Supervised machine learning to understand stock returns during the crisis
Firm-level stock returns differ enormously in reaction to COVID-19 news. Davis and al., in this paper, analyze these reactions by processing information found in the “Risk Factors” subsection of the pre-pandemic company 10-K filings using two text-analytic approaches: expert-curated dictionaries and supervised machine learning (ML). Despite methodological differences, dictionary and ML approaches yield remarkably congruent return predictions. Importantly though, ML operates on a vastly larger feature space, yielding richer characterizations of risk exposures and outperforming the dictionary approach in goodness-of-fit. Their analysis indicated that:
- Bad pandemic news triggers large negative abnormal return reactions for firms with high exposures to travel, lodging, traditional retail, energy, aircraft production, residential construction, REITs and restaurants, among others;
- The same news triggers positive abnormal return reactions for firms with high exposures to drug trials, ecommerce, basic foodstuffs, web-based services, video games, financial management, and metals and minerals that feed into supply chains for semiconductors, electrical equipment and cloud computing.
Firm-level Risk Exposures and Stock Returns in the Wake of COVID-19
Authors: Steven J. Davis, Stephen Hansen, Cristhian Seminario-Amez
From: University of Chicago, Imperial College
Big data to analyze expenditure patterns during the crisis
In this paper, Carvalho and al. use the universe of transactions mediated by Banco Bilbao Vizcaya Argentaria, S.A (BBVA) to build a daily expenditure measure and assess the ability of transaction data to capture the economic dynamics in Spain during the current crisis. The main dataset builds up from 1.4 billion individual card transactions, the universe of BBVA transactions—be it from the cards issued via the bank or the Points of Sales terminals it operates—and provides a daily account of expenditures. They find:
- Expenditure in commodities related to basic necessities, such as foodstuffs and health/pharmacy, or that cater goods with very low demand elasticity, such as tobacco, more than doubled during the lockdown period, relative to the same period in the previous year;
- Expenditure in goods and services with higher demand elasticities (related to food and entertainment away from one’s residence, fashion, or personal services) all but vanished;
- By exploring regional variation in expenditure growth, they found no evidence that differential exposure to the pandemic across regions affected regional expenditure dynamics.
Tracking the COVID-19 Crisis with High-Resolution Transaction Data
Authors: Vasco M. Carvalho, Juan R. Garcia, Stephen Hansen, Álvaro Ortiz, Tomasa Rodrigo, José V. Rodríguez Mora, José Ruiz
From: University of Cambridge, BBVA Research, Imperial College, University of Edinburgh
Tracking business closures with “alternative” data
Business exit is common in the U.S., with about 7.5 percent of firms exiting annually in recent years, and is also countercyclical. Both the high level and the cyclicality of exit are driven by very small firms. Information on the number of small firms that go out of business is not readily available as most of them close without declaring bankruptcy. Crane and al. use direct as well as indirect measures to understand the impact of COVID-19 on firm closures. They use data from:
- Womply, a credit card transaction processor, showing the share of firms that have ceased processing point-of-sale transactions;
- Homebase, a provider of clock-in/clock-out tracking software, showing the share of firms that have stopped reporting clock events;
- SafeGraph, a data company that aggregates anonymized location data from numerous mobile device applications in order to provide insights about physical places;
- Yelp, the online platform for business ratings;
- Energy Information Administration which reports the number of U.S. commercial and industrial electricity users;
- Costar which reports commercial real estate vacancy rates.
They find that exit expectations appear historically elevated among small businesses in mining, transportation & warehousing, information, education, leisure & hospitality, and other services (this latter sector includes many “local” businesses like beauty salons, auto mechanics, and churches). There are sectors which appear set to do better than average, such as construction, finance & insurance, real estate, and professional, scientific, and technical services.
Business Exit During the COVID-19 Pandemic: Non-Traditional Measures in Historical Context
Authors: Leland D. Crane, Ryan A. Decker, Aaron Flaaen, Adrian Hamins-Puertolas, Christopher Kurz
From: Federal Reserve Board
