The ongoing COVID-19 pandemic has severely impacted the US economy, with rising unemployment (Coibon et al. 2020a), income losses for many households (Cajner et al. 2020), reduced consumer spending (Coibon et al. 2020b), and overall increases in economic uncertainty (Baker et al. 2020a). Baldwin and Weder di Mauro (2020) provide a comprehensive overview of the economic consequences of the pandemic. This large economic shock also had profound implications for the consumer credit market.

Using credit card data from the Federal Reserve’s monthly Y-14M reports, in a new paper (Horvath et al.  2020), we examine the early impact of the COVID-19 shock on both the use and availability of credit in the US consumer credit card market through March 2020. We estimate the local effect of both pandemic severity and policy responses in the form of non-pharmaceutical interventions (NPIs) on use and availability of credit. Moreover, we examine differences in credit market outcomes for borrowers of different creditworthiness.

While most counties in the US still had not registered any confirmed COVID-19 cases by mid-March, some regions were already affected severely at this point. Figure 1 visualizes data on COVID-19 cases from the Johns Hopkins COVID-19 Data Repository (Dong et al. 2020). As can be seen, there were clusters of affected counties in Washington State, the Bay Area, New York State, and Southern Florida by mid-March.

Figure 1 Confirmed COVID-19 cases per 100,000 as of 15 March 2020

Notes: This figure illustrates the number of confirmed COVID-19 cases per 100,000 population across US counties as of 15 March 2020.
Source: Johns Hopkins COVID-19 Data Repository (Dong et al. 2020) and authors’ own calculations.

At the same time, many counties had already enacted some NPIs, such as large gathering bans and the closure of public venues, schools, and universities by mid-March. We use data from the Coronavirus Intervention Dataset provided by Keystone Strategy (Keystone 2020) to construct a simple county-level measure of NPI stringency, by adding up the number of NPIs. As shown in Figure 2, by mid-March most counties with NPIs had inherited their NPI measures from state legislation and were therefore subject to a high degree of NPI stringency relative to their number of confirmed cases. This allows us to disentangle the effect of the pandemic itself from the effect of NPIs. It is important to note that by mid-March, the most restrictive NPIs, such as shelter-in-place orders and lockdowns, had not yet been implemented in any county. Therefore, our analysis does not inform the discussion on these more restrictive public health interventions.

Figure 2 Non-Pharmaceutical Intervention stringency as of 15 March 2020

Notes: This figure illustrates the simple NPI stringency indicator across US counties, calculated as the number of NPIs as of 15 March 2020.
Source: Keystone Strategy Coronavirus Intervention Dataset (Keystone 2020) and authors’ own calculations.


We find a reduction in consumer credit use in response to local pandemic severity. Borrowers in severely affected counties (defined as 15 confirmed cases per 100,000 population as of 15 March) exhibited, on average, a 5.4% reduction in balances and a 6.1% reduction in transactions compared to borrowers in unaffected counties. Figures 3 shows how reductions in credit card transaction volumes are geographically distributed across US counties. The map illustrates that transaction volumes decreased the most in Washington State, California, the region around New York City, and Southern Florida. This pattern is consistent with the geographical distribution of confirmed COVID-19 case severity shown in Figure 1.

READ ALSO  “Buy British”: The viability of a nationalist commercial policy

Figure 3 The geographical distribution of changes in average credit card transaction volumes

Notes: This figure illustrates changes in average credit card transaction volumes across counties in the US. The plotted variable is the average year-to-year change (2019 to 2020) in month-to-month changes (February to March) in credit card transactions.
Source: Federal Reserve’s Y-14M reports and authors’ own calculations.

However, as Figures 4 and 5 show, there was substantial heterogeneity across borrowers in response to the pandemic. In particular, reductions in both balances and transactions were driven by creditworthy borrowers, whereas outstanding monthly balances even increased for the riskiest borrowers in affected counties. Notably, in our related paper (Horvath et al. 2020), we report qualitatively similar estimates for the effect of NPIs on the use of credit. These results are, however, smaller in magnitude. Hence, we conclude that in the early stages of the COVID-19 crisis the pandemic itself was the main driver of changes in consumer credit use.

Figure 4 The effect of case severity on credit card balances across FICO score classes

Notes: This figure illustrates the effect of confirmed COVID-19 cases on credit card balances across different borrower classes.
Source: Federal Reserve’s Y-14M reports and authors’ own calculations.

Figure 5 The effect of case severity on credit card transactions across FICO score classes

Notes: This figure illustrates the effect of confirmed COVID-19 cases on credit card transactions across different borrower classes.
Source: Federal Reserve’s Y-14M reports and authors’ own calculations.

These findings are surprising in the light of existing studies on consumer responses to negative economic shocks. Economic consumption theory (Carroll 2001) predicts that households with low levels of liquid wealth exhibit a stronger consumption response to negative income shocks. This prediction is confirmed by recent empirical evidence (e.g. Bunn et al. 2018). In our setting, as borrowers in lower FICO score classes tend to have lower levels of liquid wealth (Baker 2018) and as the COVID-19 income shocks disproportionately affected low-wage occupations (Cajner et al. 2020), we would expect to see a stronger negative consumption response for low FICO score classes. Since we find the opposite effect, we provide two alternative explanations for our results.

First, our results might indicate a disruption to consumption patterns. The COVID-19 shock had heterogeneous effects on consumer spending across different categories of goods and services (Baker et al. 2020b). Discretionary spending categories (e.g. recreation, travel, and entertainment expenses) declined the most, while non-discretionary expenses (e.g. utilities, food, and childcare) declined only modestly (Coibion et al. 2020b).1 As non-discretionary expenses make up a larger share of the consumption of less creditworthy borrowers, it is relatively harder for them to reduce spending and thus their consumer credit use. Our findings corroborate evidence that a higher pre-shock share of spending on ‘non-essential’ goods and services was associated with a larger reduction in total spending in the wake of the COVID-19 pandemic (Andersen et al. 2020). In our related paper (Horvath et al. 2020), we provide further evidence for such an interpretation of our results.

READ ALSO  Bloomberg raises $16m to help ex-felons in Florida vote

Second, our results are also consistent with recent findings on the heterogeneity of consumer credit responses to economic uncertainty shocks (Bloom 2009). Di Maggio et al. (2017) find that local economic uncertainty is associated with an increase in the credit card balances of less creditworthy borrowers and a decrease in the consumer credit demand of more creditworthy borrowers. The underlying mechanism is heterogeneity in the pecuniary costs of default between more and less creditworthy borrowers. Low FICO score borrowers, with limited access to credit, have a lower cost of default than high FICO score borrowers, which increases their incentives to engage in risk shifting. In this respect, our empirical findings capture the effect of the economic uncertainty shock caused by the COVID-19 epidemic (Baker et al. 2020a).

With regard to the availability of credit, we focus on newly issued credit cards. We find a 48% reduction in the originations of new credit cards in the second half of March, which is more pronounced in counties affected by the pandemic itself and in counties with more stringent NPIs. These very large magnitudes are in line with anecdotal evidence from the financial industry. As reported by the American Banker, “card originations were down 55% during the first two weeks of April compared with average February levels” (Wack 2020).   

Notably, the number of credit card originations decreased the most for borrowers both in the highest and in the lowest FICO score buckets. This reduction in credit card originations at both ends of the FICO score distribution could reflect both demand and supply effects respectively (see also Agarwal et al. 2018). While low FICO score borrowers were likely willing to borrow, which is in line with our result for the use of credit, banks might have been reluctant to lend to these borrowers. Concordant with this, we find a reduction in credit limits and an increase in APR spreads for new credit cards issued to the least creditworthy borrowers (as measured by FICO scores) in affected counties. Thus, at the lower end of the FICO score distribution, the reduction in credit card originations was likely primarily driven by supply effects. Conversely, banks were likely still willing to lend to high FICO score borrowers, as we find no reduction in credit limits and a decrease in APR spreads for new credit cards issued to the most creditworthy borrowers in affected counties. However, high FICO score borrowers likely had a low propensity to borrow, which is consistent with our findings on the use of credit. Hence, for creditworthy borrowers, the reduction in credit card originations was likely primarily driven by demand effects. This heterogeneity in the availability of credit across borrower types is consistent with a flight-to-safety response of banks to the COVID-19 shock.


The COVID-19 shock already had strong effects on the use and availability of consumer credit in the US by March 2020, and these effects varied considerably across borrower types. While creditworthy borrowers reduced their credit card balances and transaction volumes, there was an increase in outstanding balances for the riskiest borrowers. Moreover, we report a sizeable reduction in new credit card originations, which was likely driven by supply effects for risky borrowers and by demand effects for creditworthy borrowers. Consistent with a flight-to-safety response of banks to the COVID-19 shock, we find a reduction in the credit limits and an increase in the APR spreads of newly issued credit cards to the riskiest borrowers in affected counties.

READ ALSO  Commercial ties to deepen with new model


Agarwal, S, S Chomsisengphet, N Mahoney and J Stroebel (2018), “Do Banks Pass through Credit Expansions to Consumers Who want to Borrow?”, Quarterly Journal of Economics  133(1):129-190.

Andersen, A, E Hansen, N Johannesen and A Sheridan (2020), “Consumer responses to the COVID-19 crisis”,, 15 May 2020.

Baker, S (2018), “Debt and the Response to Household Income Shocks: Validation and Application of Linked Financial Account Data”, Journal of Political Economy 126(4):1504-1557.

Baker, S, N Bloom, S Davis and S Terry (2020a), “COVID-induced economic uncertainty and its consequences”,, 13 April 2020.

Baker, S, R Farrokhnia, S Meyer, M Pagel and C Yannelis (2020b), “Income, liquidity, and the consumption response to the COVID-19 pandemic and economic stimulus payments”,, 17 June 2020.

Baldwin, R and B Weder di Mauro (2020), “Economics in the time of COVID-19: A new eBook”,, 6 March 2020.

Bloom, N (2009), “The Impact of Uncertainty Shocks”, Econometrica 77(3):623–685.

Bunn, P, J Le Roux, K Reinold and P Surico (2018), “The consumption response to positive and negative income shocks”, Journal of Monetary Economics 96:1-15.

Cajner, T, L Crane, R Decker, J Grigsby, A Hamins-Puertolas, E Hurst, C Kurz and A Yildirmaz, (2020), “The U.S. Labor Market During the Beginning of the Pandemic Recession”, Becker Friedman Institute Working Paper 2020-58.

Carroll, C D (2001), “A Theory of the Consumption Function, With and Without Liquidity Constraints”, Journal of Economic Perspectives 15(3):23-45.

Coibion, O, Y Gorodnichenko and M Weber (2020a), “Labour markets during the Covid-19 crisis: A preliminary view”,, 14 April 2020.

Coibion, O, Y Gorodnichenko and M Weber (2020b), “The cost of the COVID-19 crisis: Lockdowns, macroeconomic expectations, and consumer spending”,, 12 May 2020.

Di Maggio, M, A Kermani, R Ramcharan and E Yu (2017), “Household Credit and Local Economic Uncertainty”, Federal Reserve Bank of Philadelphia Working Paper 17-21.

Dong, E, H Du and L Gardner (2020), “An interactive web-based dashboard to track COVID-19 in real time”, The Lancet: Infectious Diseases 20:533-534.

Horvath, A, B Kay and C Wix (2020), “The COVID-19 Shock and Consumer Credit: Evidence from Credit Card Data”, Covid Economics 36:129-177.

Keystone (2020), Accessed May 7, 2020.

Wack, K (2020), “Credit card lenders clamp down to mitigate coronavirus risk”, American Banker.


1 We note that it is not always obvious which goods and services should be classified as “discretionary” and “non-discretionary”. However, we adopt this terminology as well as the corresponding classification of expenditures from the existing literature (see e.g. Coibion et al., 2020b). Some of the spending categories that are called discretionary were also substantially harder to consume during the pandemic.