COVID-19, bank deposits, and lending.

During the pandemic, households accumulated savings in their deposit accounts as a result of a reduction in their spending, which occurred due to the restrictions on their mobility. This led to a significant increase in bank deposits for banks located in counties with a larger reduction in spending. Banks, in turn, used these additional funds to issue more real estate loans. This implies that policies that might affect household spending would lead to changes in the volume of deposits in the banking system, which have consequences on banks' loan supply.

Introduction

Banks experienced a large increase in their deposits during the COVID-19 pandemic. At the very onset, from March 11 to April 1, 2020, aggregate bank deposits increased by almost $1 trillion for banks in the United States (Li et al., 2020). The increase continued by another $1 trillion until the end of 2020 (Levine et al., 2021). As shown in Fig. 1, this aggregate increase is also reflected in the average deposit size on banks’ balance sheets, which increased by about $1.4 billion from the end of 2019 ($4.86 billion) until the end of 2020 ($6.26 billion).

Fig. 1

Change in average deposits over time. This figure plots the average total deposits for all types of banks from Q1 2018 until Q4 2020.

In this paper, we aim to shed light on the reasons behind this tremendous deposit inflow to the banking system and, subsequently, study how banks utilized this additional funding in their loan provision during the onset of the pandemic. We show that, as a response to the COVID-19 pandemic, households accumulated savings due to the decrease in their spending, which has led to an increase in banks’ core deposits. This inflow of funds to the banking system consequently resulted in more real estate lending.

We begin our analysis by documenting that banks, which are located in counties with a higher exposure to the pandemic, experienced a higher increase in their deposits. To measure the severity of the pandemic in each county, we focus on the number of COVID-19 cases per capita. We subsequently calculate each bank’s exposure to the pandemic through its branches and define a bank-level exposure measure as the weighted average COVID-19 cases per capita, where weights are the share of deposits the bank holds in each county that it has a branch. This measure enables us to quantify the exposure of each bank to local county-level COVID-19 outbreaks through its branches in these counties.

Using this bank-level exposure measure, we then study whether banks located in counties with higher infections experienced a higher increase in their deposits. We expect that the increase would happen in banks’ core deposits rather than brokered deposits — households located in the U.S. increased their savings in their deposit accounts, which would be reflected on banks’ core deposits whereas brokered deposits are obtained through a deposit broker. Our results show that banks with a higher exposure to the pandemic experienced a significant increase in their core deposits while their brokered deposits did not change, which is consistent with our expectations. We find that a 1 percentage point increase in the COVID-19 exposure of banks (i.e., 1 percentage point higher weighted average COVID-19 cases per capita) led to a significant increase in their core deposits by about 0.23 percentage point of their total assets.

We focus on one potential reason that might have motivated households to accumulate deposits in their banks: the “forced savings” view, which argues that households saved as a result of not being able to spend their money on leisure activities or travel. According to this view, we would expect that households with a higher reduction in spending would save more money. To investigate this, we use county-level consumer spending data during the pandemic, which is provided by the “Opportunity Insights Data Dictionary”, and calculate the weighted average reduction in spending at the bank level. We find that core deposits increased for banks that are active in regions with a higher reduction in spending. These banks experienced a significant increase in their core deposits by 0.39 percentage point of their total assets for a 1 percentage point higher exposure to the pandemic and, as a result, their total deposits increased by 0.33 percentage point of their total assets. For banks with a lower reduction in spending, on the other hand, a higher exposure to the pandemic did not lead to an increase in deposits. Thus, this finding supports the “forced savings” view that not being able to spend increased household savings in banks’ deposit accounts.

We further expect that such “forced savings” can happen only for households that did not lose their jobs during the pandemic. To examine this conjecture, we calculate the increase in unemployment rates for each county. According to our results, only banks that are located in the counties with a lower increase in unemployment experienced an increase in their deposits. This increase happened only when households experienced a large reduction in spending at the same time. We find that banks with a 1 percentage point higher exposure to COVID-19 experienced a significant increase in their total deposits by about 0.42 percentage point of their total assets if they are located in the counties with a higher spending reduction and a lower increase in unemployment. The increase in total deposits is again driven by an increase in core deposits.

As a next step, we study whether banks that experienced a large inflow of deposits use this additional funding to provide more loans during the pandemic. We find that banks with a higher exposure to the pandemic relatively increased their real estate loans if they were located in counties with a higher spending reduction which is consistent with our above described findings that total deposits increased for banks located in these counties. According to our results, banks with a 1 percentage point higher exposure to COVID-19 relatively increased their real estate loans significantly by about 0.06 percentage point of their total assets. The increase in real estate loans might be explained by the unexpected increase in housing demand and prices during the COVID-19 pandemic (Zhao, 2020, Bhat et al., 2021).

Last, one expects that banks with lower capital ratios might be more hesitant to increase their loans when they experience an inflow of deposits. This could be explained with their tendency to lower risk taking in an attempt to avoid any reductions on their capital (Cornett et al., 2011) or with their lower performance during times of distress (Berger and Bouwman, 2013). To investigate this, we study the impact of banks’ pre-pandemic capital ratios on their likelihood of issuing additional loans with an increase in their deposits. We find that high-capital banks, i.e., the ones with capital ratios (2019Q4) above the median, increased their lending when they experienced a large inflow of deposits, while low-capital banks’ lending did not change. This finding highlights the importance of bank capital for loan provision during crises (Kashyap and Stein, 2000, Kishan and Opiela, 2000, Meh and Moran, 2010, Puri et al., 2011, Dursun-de Neef, 2019, Dursun-de Neef et al., 2022).

To summarize, in this paper we show that, during the pandemic, households accumulated their savings in their deposit accounts as a result of a reduction in spending — they were not able to spend money on leisure activities due to the mobility restrictions. This gave rise to a large increase in bank deposits and, in turn, banks used these additional funds to issue more real estate loans. Our results imply that policies that might have an impact on household spending would lead to changes in the volume of deposits in the banking system, which have consequences on loan supply and economic activities as documented by Becker (2007) (see also, Chang et al., 2010, Acharya et al., 2011, Lin, 2020, Blickle, 2022).

This paper is organized as follows: Section 2 discusses the related literature. We present our main hypotheses in Section 3. Sections 4, 5 describe our empirical methodology, data, and main results. Last, Section 6 concludes.

Literature review

The literature that studies changes in bank deposits during the pandemic remains scarce.1 We are aware of only two articles: First, Li et al. (2020) briefly discuss the changes in deposits that occurred in the first quarter of 2020 and document that the effect of having more pre-crisis unused loan commitments on deposits depends on the size of the banks. Large banks with more unused loan commitments experienced a decrease in their deposits whereas their small peers experienced an increase in their deposits. Second, Levine et al. (2021) study different reasons for why deposits increased. The authors use proprietary deposit rates data from ‘RateWatch’ and show that deposit rates at bank branches in counties with higher COVID-19 infection rates fell by more than rates at other branches, suggesting that the deposit inflows were triggered by supply rather than demand for deposits. The authors further argue that the precautionary savings motive may have been a dominant behavior behind the surge in the supply of deposits and a corresponding drop in deposit rates: A positive relationship between local COVID-19 cases and the intensity of online searches (e.g., on topics such as unemployment or saving) is found, suggesting that COVID-19 cases are positively correlated with individuals’ expectations about job losses and spending cuts. In our paper, we instead focus on the “forced savings” view, which postulates that the accumulation of deposits was not intentional: Depositors experienced an increase in their deposit accounts as they had fewer occasions to spend their money on leisure activities. Consistent with this view, we show that the increase in deposits happens mainly for banks that were active in counties with a higher reduction in spending. We argue that the reduction in household spending led to windfall increases in deposit accounts.

On the other hand, the literature that examines the consequences of the COVID-19 crisis on banks’ loan supply is growing quickly.2  Li et al. (2020) were the first to document that upon the onset of the pandemic, U.S. banks increased their loans significantly more if they had higher unused loan commitments. The authors show that although banks increased their lending, their total credit supply (i.e., total loans plus commitments) did not change. Similarly, Greenwald et al. (2021) highlight that U.S. banks that experienced larger credit line drawdowns restricted term lending to small firms more during the COVID-19 crisis. Beck and Keil (2022) show that U.S. banks faced an increase in their loan loss provisions and non-performing loans. Hasan et al. (2021) find that syndicated loan spreads increased as either the lender or the borrower became more exposed to the pandemic. In addition, Colak and Öztekin (2021) study the effect of the pandemic on global lending by analyzing banks from 125 countries and Duan et al. (2021) examine changes in banks’ systemic risk for over 1500 listed banks from 64 countries. Focusing on European banks, Dursun-de Neef and Schandlbauer (2021) show that worse-capitalized banks increased their loan supply significantly more during the pandemic while holding their delinquent loans lower than their better-capitalized peers, which is attributed to zombie lending by worse-capitalized banks in an attempt to avoid write-offs on their capital. We contribute to this literature by showing that households accumulated their savings in their deposit accounts as a result of a reduction in their spending — they were not able to spend money on leisure activities due to reduced mobility. According to our results, this led to a large increase in bank deposits and, in turn, banks used these additional funds to issue more real estate loans.

Several other papers study the impact of the COVID-19 crisis on the economy. Using a survey of 5800 small businesses at the end of March 2020, Bartik et al. (2020) document the fast negative business consequences of COVID-19 such as mass layoffs and closures. Acharya and Steffen (2020) document that firms drew down their credit lines and raised their cash holdings. Moreover, further analyses examine the stock market reaction (e.g., Baker et al., 2020a, Gerding et al., 2020, Pagano et al., 2020) or household spending (e.g., Coibion et al., 2020, Baker et al., 2020b, Chen et al., 2021). Baker et al. (2020b) show that initial spending increased sharply, particularly in retail, credit card spending and food items to stockpile these goods; however, this was quickly followed by a sharp decrease in overall spending. Households reacted most strongly in states with reduced mobility due to shelter-in-place and stay-at-home orders.3  Horvath et al. (2021) analyze monthly U.S. credit card data and find that in counties that were more severely affected by the pandemic, more creditworthy borrowers reduced their credit card balances and transactions while banks engaged in “flight-to-safety” and reduced credit limits to risky borrowers.

Hypotheses development

Banks experienced a tremendous surge in their deposits at the onset of the pandemic: Aggregate deposits increased by almost $1 trillion at the very onset, i.e., from March 11 to April 1, 2020, and this increase continued by another $1 trillion until the end of 2020 (Levine et al., 2021, Li et al., 2020). It is thus interesting to analyze the reason behind this large inflow of deposits and the implications of this on banks’ loan supply.

We first investigate whether the increase in deposits was likely an outcome of “forced savings” that households saved as a result of not being able to spend their money. In general, lockdowns, shelter-in-place policies, and stay-at-home orders led to a large reduction in mobility (Moreland, 2020) and, as a consequence, households were likely to save more during the pandemic as they have had fewer occasions to spend their money (Baker et al., 2020b). As a result, we expect that banks, which are located in counties where the reduction in spending was higher, would experience a higher increase in their deposits compared to banks located in the counties with a lower reduction in spending. Moreover, we expect that people would be able to save only under the condition that they did not lose their jobs, and hence, deposits would increase primarily in counties where the increase in unemployment was lower. This leads us to the following hypothesis:

H1: Deposits increased with the exposure to the pandemic mainly for banks operating in the counties with a higher reduction in spending. This was more pronounced if the counties did not experience a large increase in unemployment.

We subsequently analyze whether banks that experienced a large increase in their deposits financed relatively more loans during the pandemic. Given that there was a significant increase in housing demand and house prices during COVID-19 as documented by Zhao (2020) and Bhat et al. (2021), one might expect a relative increase in real estate loans. Moreover, there is a large existing literature that highlights the importance of bank capital for loan provision during crises (Kashyap and Stein, 2000, Kishan and Opiela, 2000, Meh and Moran, 2010, Puri et al., 2011, Dursun-de Neef, 2019, Dursun-de Neef et al., 2022). Hence, banks with lower capital ratios might be more hesitant to increase their loans when they experience an inflow of deposits to avoid an increase in their risk taking (Cornett et al., 2011) or due to their lower performance during times of distress (Berger and Bouwman, 2013).

H2: Banks with a higher exposure to the pandemic issued more real estate loans if the counties that they are located in had a larger reduction in spending. This was more pronounced if banks had sufficient capital at hand when entering the COVID-19 pandemic.

Data

Measuring banks’ exposure to COVID-19

Our bank-specific measure of COVID-19 exposure is calculated as the weighted average COVID-19 cases per capita in all counties in which a bank has branches, where the share of bank deposits in each county is used as weights. The weights reflect the geographic distribution of the bank’s deposits.4 We call this measure “Weighted_Covid_19” and it is defined as: for bank . is the proportion of bank ’s deposits in county . We rely on the amount of branch deposits in June 2019, since this is the last available Summary of Deposits data before the COVID-19 outbreak. is the number of COVID-19 cases per capita in county in each quarter of 2020.

This measure is at the bank level and it quantifies the exposure of each bank to the COVID-19 outbreak through its branches in each county all around the United States.

Data sources, sample selection, and descriptive statistics

Our analysis focuses on U.S. commercial banks between 2018Q1 and 2020Q4. We collect all bank data from SNL Financial, part of S&P Global. Besides balance sheet and income statement data, the database provides detailed information on bank branches, which comes from the Summary of Deposits filings of the Federal Deposit Insurance Cooperation (FDIC). To eliminate outliers, we winsorize all variables and we exclude non-depository institutions (defined as banks whose total deposits to total assets ratio is less than 2%), bank-years with negative values of total equity, total loans and Tier-1 capital, as well as extreme growth rates of deposits and loans where the yearly growth rate exceeds 100% (see Table 1 for the definitions of variables). The final unbalanced sample consists of 4,860 banks (55,062 bank-quarter observations).

Table 1

Definitions for the key variables.

Variable		Definition
Main dependent variables
ΔDeposits/laggedtotalassets		Change in (total, core, brokered) deposits divided by lagged total assets
ΔLoans/laggedtotalassets		Change in (real estate, consumer, C&I) loans divided by lagged total assets
Main independent variables
Weighted_Covid_19		Weighted average COVID-19 cases per capita in all counties in which a bank has branches, where the share of bank deposits in each county is used as weights: ∑j=1NDep_bank_i,j×Covid_19_j,t
Crisis		Indicator variable that is equal to one for 2020 and zero otherwise
Low (high) spending reduction		Indicator variable that is equal to one if the bank-level weighted average quarterly reduction in spending (as measured by ‘Opportunity Insights’ data) in 2020 is below (above) the median and zero otherwise
Low (high) unemployment increase		Indicator variable that is equal to one if the bank-level weighted average change in the counties’ average unemployment rate between 2020Q4 and 2019Q4 is below (above) the median and zero otherwise
Low (high) Δ deposits		Indicator variable that is one if the change in the deposit to asset ratio between 2020Q4 and 2019Q4 is below (above) the median and zero otherwise
Low (high) capital		Indicator variable that is one if the 2019Q4 capital to asset ratio is above (below) the median and zero otherwise
Low (high) HPI		Indicator variable that is one if the 2019Q4 bank-level weighted average house price indices (HPIs) are below (above) the median and zero otherwise
Bank characteristics
Total equity		Total equity divided by total assets to proxy Capital adequacy (C)
Loan loss reserves		Loan loss reserves divided by total loans to proxy Asset quality (A)
Net interest income		Net interest income divided by total assets to proxy Management quality (M)
Return on assets		Net income divided by total assets to proxy Earnings (E)
Cash		Cash divided by total assets to proxy Liquidity (L)
Deposits		Total deposits divided by total assets to proxy Sensitivity to market risk (S)
Size		Natural logarithm of total assets
Unused commitments		Unused commitments divided by the sum of total assets and unused commitments
Weighted MSA characteristics
County population		Total population in a county (in log)
Income per capita		Income per capita at the county level (in log)
Median age		The median residents’ age at the county level

Panel (A) in Table 2 shows the summary statistics for the entire time period. The average bank in our sample has $3.9 billion in total assets. Banks are generally well-capitalized. The average total equity to total assets ratio is 11.6% and the average Tier-1 regulatory capital ratio is 16.9%. The average bank has a return on assets of 0.27% and net interest income of 1.0%. The largest asset class is loans with 64.7% followed with securities (19.4%). Banks hold on average 9.8% of cash. They finance their assets mainly by deposits with 84.1%. Only 4.3% of total assets are financed by non-depository debt.

Table 2

Summary statistics.

	Mean	St. dev.	p1	p25	Median	p75	p99
(A) All quarters
Total Assets (000)	3940512	59924715	19651	109187	230082	548883	41760725
Total Loans (000)	2101753	27347144	8077	65071	150516	379251	25524798
Total C&I Loans (000)	482116	6857193	146	6402	17393	53522	6180927
Total Consumer Loans (000)	348271	5501313	67	1633	3950	10013	2185674
Total RE Loans (000)	981841	11047295	1808	41611	106070	278620	13181234
Total Deposits (000)	3086802	45884295	16582	91977	194391	462825	32939047
Core Deposits (000)	2491475	36967702	15629	84198	176983	415263	28453056
Brokered Deposits (000)	543962	5166582	103	4000	13088	49988	9019967
Equity/assets	0.116	0.030	0.068	0.096	0.109	0.128	0.232
Nonaccrual loans/loans (%)	0.794	1.103	0.000	0.090	0.410	1.020	5.930
Interest income/assets	0.010	0.002	0.006	0.009	0.010	0.011	0.017
Net income/assets	0.003	0.002	−0.004	0.002	0.003	0.004	0.009
Cash/assets	0.098	0.084	0.010	0.038	0.073	0.130	0.439
Deposits/assets	0.841	0.055	0.639	0.815	0.853	0.881	0.922
Total loans/assets	0.647	0.156	0.177	0.557	0.678	0.764	0.899
Tier 1 ratio	0.169	0.070	0.097	0.125	0.148	0.188	0.485
(B) 2018Q1–2019Q4
Total Assets (000)	3637466	55540401	19311	104635	217935	516669	37919000
Total Loans (000)	2001865	26570796	8198	63278	145047	360329	24539599
Total C&I Loans (000)	444086	6531676	137	5840	15462	45697	5720311
Total Consumer Loans (000)	338526	5408863	66	1613	3928	9952	2088699
Total RE Loans (000)	947581	10935544	1886	40840	103046	270578	12493246
Total Deposits (000)	2809389	41825877	16297	88125	184644	433810	30031316
Core Deposits (000)	2242678	33372256	15363	80374	167890	387915	25754538
Brokered Deposits (000)	494562	4841917	113	4007	12983	48915	7517939
Equity/assets	0.116	0.030	0.068	0.097	0.110	0.129	0.232
Nonaccrual loans/loans (%)	0.804	1.114	0.000	0.090	0.420	1.030	5.980
Interest income/assets	0.011	0.002	0.007	0.010	0.011	0.012	0.017
Net income/assets	0.003	0.002	−0.004	0.002	0.003	0.004	0.009
Cash/assets	0.089	0.079	0.010	0.034	0.064	0.115	0.429
Deposits/assets	0.841	0.055	0.639	0.814	0.853	0.881	0.922
Total loans /assets	0.656	0.156	0.181	0.568	0.688	0.772	0.903
Tier 1 ratio	0.171	0.071	0.097	0.125	0.150	0.190	0.485
(C) 2020Q1–2020Q4
COVID-19 cases per capita (weighted)	0.018	0.023	0.000	0.001	0.008	0.026	0.086
Reduction in spending (weighted)	0.001	0.098	−0.270	−0.026	0.000	0.050	0.192
Change in unemployment (weighted)	0.017	0.018	−0.014	0.005	0.013	0.025	0.075
House price inflation (weighted)	212.511	250.552	0.000	29.811	144.890	280.880	1246.140
Total Assets (000)	4589773	68375249	20606	120793	259883	626354	50044903
Total Loans (000)	2315757	28940032	7822	69425	162777	421346	32217112
Total C&I Loans (000)	563595	7506783	165	8154	23539	73732	7637067
Total Consumer Loans (000)	369174	5694661	70	1678	4002	10199	2490597
Total RE Loans (000)	1055217	11282875	1643	43759	111428	295043	14916778
Total Deposits (000)	3681146	53550255	17625	102295	218862	525972	42169284
Core Deposits (000)	3024123	43677547	16288	93761	200373	479232	35010100
Brokered Deposits (000)	661530	586618	56	3749	13497	54094	12748313
Equity/assets	0.114	0.029	0.067	0.095	0.108	0.126	0.232
Nonaccrual loans/loans (%)	0.773	1.076	0.000	0.080	0.400	0.990	5.780
Interest income/assets	0.010	0.002	0.005	0.008	0.009	0.011	0.016
Net income/assets	0.003	0.002	−0.004	0.002	0.003	0.003	0.009
Cash/assets	0.118	0.091	0.010	0.049	0.094	0.160	0.439
Deposits/assets	0.841	0.056	0.639	0.816	0.854	0.882	0.921
Total loans /assets	0.630	0.154	0.177	0.539	0.656	0.744	0.886
Tier 1 ratio	0.163	0.067	0.097	0.123	0.143	0.179	0.485

This table reports the summary statistics for our main variables. In total, our sample comprises of 4860 banks. Panel (A) shows all bank quarters from 2018Q1–2020Q4, whereas Panels (B) and (C) split the time period as pre-pandemic quarters and the four pandemic quarters of 2020.

Panel (B) shows the summary statistics for 2018Q1–2019Q4 and Panel (C) for the year 2020 (2020Q1–2020Q4). While the average amount of total deposits of commercial banks increased from 2.8 billion USD in the pre-crisis period to 3.7 billion USD in 2020, their deposits to assets ratio remained stable at 84.1%. Looking at the asset side, the average amount of total loans was 2.0 billion USD and increased to 2.3 billion USD during the crisis, while the loans to assets ratio decreased from 65.6% to 63.0%. Banks also managed to keep their equity to assets ratio stable (11.6% vs. 11.4%), whereas they increased significantly their cash to assets ratio (8.8% vs. 11.8%) during the crisis quarters. Moreover, their return on assets have decreased only slightly (0.28% vs. 0.26%).

Daily county-level data on COVID-19 cases is collected from the New York Times.5 This data is subsequently summed up to receive the overall number of cases per county per quarter in 2020. The very first COVID-19 case was in Snohomish county (Washington) in the United States on January 21st. Until the end of March, while several counties had non or only very few cases, others such as Westchester (67,787), Nassau (50,077) or Suffolk (36,038) (all New York) had the most number of cases. State wise, a large variation existed: South Dakota had only 714 cases, while New York, California, and Washington (463,593, 55,712, and 45,203) had the most. As time passed, however, the pandemic quickly spread all across the United States.

Table 3 further reports the univariate differences between banks that were either more or alternatively less affected by the pandemic. To do so, we split the sample into two sub-groups, a treatment group and a control group, where banks are more (less) exposed to COVID-19, as measured by an above (below) median COVID-19 cases per capita measure as defined in Eq. (1) above. Even without controlling for bank, time, or other characteristics, we can see that on average those banks that are affected more experienced a larger increase in their deposits and real estate loans, both normalized by lagged total assets. As a next step, we analyze in more detail in a multivariate setting how bank deposits and loans were affected by the pandemic.

Table 3

Differences between high & low COVID-19 exposure.

	Treatment group		Control group		Difference	t-stat
	Mean	Std. dev	Mean	Std. dev
Δ deposits/lagged assets	0.048	0.058	0.040	0.051	−0.007***	(−7.517)
Δ core deposits/lagged assets	0.048	0.055	0.041	0.049	−0.007***	(−7.713)
Δ brokered deposits/lagged assets	−0.000	0.011	−0.000	0.010	−0.000	(−0.598)
Δ real estate loans/lagged assets	0.006	0.020	0.004	0.018	−0.002***	(−5.441)
Δ consumer loans/lagged total assets	0.000	0.003	0.000	0.003	0.000	(0.270)
Δ C&I loans/lagged total assets	0.011	0.032	0.008	0.027	−0.003***	(−5.975)
Equity/assets	0.111	0.029	0.114	0.029	0.002***	(4.343)
Nonaccrual loans/loans	0.722	0.997	0.799	1.122	0.077***	(4.146)
Interest income/assets	0.009	0.002	0.009	0.002	−0.000	(−0.182)
Net income/assets	0.003	0.002	0.003	0.002	0.000*	(2.514)
Cash/assets	0.123	0.091	0.124	0.095	0.001	(0.865)
Deposits/assets	0.840	0.057	0.844	0.055	0.004***	(3.914)
Total loans/assets	0.637	0.154	0.612	0.153	−0.025***	(−9.147)
Tier 1 ratio	0.159	0.063	0.168	0.069	0.009***	(6.159)
COVID-19 cases per capita (weighted)	0.025	0.023	0.022	0.024	−0.003***	(−7.459)
Observations	6544		6541		13085

This table reports the univariate differences between treated and control banks in the year 2020, where treated (control) banks are those with local branches above (below) median exposure to COVID-19 infections as measured by the weighted, lagged, COVID-19 cases per capita. Columns (5) and (6) report the differences in means and -statistics.

Main results

COVID-19 outbreak and bank deposits

Our primary goal is to study whether banks located in the counties with a more severe COVID-19 outbreak experienced an increase in their deposits, and, in response, provided liquidity to the economy by a relative increase in their lending. We first focus on banks’ deposits and estimate the following regression model:

The dependent variable is the first difference of total deposits divided by lagged total assets (similar to e.g., Acharya and Mora, 2015, Cornett et al., 2011, Li et al., 2020, Dursun-de Neef and Schandlbauer, 2021). is our bank-specific weighted average COVID-19 cases per capita measure, lagged by one period, as defined in Eq. (1) above. In our main specification, our time period is from the first quarter of 2018 until the fourth quarter of 2020 (12 quarters in total), however, we show in Section 5.3 that our results hold when choosing a different pre-pandemic time period as well. is equal to one for 2020 and zero otherwise. The main coefficient of interest is , which captures the relation between the exposure to COVID-19 and the change in the amount of deposits. is omitted in the regression equation since it becomes nonzero only in 2020 during our time period.

denotes bank controls, also lagged by one period, including bank size, unused loan commitments, and a set of bank characteristics that proxy the confidential CAMELS supervisory rating. The acronym CAMELS refers to the six components of a bank’s condition: Capital adequacy, asset quality, management, earnings, liquidity, and sensitivity to market risk. Banking regulators argue that these six components can provide a comprehensive assessment of a bank’s overall condition (e.g., Lopez, 1999). The following proxies are chosen: Total equity for a bank’s capital adequacy, nonaccrual loans for a bank’s asset quality, net interest income for management quality, return on assets for earnings, cash for liquidity, and deposits for the last acronym sensitivity to market risk. Bank size is defined as the logarithm of total assets and unused loan commitments are the unused loan commitments divided by total assets plus unused loan commitments. Data limitations prevent us from also controlling for the interest rates offered on deposits, which may vary over time as a response to COVID-19 (e.g., Levine et al., 2021). We further control for a set of demographic and economic variables of the counties () that might be related to the local market: population, income per capita, and median age. These are weighted by the geographic distribution of each bank’s deposits similar to our main independent variable, , as shown in Eq. (1). To additionally control for unobserved heterogeneity, bank fixed-effects, , as well as quarter-year fixed-effects, , are included. Finally, following (Petersen, 2009), all standard errors are clustered at the bank and quarter-year level.

Column (1) of Table 4 shows that our main variable of interest, i.e., our bank-specific measure of COVID-19 cases per capita interacted with the 2020 crisis dummy, is not significant for total deposits. However, we expect that households located in the U.S. increased their savings in their deposit accounts, which would be reflected on banks’ core deposits, whereas brokered deposits are obtained through a deposit broker. When we examine core and brokered deposits separately, we are able to confirm this conjecture. We find that an increase in banks’ exposure to COVID-19 led to a significant increase in banks’ core deposits. According to the coefficient estimate reported in column (2), a 1 percentage point higher exposure to the pandemic (i.e., one percentage point higher weighted average COVID-19 cases per capita) led to a significant increase in banks’ core deposits by 0.23 percentage point of their total assets. In our sample, the average bank has total assets of 4.59 billion in 2020 and the average county has a population of 104,816. This implies that a 100-unit increase in the weighted average COVID-19 cases, on average, resulted in about 1 million USD increase in banks’ core deposits. On the other hand, brokered deposits, as shown in column (3), did not change with the COVID-19 exposure.

Table 4

The effect of COVID-19 on bank deposits.

	(1)	(2)	(3)
	ΔTotaldeposits/laggedassets	ΔCoredeposits/laggedassets	ΔBrokereddeposits/laggedassets
Crisis × Weighted_Covid_19	0.149	0.228*	−0.018
	(0.101)	(0.112)	(0.016)
Unused commitments	0.174***	0.114***	0.041***
	(0.036)	(0.035)	(0.011)
Equity	0.174	0.140	0.024
	(0.122)	(0.118)	(0.023)
Nonaccrual loans	−0.002**	−0.001*	−0.000
	(0.001)	(0.001)	(0.000)
Interest income	0.805	1.371	−0.554*
	(1.026)	(0.918)	(0.274)
Net income	−0.451	−0.545*	0.033
	(0.377)	(0.302)	(0.142)
Cash	−0.160***	−0.139***	−0.016**
	(0.029)	(0.024)	(0.006)
Deposits	−0.509***	−0.403***	−0.076***
	(0.061)	(0.050)	(0.012)
Size	−0.175***	−0.157***	−0.010***
	(0.026)	(0.025)	(0.003)
Weighted county population	0.013	0.014*	−0.001
	(0.008)	(0.008)	(0.002)
Weighted income per capita	0.045***	0.042***	−0.002
	(0.012)	(0.009)	(0.003)
Weighted median age	−0.005	−0.007	0.006
	(0.016)	(0.014)	(0.006)
Bank and time FE	Yes	Yes	Yes
Observations	54374	54292	54358
Adjusted R^2	0.221	0.215	0.022

This table shows how total, core, and brokered deposits changed with the COVID-19 crisis. All dependent variables are defined as the first differences divided by lagged total assets. is our bank-specific weighted average COVID-19 cases per capita measure. is a dummy variable that is one for 2020 and zero otherwise. Bank and quarter-year fixed effects are included. The robust standard errors, clustered at the bank and quarter-year level, are reported under the coefficients. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

To investigate the first hypothesis H1, i.e., whether the “forced savings” motivation was a driver behind the increase in core deposits, we examine whether the reduction in spending has had an impact on the increase in bank deposits. According to the “forced savings” view, we expect that banks operating in the counties with a higher reduction in spending would experience a larger increase in their deposits. To investigate this, we use the county-level credit/debit card spending during the pandemic, which is provided by the ‘Opportunity Insights Data Dictionary’ (‘Affinity Solutions’), and calculate the weighted average reduction in spending at the bank level, similarly to the calculation of the weighted average COVID-19 cases per capita as described in Eq. (1) above.6 Banks with are banks whose quarterly reduction in spending is above the median, i.e., these banks are located in the counties with a higher reduction in spending. We then estimate the following triple interaction regression model:

Table 5 shows the results. In line with hypothesis H1, we find that banks’ total and core deposits increased primarily for those banks that are active in the regions with a higher reduction in spending. As reported in columns (1) and (2), these banks experienced a significant increase in their core deposits by almost 0.4 percentage point of their total assets for 1 percentage point higher exposure to the pandemic and, as a consequence, their total deposits increased by 0.33 percentage point of their total assets. Brokered deposits, on the other hand, did not increase, but rather decreased.

Table 5

The effect of COVID-19 on bank deposits: spending reduction.

	(1)	(2)	(3)
	ΔTotaldeposits/laggedassets	ΔCoredeposits/laggedassets	ΔBrokereddeposits/laggedassets
Crisis × Weighted_Covid_19	−0.020	0.015	0.009
	(0.195)	(0.211)	(0.009)
Crisis × Weighted_Covid_19 × High spending reduction	0.326*	0.385**	−0.045**
	(0.160)	(0.164)	(0.019)
Crisis × High spending reduction	0.006	0.005	0.001***
	(0.005)	(0.005)	(0.000)
High spending reduction	−0.007*	−0.007*	−0.000*
	(0.003)	(0.004)	(0.000)
Unused commitments	0.175***	0.118***	0.031***
	(0.022)	(0.025)	(0.004)
Equity	0.318***	0.286***	0.008
	(0.074)	(0.068)	(0.012)
Nonaccrual loans	−0.001*	−0.000	−0.000
	(0.000)	(0.000)	(0.000)
Interest income	3.430***	3.452***	−0.159
	(0.812)	(0.694)	(0.126)
Net income	−0.810**	−0.746**	−0.032
	(0.275)	(0.253)	(0.042)
Cash	−0.150***	−0.128***	−0.012***
	(0.026)	(0.022)	(0.003)
Deposits/assets	−0.405***	−0.313***	−0.055***
	(0.062)	(0.049)	(0.010)
Size	−0.081***	−0.071***	−0.006***
	(0.012)	(0.012)	(0.002)
Weighted county population	0.006	0.006	0.000
	(0.004)	(0.004)	(0.001)
Weighted income per capita	0.031***	0.029***	−0.001
	(0.007)	(0.006)	(0.001)
Weighted median age	−0.004	−0.005	0.001
	(0.009)	(0.008)	(0.002)
Bank and time FE	Yes	Yes	Yes
Observations	54346	54266	54358
Adjusted R^2	0.326	0.307	0.006

This table shows how total, core, and brokered deposits changed with the COVID-19 crisis. All dependent variables are defined as the first differences divided by lagged total assets. is our bank-specific weighted average COVID-19 cases per capita measure. is a dummy variable that is one for 2020 and zero otherwise. High spending reduction is an indicator variable that is equal to one if the quarterly reduction in spending (as measured by Opportunity Insights) in 2020 is above the median, and zero otherwise. Bank and quarter-year fixed effects are included. The robust standard errors, clustered at the bank and quarter-year level, are reported under the coefficients. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

These results imply that the increase in deposits happened mainly for banks located in the counties with a higher reduction in spending. This is in accordance with the “forced savings” view that depositors accumulated savings in their deposit accounts as a result of not being able to spend due to the pandemic and subsequent lockdowns. Stated differently, higher COVID-19 infections, which led to a significant decrease in spending due to the reduction of e.g., leisure activities, resulted in an increase in households’ savings in their deposit accounts.

Cross-sectional analyses

If households saved money as a result of the decrease in their spending due to the pandemic, one expects that they could save money only if they did not lose their jobs. Following this argument, we expect that banks would experience an increase in their deposits only if the counties had a lower increase in the unemployment rate as well. To further explore this, we download county-level unemployment data from the Bureau of Labor Statistics and calculate the weighted average unemployment rate at the bank level. The change in the weighted average unemployment rate from December 2019 to December 2020 is used to examine how much the unemployment rate changed. We define a bank as facing a higher increase in unemployment () if the increase in the weighted average unemployment is above the median and as lower increase in unemployment otherwise (). We subsequently split the sample into these two subgroups and estimate the above described regression (3) separately for each.

According to our results reported in Table 6, a higher exposure to COVID-19 led to an increase in banks’ deposits only if they operate in the counties with a low unemployment increase. We find that banks with a 1 percentage point higher exposure to COVID-19 had a significant increase in their total deposits by about 0.42 percentage point of their total assets if they were located in the counties with a high reduction in spending and a low increase in unemployment. Differentiating again between core and brokered deposits, columns (2) and (3) show that the increase happens only in core deposits. These findings support the argument that households could save only if they did not lose their jobs.

Table 6

The effect of COVID-19 on bank deposits: spending reduction and unemployment increase.

	(1)	(2)	(3)	(4)	(5)	(6)
	ΔTotaldeposits/ laggedassets	ΔCoredeposits/ laggedassets	ΔBrokereddeposits/ laggedassets	ΔTotaldeposits/ laggedassets	ΔCoredeposits/ laggedassets	ΔBrokereddeposits/ laggedassets
	Low unemployment increase			High unemployment increase
Crisis × Weighted_Covid_19	−0.061	−0.001	−0.006	0.240**	0.320**	0.001
	(0.216)	(0.241)	(0.026)	(0.090)	(0.111)	(0.020)
Crisis × Weighted_Covid_19 × High spending reduction	0.421**	0.416**	−0.031	−0.251	−0.233	−0.056
	(0.154)	(0.164)	(0.041)	(0.167)	(0.190)	(0.045)
Crisis × High spending reduction	0.005	0.005	0.000	0.009*	0.008	0.001**
	(0.003)	(0.003)	(0.000)	(0.004)	(0.005)	(0.000)
High spending reduction	−0.005**	−0.005**	−0.000	−0.003	−0.002	0.000
	(0.002)	(0.002)	(0.000)	(0.002)	(0.002)	(0.000)
Unused commitments	0.126***	0.071**	0.032***	0.228***	0.170***	0.029***
	(0.028)	(0.030)	(0.006)	(0.033)	(0.034)	(0.006)
Equity	0.386***	0.343***	0.019	0.268**	0.243***	0.003
	(0.087)	(0.088)	(0.014)	(0.087)	(0.076)	(0.015)
Nonaccrual loans	−0.001	−0.000	−0.000	−0.001*	−0.001	0.000
	(0.000)	(0.000)	(0.000)	(0.001)	(0.001)	(0.000)
Interest income	3.482***	3.440***	−0.230	3.217***	3.262***	−0.057
	(1.073)	(0.949)	(0.183)	(0.835)	(0.795)	(0.146)
Net income	−0.922**	−0.859**	0.001	−0.738**	−0.655***	−0.076
	(0.381)	(0.377)	(0.096)	(0.241)	(0.208)	(0.062)
Cash	−0.150***	−0.133***	−0.009**	−0.147***	−0.122***	−0.014***
	(0.027)	(0.023)	(0.004)	(0.027)	(0.022)	(0.004)
Deposits	−0.470***	−0.369***	−0.062***	−0.352***	−0.268***	−0.049***
	(0.061)	(0.048)	(0.014)	(0.061)	(0.049)	(0.010)
Size	−0.090***	−0.081***	−0.006***	−0.073***	−0.063***	−0.006**
	(0.015)	(0.014)	(0.001)	(0.013)	(0.012)	(0.002)
Weighted county population	0.007	0.007	−0.000	0.003	0.002	0.001
	(0.005)	(0.005)	(0.001)	(0.007)	(0.006)	(0.001)
Weighted income per capita	0.022***	0.021**	−0.001	0.038***	0.036***	−0.001
	(0.006)	(0.008)	(0.002)	(0.009)	(0.008)	(0.001)
Weighted median age	−0.005	−0.006	0.001	−0.017	−0.018	0.001
	(0.009)	(0.010)	(0.003)	(0.015)	(0.014)	(0.002)
Bank and time FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	28018	27971	28030	26322	26289	26322
Adjusted R^2	0.321	0.303	0.003	0.335	0.316	0.009

This table shows how total, core, and brokered deposits changed with the COVID-19 crisis. All dependent variables are defined as the first differences divided by lagged total assets. is our bank-specific weighted average COVID-19 cases per capita measure. is a dummy variable that is one for 2020 and zero otherwise. High spending reduction is an indicator variable that is equal to one if the quarterly reduction in spending (as measured by Opportunity Insights) in 2020 is above the median, and zero otherwise. Banks with high (low) unemployment increase are the ones that experienced an above (below) median increase in their local, weighted, unemployment rate between 2020Q4 and 2019Q4. Bank and quarter-year fixed effects are included. The robust standard errors, clustered at the bank and quarter-year level, are reported under the coefficients. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

COVID-19 outbreak and bank lending

It is well known that banks finance their lending mostly via their deposits (see e.g., Dursun-de Neef and Schandlbauer, 2020). We thus expect that an increase in deposits might lead to an increase in banks’ lending during the pandemic. Given that there was a significant increase in both housing demand and house prices during COVID-19 as documented by Zhao (2020) and Bhat et al. (2021), one might expect a relative increase in banks’ real estate loans. In the previous section, we show that the increase in total deposits with respect to the COVID-19 exposure happened mainly for banks located in counties with a higher reduction in spending. Following this, one further expects a relative increase in banks’ real estate lending mainly for banks active in such high-spending-reduction counties. To investigate these conjectures, we estimate a similar regression model as in Eq. (3), where the dependent variable is the change in real estate loans divided by lagged total assets (similar to e.g., Cornett et al., 2011, Li et al., 2020, Dursun-de Neef and Schandlbauer, 2021). The independent variables and the fixed effects are the same as above, except that we use the one quarter lagged spending reduction indicator to account for the fact that banks’ lending might react with a time lag. The standard errors are again clustered at the bank and quarter-year level.

Table 7 shows the results for real estate loans, consumer loans, and commercial & industrial (C&I) loans. In line with our expectations and with hypothesis H2, we find that higher exposure to the pandemic led to a relative increase in real estate loans for high-spending-reduction banks. As presented in column (1), a 1 percentage point higher COVID-19 exposure led to a significant relative increase in banks’ real estate loans by about 0.06 percentage point of their total assets for banks that are operating in the counties with a higher spending reduction. Other types of loans were not adjusted as a response to COVID-19.

Table 7

The effect of COVID-19 on bank loans: spending reduction.

	(1)	(2)	(3)
	ΔRealestateloans/ laggedassets	ΔConsumerloans/ laggedassets	ΔC&Iloans/ laggedassets
Crisis × Weighted_Covid_19	−0.061**	0.010	0.032
	(0.021)	(0.006)	(0.037)
Crisis × Weighted_Covid_19 × High spending reduction	0.059**	−0.007	−0.135
	(0.020)	(0.004)	(0.087)
Crisis × High spending reduction	0.000	−0.000	0.004
	(0.000)	(0.000)	(0.002)
High spending reduction	−0.001**	−0.000	0.000
	(0.000)	(0.000)	(0.001)
Unused commitments	0.169***	0.002**	0.048***
	(0.010)	(0.001)	(0.005)
Equity	0.060*	0.001	0.109**
	(0.029)	(0.003)	(0.042)
Nonaccrual loans	−0.001***	−0.000**	−0.001***
	(0.000)	(0.000)	(0.000)
Interest income	−1.879***	−0.133**	−0.764***
	(0.271)	(0.043)	(0.152)
Net income	−0.021	0.042**	0.027
	(0.084)	(0.019)	(0.070)
Cash	0.015*	0.001	0.017***
	(0.008)	(0.001)	(0.003)
Deposits	−0.037***	−0.001	−0.001
	(0.009)	(0.001)	(0.010)
Size	−0.016***	−0.001***	−0.009*
	(0.004)	(0.000)	(0.004)
Weighted county population	0.002	0.000	0.004*
	(0.001)	(0.000)	(0.002)
Weighted income per capita	−0.001	0.001*	0.000
	(0.002)	(0.000)	(0.002)
Weighted median age	0.005	−0.002***	0.005
	(0.003)	(0.000)	(0.003)
Bank and time FE	Yes	Yes	Yes
Observations	54349	54358	54368
Adjusted R^2	0.212	0.182	0.469

This table shows how real estate, consumer, and C&I loans changed with the COVID-19 crisis. All dependent variables are defined as the first differences divided by lagged total assets. is our bank-specific weighted average COVID-19 cases per capita measure. is a dummy variable that is one for 2020 and zero otherwise. High spending reduction is an indicator variable that is equal to one if the lagged quarterly reduction in spending in 2020 (as measured by Opportunity Insights) is above the median, and zero otherwise. Bank and quarter-year fixed effects are included. The robust standard errors, clustered at the bank and quarter-year level, are reported under the coefficients. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

We first confirm that banks, which experienced a larger increase in their deposits, were indeed more likely to increase their real estate loans. To do so, we split the overall sample according to the median change in banks’ deposits during the pandemic, i.e., we calculate the difference between the 2020Q4 deposits to asset ratio and the 2019Q4 value, form two sub-groups of banks (those with “high deposits”, and those with “low deposits”), and estimate the above described regression separately for each sub-group. We find in Columns (1) and (2) of Table 8 that the increase in real estate loans comes from the banks that experienced a high deposit increase during the pandemic. A 1 percentage point higher exposure to the pandemic led to a relative increase in banks’ real estate loans by 0.1 percentage point of their total assets for banks that have experienced a larger increase in their deposits, whereas we find no significant change for other banks.

Table 8

The effect of COVID-19 on bank loans: cross-sectional results.

	(1)	(2)	(3)	(4)	(5)	(6)
	ΔRealestateloans/laggedassets
	Low	High	Low	High	Low	High
	Δdeposits	Δdeposits	capital	capital	HPI	HPI
Crisis × Weighted_Covid_19	−0.087*	−0.034	−0.041	−0.069**	−0.023	−0.137***
	(0.046)	(0.020)	(0.026)	(0.027)	(0.029)	(0.035)
Crisis × Weighted_Covid_19 × High spending reduction	0.002	0.110***	0.019	0.091**	0.019	0.161***
	(0.039)	(0.033)	(0.021)	(0.036)	(0.022)	(0.045)
Crisis × High spending reduction	0.000	−0.000	−0.001*	0.001*	0.001	−0.001
	(0.000)	(0.000)	(0.000)	(0.001)	(0.000)	(0.001)
High spending reduction	−0.001*	−0.001	−0.000	−0.001**	−0.001	−0.001
	(0.001)	(0.001)	(0.000)	(0.001)	(0.000)	(0.001)
Unused commitments	0.182***	0.160***	0.157***	0.181***	0.156***	0.203***
	(0.014)	(0.013)	(0.016)	(0.011)	(0.010)	(0.018)
Equity	0.052	0.088*	0.032	0.062	−0.006	0.174***
	(0.038)	(0.041)	(0.026)	(0.035)	(0.028)	(0.051)
Nonaccrual loans	−0.001***	−0.001***	−0.002***	−0.001***	−0.001***	−0.001**
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Interest income	−1.777***	−2.145***	−2.099***	−1.727***	−1.934***	−1.717***
	(0.229)	(0.477)	(0.401)	(0.324)	(0.302)	(0.391)
Net income	−0.190	0.102	0.001	−0.038	0.018	−0.143
	(0.124)	(0.113)	(0.154)	(0.080)	(0.102)	(0.163)
Cash	0.014	0.016	0.013	0.015	0.010	0.021*
	(0.009)	(0.009)	(0.007)	(0.010)	(0.007)	(0.010)
Deposits	−0.052***	−0.019	−0.031**	−0.045***	−0.036***	−0.040***
	(0.009)	(0.011)	(0.010)	(0.012)	(0.010)	(0.011)
Size	−0.013**	−0.018***	−0.019***	−0.014***	−0.024***	−0.001
	(0.005)	(0.005)	(0.005)	(0.004)	(0.005)	(0.004)
Weighted county population	0.002	0.001	−0.002	0.005**	0.001	0.001
	(0.002)	(0.002)	(0.001)	(0.002)	(0.002)	(0.004)
Weighted income per capita	0.002	−0.003	0.001	−0.002	−0.001	0.001
	(0.003)	(0.002)	(0.004)	(0.002)	(0.003)	(0.005)
Weighted median age	−0.001	0.010**	0.010*	0.001	0.006	−0.004
	(0.005)	(0.004)	(0.005)	(0.008)	(0.004)	(0.010)
Bank and time FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	25582	25597	26789	26022	39483	13328
Adjusted R^2	0.215	0.205	0.207	0.212	0.192	0.259

This table shows how real estate loans changed with the COVID-19 crisis. All dependent variables are defined as the first differences divided by lagged total assets. is our bank-specific weighted average COVID-19 cases per capita measure. is a dummy variable that is one for 2020 and zero otherwise. High spending reduction is an indicator variable that is equal to one if the lagged quarterly reduction in spending in 2020 (as measured by Opportunity Insights) is above the median, and zero otherwise. Banks with high (low) Deposits have above (below) the median changes in deposits to assets ratios between 2020Q4 and 2019Q4. Banks with high (low) capital have above (below) the median 2019Q4 capital to assets ratios. Banks that are above (below) the 75th percentile weighted average HPI (2019Q4) are high HPI (low HPI). Bank and quarter-year fixed effects are included. The robust standard errors, clustered at the bank and quarter-year level, are reported under the coefficients. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

As a next step, we examine whether banks’ pre-pandemic capital ratios affect their lending behavior as one expects that banks with lower capital ratios might be more hesitant to increase their loans. This could be explained with their tendency to lower their risk taking in an attempt to avoid any reductions on their capital during the pandemic (Cornett et al., 2011) or with their lower performance during times of distress (Berger and Bouwman, 2013). To analyze this conjecture, we split the sample according to the median 2019Q4 equity to assets ratios. Table 8 highlights that bank loans were indeed primarily increased for banks whose pre-pandemic capital ratios were higher. Column (4) shows that a 1 percentage point higher exposure to the pandemic led to a relative increase in banks’ real estate loans by almost 0.1 percentage point of their total assets for banks with higher pre-pandemic capital ratios. This finding highlights the importance of bank capital for loan provision during crises (Kashyap and Stein, 2000, Kishan and Opiela, 2000, Meh and Moran, 2010, Puri et al., 2011, Dursun-de Neef, 2019, Dursun-de Neef et al., 2022).

Last, we divide our sample as banks operating in counties with higher versus lower house prices. House price indices (HPIs) at the county level are downloaded from the Federal Housing Finance Agency (Bogin et al., 2019). We calculate the bank-level weighted average HPIs similar to the weighted average COVID-19 cases per capita and define banks that are above the 75th percentile weighted average HPI (2019Q4) as high-HPI banks and the ones below as low-HPI banks. We find that the significant relative increase in the real estate loans happened mainly for the banks that are active in counties with higher house prices. As reported in Column (6) of Table 8, a 1 percentage point higher exposure to the pandemic led to a relative increase in banks’ real estate loans by almost 0.2 percentage point of their total assets for banks located in counties with higher house prices.

Overall, these findings imply that banks that experienced an increase in their deposits issued relatively more real estate loans during the pandemic. This relative increase is seen mainly for banks with higher pre-pandemic capital ratios and for banks that were located in counties with higher house prices.

Robustness checks

In this subsection, we show that our main results are stable to using alternative regression specifications. First, columns (1) to (3) of Tables A1 and A2 in the Online Appendix show that changing the start of the sample period from 2017 to 2019 does not alter the size or the significance of our main results, both for core deposits and for real estate loans.

Second, in our empirical identification, we assume that banks that have higher exposure to COVID-19 would have increased their deposits and their loans in a similar rate in the absence of the COVID-19 shock compared to the banks that have lower exposure, i.e., the selection into different levels of exposure was random. To support this assumption, we examine whether exposure to COVID-19 has any significant impact on the changes in banks’ core deposits and real estate loans in the pre-pandemic period from 2017Q1 to 2018Q4. As reported in columns (4) of Tables A1 and A2, our results show that the parallel trends assumption holds for our analysis. This implies that the exposure to COVID-19 is random across banks such that banks’ exposure to COVID-19 does not explain the trend in these bank characteristics in the pre-pandemic period.

Last, Table A3 in the Online Appendix shows that the above described increase in core deposits is also visible when differentiating between insured and uninsured deposits, where the former consists of deposits below $250,000, which are insured by the FDIC. According to our results, the increase happens only in insured deposits. This finding supports our conjecture that it is the households that increased their deposits during the pandemic as we expect that the majority of households are small savers with a savings account below the insurance limit.

Conclusion

In this paper, we examine the motivation behind the increase in bank deposits during COVID-19 and further investigate whether those banks that experienced an increase in their deposits used this additional funding to issue more loans during the pandemic. We first analyze whether the accumulation of deposits was a result of a reduction in spending due to the pandemic. According to this “forced savings” view, not being able to spend led to windfall increases in deposit accounts. Next, we study whether banks that experienced an inflow of deposits had a relative increase in their lending.

To investigate this, we use the weighted average COVID-19 cases per capita as the bank-level exposure to the pandemic, where the weights are the fraction of bank deposits in each county. We find that banks with a higher exposure to the pandemic experienced a significant increase in their core deposits. Our results show that a 1 percentage point higher COVID-19 exposure led to a significant increase in banks’ core deposits by about 0.23 percentage point of their total assets. Consistent with the “forced savings” view, we find that the increase in core deposits happened mainly for banks that are located in the counties with a larger reduction in spending: A 1 percentage point higher exposure to the pandemic significantly increased these banks’ core deposits by almost 0.4 percentage point of their total assets. This is reflected as a significant increase in total deposits by about 0.33 percentage point of their total assets. In addition, we find that such “forced savings” was more pronounced for banks located in counties with a lower increase in unemployment.

As a result of the large inflow of deposits, one might expect to see a relative increase in bank lending during the pandemic. We find that banks with a higher exposure to the pandemic increased their real estate loans if they were located in counties with a higher reduction in spending: A 1 percentage point higher exposure to COVID-19 led to a relative increase in banks’ real estate loans by about 0.06 percentage point of their total assets. We show that the increase happened mainly for banks with higher capital ratios and for banks located in counties with higher house prices.

To summarize, our results indicate that, during the pandemic, households were not able to spend money on leisure activities due to the mobility restrictions, and, as a result, they accumulated savings in their deposit accounts. This led to a large increase in bank deposits and banks, in turn, used these additional funds to issue more real estate loans. This suggests that policies that might have an impact on household spending would generate changes in the volume of deposits in the banking system, which have consequences on loan supply and economic activities, consistent with the findings of Becker (2007) (see also, Chang et al., 2010, Acharya et al., 2011, Lin, 2020, Blickle, 2022).

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.