulating the effects of the CTC expansion. Finally, we discuss potential behavioral effects of the CTC expansion and how they might affect our static CID-based estimates.
Exploring the Gender Gap in Retirement Saving: Using IRS Data to Understand Varying Contribution Behavior between Men and Women
Account-based retirement savings have risen in importance, and thus, individuals increasingly are managing their retirement accumulations. With $9.1 trillion in assets at year-end 2018, individual retirement accounts (IRAs) are a significant component of the U.S. retirement system, serving as not just a contributory savings vehicle but also as a consolidator of rollovers from employer-sponsored pension plans. While retirement plan designs are gender neutral, women’s results regarding retirement accumulations appear to come up short when compared with men’s results The question remains, does having access to a defined contribution (DC) plans and individual retirement accounts (IRAs) help level the playing field. This paper analyzes the varying contribution behavior between men and women, accounting for both IRAs and DC plans contributions. We also examine differences in IRA balances and the role that rollovers and withdrawal behavior plays in the savings difference between men and women. In this study, we use IRS information filings (Forms 5498, 1099-R, and W-2) and tax return data (Form 1040) between 2008 and 2018 to analyze whether there are systemic gender differences between IRAs and DC plan behaviors controlling for age, income, and marital status. This research marks the first paper to combine IRS administrative IRA and DC plan data to directly study the existence and size of the retirement savings gender gap.
Addressing Nonresponse Bias in Household Surveys Using Linked Administrative Data
Nonresponse has been increasing in household surveys. To address nonresponse bias, we use address-linked administrative data to identify individuals in respondent and nonrespondent households. We link them to income, demographic, and socioeconomic information from administrative data, prior surveys, and the decennial census. We use entropy balancing to adjust survey weights to match a high-dimensional vector of moment constraints. In the 2020 CPS ASEC, nonresponse biased income estimates up by 2 to 3 percent. In other years, we do not find evidence of bias in income or poverty. We produce public-use weights that address this nonresponse bias while protecting respondent confidentiality.
Using WIC Administrative Data to Evaluate the Supplemental Poverty Measure
The Supplementary Nutrition Program for Women, Infants, and Children program (WIC) is designed to provide food assistance and nutritional screening to low-income pregnant, postpartum women and their infants, and to low income children up to the age of 5. The Supplemental Poverty Measure (SPM) is an alternative poverty measure produced by the Census Bureau since 2011 using the Current Population Survey Annual Social and Economic Supplement (CPS ASEC). Unlike the official poverty measure, the SPM incorporates transfer income from programs such as WIC, SNAP, and TANF. Past research suggests that survey responses to transfer program questions produce undercounts of both participation and benefit amounts (Meyer and Mittag 2015, Shantz and Fox 2018, Mittag 2019). This underreporting can deteriorate data quality and distort SPM poverty rates. To address the impact of WIC misreporting on the SPM, this paper links state administrative data on the WIC program to the CPS ASEC covering calendar years 2009-2017. We use this linkage to directly compare self-reported WIC participation to the administrative data (e.g., false positive and false negative rates). We also assess the degree to which annual household WIC benefit amounts line up between the two data sources and the impact of any differences on overall SPM poverty and SPM poverty for various demographic groups.
U.S. Census Bureau
U.S. Census Bureau
Victoria L. Bryant
Internal Revenue Service
H0 - General
J0 - General