Essays on the Macroeconomic and Measurement Consequences of Government Systems

Abstract

In Chapter 2, I study the macroeconomic consequences to delaying a fiscal stabilizer. Specifically, I study how delays to unemployment insurance benefits during the pandemic recession (fiscal stabilizer) affected consumption (macroeconomic consequence). The United States experienced an unprecedented increase in unemployment insurance (UI) claims starting in March 2020. State UI-benefit systems were inadequately prepared to process these claims. In states that used an antiquated programming language, COBOL, to process claims, potential claimants experienced a larger increase in administrative difficulties, which led to longer delays in benefit disbursement. Using daily debit and credit card consumption data from Affinity Solutions, I employ a two-way fixed-effects estimator to measure the causal impact of having an antiquated UI benefit system on aggregate consumption. Such systems led to a 2.8-percentage-point decline in total credit and debit card consumption relative to card consumption in states with more modern systems. I estimate that the share of claims whose processing was delayed by over 70 days rose by at least 2.1 percentage points more in COBOL states relative to non-COBOL states. Based on a back-of-the-envelope calculation using 2019 data, my results suggest that the decline in consumption in COBOL states in 2020 after the pandemic-emergency declaration corresponds to a real-GDP decline of at least $105 billion (in 2019 dollars).

In Chapter 3, Joonkyu Choi, Samuel Messer, Veronika Penciakova, and I study how business formation patterns in 2020 were affected by antiquated UI benefit systems. New business formation surged after the pandemic recession, but the causes of this surge are not well understood. The expansion of UI benefits under the CARES Act, coupled with the reduction of work search, provided unemployed potential entrepreneurs with the funds and time needed to develop business ideas. States that used an antiquated programming language, COBOL, to process claims experienced a lower growth rate in UI payments per unemployed than states with more modernized systems. Using business application data from the Business Formation Statistics, we employ a two-way fixed-effects estimator to measure the causal impact of having an antiquated UI benefit system on business formation. Such systems led to a 6.6 percent decline in business applications per capita in COBOL states relative to more modernized states from March 2020 to July 2020. We also find some evidence of business quality deterioration while the Federal Pandemic Unemployment Compensation program was in effect. Our findings highlight the potential role of UI policy in contributing to economic recoveries by fostering entrepreneurship.

In Chapter 4, the RESET team Gabriel Ehrlich, John Haltiwanger, David Johnson, Ron Jarmin, Seula Kim, Jake Kramer, Edward Olivares, R. Rodriguez, Mathew D. Shapiro, and I use point of sales (POS) data to construct real sales and compare these POS generated statistics to official statistics. Businesses, individuals, and government policymakers rely on accurate and timely measurement of nominal sales, inflation, and real output, but current official statistics face challenges on a number of dimensions. First, these key indicators are derived from surveys conducted by multiple agencies with different time frames, yielding a complex integration process. Second, some of the source data needed for the statistics (e.g., expenditure weights) are only available with a considerable lag. Third, response rates are declining, especially for high-frequency surveys. Focusing on retail trade statistics, we document important discrepancies between official statistics and measures computed directly from item-level transactions data. The long lags in key components of the source data delay recognition of economic turning points and lead to out-of-date information on the composition of output. We provide external data sources to validate the transactions data when their nominal sales trends differ importantly from official statistics. We then conduct counterfactual exercises that replicate the methodology that official statistical agencies use with the transactions data in the construction of nominal sales indices. These counterfactual exercises produce similar results to the official statistics even when the official nominal sales and item-level transactions data exhibit different trends.