Replication package for: Digital Capital as Liquidity Insurance: Evidence from Brazilian Bank Branches during the Pandemic

Replication package for the paper "Digital Capital as Liquidity Insurance: Evidence from Brazilian Bank Branches during the Pandemic". It contains the public raw data (ESTBAN, COSIF and Pix statistics of the Banco Central do Brasil; fixed-broadband accesses from Anatel; population and GDP from IBGE; COVID-19 deaths from Brasil.io; the Oxford COVID-19 Government Response Tracker; IBGE boundaries), two Python scripts that reproduce every number, table and figure of the paper (DG05_1_estimation.py and DG05_2_figures.py) and a single Excel workbook with all derived data. See README.txt for instructions, the list of sources and the correspondence between each result of the paper and the outputs. Because of the upload limits, the package DG05_replication_package.zip (172.9 MB, 54 files) is stored as 58 split volumes (DG05_replication_package.zip.001 to .058). Download all 58 into the same folder and either open the first volume with 7-Zip, or join them first (Linux/macOS: cat DG05_replication_package.zip.0* > DG05_replication_package.zip; Windows: copy /b DG05_replication_package.zip.0* DG05_replication_package.zip) and unzip the result. MD5 of the joined file: 8c035b21df339a488d4f6e241234cb92. When the COVID-19 pandemic impaired access to bank branches, did banks that had invested in digital capital protect their funding, and did the protection pay for the investment? We follow the branch networks of 92 Brazilian banks (20 of which identify the effect) in 3,197 municipalities from 2017 to 2022 and exploit a simple complementarity: a bank's digital capital can replace its branches only where its customers can go online. Comparing branches of the same bank across municipalities of the same state, and different banks within the same municipality, we find that one standard deviation more pre-pandemic digital capital raised demand deposits by 5.54 per cent in municipalities with one standard deviation more broadband during 2020. The effect has no pre-trend, rises with local connectivity and disappears in 2021. It did not change branch results, and the evidence on credit and on whether the gain created new deposits is inconclusive. Valued at the policy rate of 2020, the funding gained did not pay for the digital capital behind it; under alternative counterfactual assumptions, the implied break-even policy rate ranges from 3.3 to 21.4 per cent. Digital capital insures bank funding against the loss of the physical channel, but the insurance is temporary and, on its own, worth paying for only when interest rates are high and such disruptions recur.