Why do households repay their debt during the COVID-19 crisis? A VAR analysis using neural networks.
Mamatzakis, Emmanuel and Ongena, S. and Tsionas, M.G. (2022) Why do households repay their debt during the COVID-19 crisis? A VAR analysis using neural networks. Working Paper. Birkbeck, University of London, London, UK. (Unpublished)
|
Text
Covid household debt April midas Biron.pdf - Published Version of Record Download (1MB) | Preview |
Abstract
In this paper, we investigate whether COVID-19 has had an impact on household finances, like household debt repayments. To do so, the paper employs a vector autoregressive (VAR) model that nests neural networks and uses Mixed Data Sampling (MIDAS) techniques. We use data information related to COVID-19, financial markets, and household finances. Our results show that household debt repayments’ response to the first principal component of COVID-19 shocks is negative, albeit of low magnitude. However, when we employ specific COVID-19 related data like vaccines and tests the responses are positive, insinuating the complexities. Overall, though, main COVID-19 data such as confirmed cases and confirmed deaths negatively affect household debt repayments. We also report low persistence in household debt repayments. Generalized impulse response functions confirm the main results. As draconian measures, the lockdowns are eased it appears that the COVID-19 shocks are diminishing, and household financial data converge to the levels prior to the pandemic albeit with some lags.
Metadata
Item Type: | Monograph (Working Paper) |
---|---|
Keyword(s) / Subject(s): | COVID-19, household debt, ANN, VAR, MIDAS |
School: | Birkbeck Faculties and Schools > Faculty of Business and Law > Birkbeck Business School |
Research Centres and Institutes: | Accounting and Finance Research Centre |
Depositing User: | Emmanuel Mamatzakis |
Date Deposited: | 04 Oct 2022 05:20 |
Last Modified: | 02 Aug 2023 18:17 |
URI: | https://eprints.bbk.ac.uk/id/eprint/48452 |
Statistics
Additional statistics are available via IRStats2.