PTDid Covid make banks lend less?
During the pandemic, governments cut interest rates and released money so that credit would reach firms and households. But banks did not always follow the script. The study followed 1,153 banks in 65 countries, from 2016 to 2023, to see where lending slowed down. Explained in 12 screens.
One column per year. Blue: 767 banks in developed countries; red: 386 in emerging countries. The heights are equal on purpose: the same 1,153 banks appear in every year (Table 3 and Appendix A).
In a crisis, the bank stands between the government and those who need credit
The bank takes money from savers and lends it to those who need it. During Covid, governments used banks to keep money flowing: lower interest rates, more liquidity, loan guarantees. But uncertainty also makes banks afraid of not being paid back.
illustration Arrow thickness shows the idea, not numbers. The scenarios summarize sections 1.1, 2.1.4 and 2.1.5; the GDP figures come from studies cited by the dissertation (World Bank; Alon and colleagues, 2023).
Covid killed proportionally more in lower-income countries
Rich countries recorded more cases. But for every 100 cases, deaths were three times more frequent in low-income countries. Emerging economies have fewer resources to protect themselves from a health crisis, and that is what led the study to separate developed and emerging countries.
World Health Organization data up to December 2024, cited in section 2.1.1; World Bank income groups. In the study, "developed" means high income and "emerging" groups the other three.
Five studies looked at lending in the pandemic with the same method
All of them compare a hard-hit group with a less affected one, before and after Covid (the method on screen 07). Bet: did the highlighted group get less, the same or more credit than the other?
1,153 banks in 65 countries, eight years each
Publicly listed commercial banks, with positive equity and assets above US$ 1 million, followed from 2016 to 2023: four years before Covid and four after. That makes 9,224 observations (one bank in one year). Data from Capital IQ, the World Bank and the WHO.
Each dot = 1 bank. Banks and Covid cases from 2020 to 2023 by country: Appendix A (totals checked: 767 and 386 banks).
The study measures lending as a share of the bank’s assets
Loan growth compares this year’s share of loans in total assets with last year’s. Play with the made-up bank on the right.
Made-up bank. Formula from Table 2: growth = (loans ÷ assets this year) ÷ (loans ÷ assets last year) − 1. Our reading: because it is a share, the measure can fall even when the bank lends more, if assets grow faster.
Mean, standard deviation and extremes from Table 3 (full sample). The red circle is the made-up bank.
The study compares how much each group changed after 2020
The method is called difference-in-differences (DID). Countries with Covid cases per million inhabitants at or above the median form the high-contagion group; the rest form the comparison group. The same logic compares emerging and developed countries.
illustration Made-up numbers to show the logic. The method’s premise: without Covid, both groups would have followed the same trend. In the real model, the study also holds constant the banks’ size, cash and capital, GDP growth and the quality of each country’s government.
Matching made the groups more alike
High-contagion banks were different from the others. To compare like with like, the study redid the analysis with propensity score matching (PSM): for each observation in the high-contagion group, it picked the most similar one in the comparison group, without repeats.
Standardized gap between the groups, in % (bias), Table B.1. Zero = identical groups. Country governance came closest; cash barely changed. The table itself flags that the overall imbalance (indicator B: 63.1% → 27.8%) stayed above the 25% reference.
Each hypothesis got a clear answer from the model
The study started from three hypotheses. Before seeing the columns, bet: did the model confirm each one?
Table 5: M1 = main test (1,153 banks), M2 = matched test (947). Height = significance: three steps for p < 0.01, two for p < 0.05. We show only direction and significance, because the size of the coefficients does not fit the scale of the measure (see the limits).
At first glance, lending grew less after 2020
In the simple correlation, the years after Covid go together with lower loan growth. In the model, the "after Covid" term turns positive, which is why the first hypothesis was not confirmed. Tap the three cards.
Correlations from Table 4 (unmatched panel); in the matched one, −0.07 and 0.91. Coefficients and p from Table 5, model 1.
Five features of the banks and the countries also enter the calculation
Theory expected all of them to make lending grow more. Bet on each: with more of it, does lending grow more, less, or is there no clear effect?
Expected sign: Table 2. Sign found: Table 5. "Hint" = p between 0.05 and 0.10. R² from Table 5: the share of the differences in loan growth that the model explains.
What changes for those who regulate, lend or invest
Governments and regulators
Low rates and easy money are not enough if banks pull back. Knowing where lending slows most (the hardest-hit and emerging countries) helps tailor measures to each country.
Banks
Spotting periods of greater uncertainty early, when they come from outside shocks, helps adjust capital before the crisis.
Investors
The different reactions of banks in each group of countries can be taken into account when diversifying portfolio risk.
Researchers
Suggested next steps: check whether who owns the bank changes its reaction, and look at other credit channels, such as receivables funds, peer-to-peer loans and bonds.
What this study does not say
- Only listed banks. Privately held banks were left out for lack of data. · according to the author
- Direction, not size. The coefficients in Table 5 are too large for the scale of the measure (a mean of 0.3% and a standard deviation of 9.6 points), so the page shows only which way lending went and how reliably. · our note
- The model explains little. 3.1% of the differences in loan growth (0.8% with matching): lending depends on much more. · our note
- The group is the country, not the bank. Every bank in a high-contagion country is in the same group, and the US has 37% of the banks. The dissertation does not show whether the groups followed the same trend before 2020. · our note
Credits
ROBIAL, Karina. O impacto da Covid na concessão de crédito bancário. Master’s dissertation (Business Administration), Mackenzie Presbyterian University, São Paulo, 2025. Advisor: Prof. Dr. Michele Nascimento Jucá.
Research supported by CAPES (Brazil’s federal agency for graduate education). Data from Capital IQ, the World Bank and the WHO; analysis in Stata. This version simplifies the work for a general audience; the formulas and full tables are in the original text.