Data accessibility. In the electronic supplementary material, we make available codes and data—both expenditure series and necessary covariates—pertaining to the national-, province-and broad category-level data, allowing researchers to fully replicate key COVID-19 results in the paper (figures 1, 2, 4 and table 1) and to conduct their own national, subnational or expenditure-category analysis in this context. Note, however, that this study builds from proprietary card transaction data from BBVA, a Spanish commercial bank. Both the individual-level card source data and aggregations at the postal code level or highly disaggregated category level involve highly sensitive personal information about customers and/or may disclose proprietary commercial information on local bank activities. Therefore, we are unable to render data fully publicly accessible beyond what is deposited in the electronic supplementary material. In particular, we are unable to publicly share replication materials involving: historical time series for Spain, cross-country expenditure data, detailed category of expenditure information or postal code level data. Individual researchers interested in these more detailed datasets should direct their query to BBVA Research. The Spanish Household Budget Survey is publicly accessible data, and can be obtained from the web page of the ‘Instituto Nacional de Estadística’ The data on income at census tract level (CUSEC) from where the income at postal code level is calculated is also public, and can also be obtained from https://www.ine.es/ experimental/atlas/exp_atlas_tab.htm the web page of the ‘Instituto Nacional de Estadística’ The data on incidence of the pandemic at Madrid Health District level can be obtained from: https://www.comunidad.madrid/servicios/ salud/2019-nuevo-coronavirus. Authors’ contributions. V.M.C., S.H. and J.V.R.M. analysed the data and wrote the manuscript. J.R.G., A.O., T.R. and P.R. analysed the data. Competing interests. We declare we have no competing interests. Funding. V.M.C. gratefully acknowledges funding from the Leverhulme Trust and the European Research Council, grant no. 101001221, MICRO2MACRO. S.H. gratefully acknowledges funding from the European Research Council, grant no. 864863. J.V.R.M. gratefully acknowledges the support of the UK Economic and Social Research Council (ESRC), award reference ES/L009633/1. Acknowledgements. The authors thank four anonymous reviewers and editors for their valuable suggestions. We also thank Raman Singh Chhina for superb research assistance, and Gary Koop for inspiring comments on a previous draft.