key: cord-0973508-8l7ubjyi
authors: Thorbecke, Willem
title: Understanding the transmission of COVID-19 news to French financial markets in Early 2020
date: 2022-02-07
journal: International Economics
DOI: 10.1016/j.inteco.2022.02.001
sha: 18bfd5afb8a978196d1294329863b39f73936518
doc_id: 973508
cord_uid: 8l7ubjyi

News of COVID-19 cases roiled the French stock market in 2020. Finance theory indicates that changes in returns across many assets are driven by economy-wide rather than firm-specific factors. To identify these factors, this paper investigates the time series exposure of 174 French assets to macroeconomic variables. It then uses these exposures to examine the cross-sectional pattern of asset price changes due to coronavirus news. The results indicate that investors responded to COVID-19 news by bidding down the prices of assets that do badly when oil prices fall and the euro appreciates and by bidding up the prices of assets that do well when the European Central Bank eases. Banking sector stocks were not harmed by COVID-19 news, indicating that fears of a sovereign-bank nexus were not driving the response.

Coronavirus news roiled financial markets in early 2020. The French stock market fell 44 percent during the first three and a half months of 2020. A year later it had yet to regain its pre-crisis value. The pandemic could flare up again. Policymakers, investors, and portfolio managers are all concerned about the link between the COVID-19 crisis and stock prices. investigated how increases in the number of new COVID-19 cases in early 2020 affected 10-year sovereign yields in 15 Eurozone countries relative to 10-

year German sovereign yields. Using the local projection methods of Jorda (2005) , they reported that an increase of ten COVID-19 cases per million people between 2 January and 5

March 2020 raised sovereign bond spreads by 0.021 percentage points (ppt) immediately and by 0.24 ppt after five business days. When Ortmans and Tripier extended their sample period beyond 12 March, they found that increases in COVID-19 cases no longer affected spreads.

Employing samples up to 9 March 2020, they also reported that 10 new cases per million people led to an 11 percent drop in Eurozone stock prices. Extending the sample period, they again found that COVID-19 cases stopped affecting returns. They concluded that interventions by the ECB on 12 March 2020 broke the link between news of COVID-19 cases and turmoil in European stock and bond markets.

On 12 March the ECB announced several steps to help the economy address the pandemic (Arnold and Stubbington, 2020) . These included increasing quantitative easing (QE) purchases in 2020 by €120bn, providing subsidized loans to banks to stimulate small business lending, and offering loans to banks at rates below the yields that banks received from deposits at the ECB. These policies increased liquidity to the banking sector (Couppey-Soubeyran, Perego, and Tripier, 2020) . At the same time, ECB President Christine Lagarde caused controversy by J o u r n a l P r e -p r o o f saying that the role of the ECB was not to reduce yield spreads between the sovereign bonds of Germany and other Eurozone countries. Later in the day however she and other ECB officials downplayed her comment. This paper investigates why COVID-19 cases roiled financial markets before the ECB's intervention on 12 March. These are several reasons why coronavirus news could affect stock returns. Investors might have expected health concerns or legal shutdown requirements to restrict spending on items requiring close contact such as hotels, transportation, and restaurant meals (Chetty et al., 2020) . Concerns that international trade would be hindered and supply chains disrupted could have harmed firms dependent on global value chains (Shih, 2020) . Increased government borrowing to offset the health and economic costs of the crisis could raise sovereign yields and harm banks that hold government bonds . At an extreme, explosive increases in sovereign yield spreads relative to German bond yields could raise the risk of a breakup of the single currency (Jones, 2020) .

The coronavirus crisis also unleashed two disinflationary shocks in Europe. First, it caused the euro to appreciate as investors withdrew funds from investments in other currencies and flocked to the safety of the euro and as they expected the Federal Reserve to lower interest rates more than the ECB would (see, e.g., Martin and Szalay, 2020) . Second, it contributed to a drop in oil prices, as investors foresaw a drop in spending on transportation and as oil producing countries struggled to address the shock. Figure 1 shows the evolution of the euro and Brent crude oil prices as the crisis hit Europe. The euro appreciated by 6 percent between 21 February and 9 March 2020. Since a one-standard deviation shock to the euro equals 0.59 percent, a 6 percent change represents a large appreciation. Brent crude oil prices also fell 53 percent over J o u r n a l P r e -p r o o f this period. News of the coronavirus crisis could have affected stock prices by impacting the euro and crude oil prices.

To investigate why news of the number of coronavirus cases affected financial markets before the ECB's intervention on 12 March 2020, this paper examines in detail the response of French stocks to news of French COVID-19 cases. It first investigates the time series exposure of 174 French assets to the euro, Brent crude oil prices, and several other macro variables over 20 years. It then investigates the cross-sectional relationship between assets' exposures to macro variables with their exposure to COVID-19 cases in 2020. The results indicate that there is a close relationship between stocks harmed by appreciations of the euro and by falls in oil prices and stocks harmed by increases in the number of coronavirus cases. Also, the results indicate that stocks that gain from expansionary ECB policy also gain from increases in the number of cases. This suggests that investors expected an increase in the number of cases to trigger expansionary policy. Finally, there is no evidence that banks in general were harmed by increases in the number of cases. Banks such as Natixis that had extended loans to firms dependent on the oil industry were harmed by increases in the number of cases, but other banks were not. This indicates that the sovereign/bank "doom loop", whereby reductions in the value of banks' holding of sovereign bonds pressure governments to borrow more to bail out banks, was not driving the response of French stock prices to coronavirus cases in early 2020.

The next section presents the theoretical framework motivating the empirical work and reviews the literature. Section 3 discusses the data and methodology. Section 4 presents the results. Section 5 concludes.

The COVID-19 crisis brought a large unexpected drop in French stock returns. Finance theory indicates that the unexpected return on an asset equals the inner product of a vector of factor loadings and a vector of innovations in macroeconomic factors plus an error term capturing idiosyncratic risk:

where URi,t equals the unexpected return on asset i at time t, ij is the factor loading or beta of asset i to factor j, fj represents news about macroeconomic factor j and i is a mean-zero error term. Ross (1975) , Cox, Ingersoll, and Ross (1985) , and others showed that, assuming i is sufficiently uncorrelated across assets that its influence can be diversified away in large portfolios, the absence of arbitrage implies that the expected return on an asset equals the riskfree rate plus the inner product of a vector of factor loadings with a vector of risk premia:

where Ei is the ex-ante required return on asset i, 0 is the risk-free rate, and j is the risk premium associated with factor j (see Ross, 2001) .

Adding equation (2) to equation (1), the actual return (Ri,t) on asset i is given by:

Ross (2001) noted that equation (3) can be viewed as a snapshot of any intertemporal model,

where the factors represent innovations in the underlying state variables. Since idiosyncratic risk in this framework is specific to each asset, large changes across assets will be driven by macroeconomic rather than firm-specific factors. This paper investigates the economy-wide factors that led to a collapse in French stock prices at the outset of the COVID-19 crisis.

Equations (1) -(3) indicate that there is both a time series and a cross-sectional dimension to asset pricing. The coefficient ij captures the time series exposure of asset i to J o u r n a l P r e -p r o o f macroeconomic factor j. At a single point in time, the cross-section of asset returns will respond differently to economy-wide news depending on their betas with macroeconomic variables.

Suppose there is news about a decrease in a macroeconomic factor (i.e., fj < 0). This will cause a predictable change in the cross-section of asset returns. Those negatively impacted (those with large positive values of ij) will see their returns fall more. Those positively impacted (those with large negative values of ij) will see their returns rise more. For instance, news about a fall in oil prices will cause a change in returns that is proportional to the assets' oil price betas. Those assets that benefit from lower oil prices such as airline stocks will see their returns rise and those that are harmed such as oil exploration company stocks will see their returns fall.

This same pattern will obtain if news causes the risk premium associated with a macroeconomic factor to change. Ferson and Harvey (1991) reported that the risk premia vary much more time than the betas. Suppose an asset is harmed by a decrease in oil prices (i.e., i,oil > 0). If investors perceive a greater risk that oil prices will fall, they will require a higher expected return to hold this asset. In terms of equation (2), i,oiloil will increase. The mechanism driving this increase in required returns is that investors in aggregate will seek to sell assets damaged by lower oil prices, driving down the prices of these assets and raising their expected returns (see, e.g., Fischer and Merton, 1984, and Thorbecke, 2000) . Thus, either a decrease in oil prices (a decrease in foil in equation (3)) or an increase in the required return on assets harmed by lower oil prices (an increase in i,oiloil in equation (3)) will cause returns (Ri,t in equation (3)) for assets harmed by lower oil prices to fall more than returns for assets that benefit from lower oil prices. Thus, when investors are reacting to news of a macroeconomic variable such as oil prices, the returns on a cross-section of assets will vary proportionally to the betas of these assets with the macroeconomic variable.

Identifying the macroeconomic factors that matter for asset returns is more of an art than a science (see, e.g., Chen, Roll, and Ross, 1986 Parlapiano, Alexeev, and Dungey (2017) investigated the exposure of European firms to exchange rate fluctuations. They used the orthogonal market model to take account of common drivers of exchange rates and equity prices. Employing monthly stock return data from 1999 to 2011, they reported that many European firms are exposed to exchange rate changes. Thorbecke (2021) found that many French firms are exposed not only to exchange rates but also to the economic state in the rest of the world (ROW). He represented exchange rates Another variable that is often employed as a state variable is the return on the national stock market. Beginning with Jorion (1990) , most investigations into exchange rate exposure include the return on the country's stock market as a control variable. Researchers use the return on the national stock market to capture the effect of the overall economy on individual stock returns (see, e.g., Warner 1980, 1985) .

Based on these works, this paper employs ECB monetary policy, the price of Brent crude oil, the dollar/euro exchange rate, the return on the world stock market, and the return on the French stock market as macroeconomic factors.

There has also been research on asset returns during the COVID-19 era. Aloui (2021) investigated the impact of ECB QE policies on the euro/dollar exchange rate from July 2007 to This paper adds to this literature by investigating in detail the impact of the coronavirus crisis on the French stock market. It uses more disaggregated data than the papers cited above.

This makes it possible to investigate in detail why news of COVID-19 cases roiled French financial markets in early 2020.

To investigate why COVID-19 news affected French financial markets the first step is to 

where ∆Pi,t is the change in the log of the stock price index for firm or portfolio i, ∆Pm,t is the change in the log of the price index for the French aggregate stock market, ∆Pm,World,t is the change in the log of the price index for the world stock market, ∆Poil,t is the change in the log of the spot price for Brent crude oil, ∆ert is the change in the U.S. dollar/euro exchange rate, ∆MPt represents the change in the two-year French sovereign yield driven by ECB press conferences, and Draghii,t is a dummy variable that equals one on 26 July 2012 when ECB President Draghi said that he would do whatever it takes to save the euro and zero otherwise.

Following Chen, Roll, and Ross (1986) , causality is assumed to flow from the macroeconomic variables on the right-hand side of equation (4) to firm and portfolio returns on the left-hand side and any causality flowing in the other direction is assumed to be second order.

There are no cross-equation restrictions, so the model can be estimated equation-by-equation using ordinary least squares. Given the large sample size (5,216 observations) and the assumption that causality flows from the macroeconomic variables on the right-hand side to the firm-or sectorspecific returns on the left-hand side, ordinary least squares should provide precise estimates of the time series parameters.

The second step is to examine the exposure of the assets on the left-hand side of equation (4) to the number of COVID-19 cases. This is investigated over the sample period from the beginning of January 2020 to 13 March 2020. Ortmans and Tripier (2021b) reported that news J o u r n a l P r e -p r o o f of the number of cases had the strongest impact on Eurozone asset prices over this period. Data on the number of cases in France are obtained from Our World in Data. 2 The following regression is estimated:

NumCases represents the number of COVID-19 cases and the other variables are defined after equation (4). The number of cases is included in first difference form since it trends upwards over the sample period. As a robustness check it is also included in one specification in level form. The variable Draghii,t from equation (4) is excluded from equation (5) because it only takes on values of zero over the 1 January 2020 to 13 March 2020 sample period.

The third step is to examine the cross-sectional relationship between the fall in returns due to news of COVID-19 cases and the macroeconomic variables. Equation (1) indicates that news of a change in a macroeconomic variable should cause returns to change proportionally to assets' regression coefficients to the macroeconomic variable. Even if the macroeconomic variable did not change but if investors are reacting to possible changes in the future, the crosssection of asset returns should react proportionally to the assets' beta coefficients.

From equation (5), ̂6 , ∑ ∆ =1 indicates the cumulative fall in ∆Pi,t driven by news of COVID-19 cases in France between 1 January 2020 and 13 March 2020. Since

is constant across all assets, the same information is contained in the ̂6 , coefficients alone. To investigate the cross-sectional relationship between the change in asset returns driven by news of new cases in France and the macroeconomic variables, the 2 The website for OWID is https://ourworldindata.org/coronavirus-data .

J o u r n a l P r e -p r o o f estimated value of β6 from equation (5) is regressed on the estimated αi coefficients from equation (4):

The hats above the variables indicate that they are estimated values. 3 Changei represents the change in the value of asset i between 1 January 2020 and 20 January 2021. This variable would be informative if investors had foresight about how assets would fare during the pandemic period. When estimating equation (6) 

Table 1 presents the coefficients for the number of cases from estimating equation (5).

Column (2) presents stocks' exposures to the first difference of the contemporaneous number of cases and column (5) presents stocks' exposure to the level of the contemporaneous number of new cases. Columns (3) and (6) present the standard errors and columns (4) and (7) Regressing the coefficients in column (2) on the coefficients on column (5) yields a coefficient of 1.08 and a t-statistic greater than 31.

3 Equation (6) does not directly estimate equation (2) assets' response to news of COVID cases. This cross-sectional variation should help to provide precise estimates of the parameters in equation (6).

By multiplying the beta coefficients in Table 1 To investigate the cross-sectional relationship between the change in returns driven by COVID-19 news and the macroeconomic variables, Table 2 presents the results from estimating equation (6). Column (2) presents the results using the coefficients on the first difference of the number of cases from column (2) of Table 1 as the left-hand side variable. Column (4) presents the results using the coefficients on the level of the number of cases from column (5) of Table 1 as the left-hand side variable. Column (2) indicates that the stocks that did better in the face of COVID-19 news were those that benefit from a decrease in oil prices, a depreciation of the euro, and a decrease in the 2-year French interest rate driven by ECB policy. Since a decrease in interest rates driven by ECB policy corresponds to expansionary monetary policy, this implies that investors responded to COVID-19 news by bidding up the prices of assets that benefit from expansionary monetary policy. The coefficient on the actual change in returns over 2020

indicate that investors responded to COVID news by bidding up the prices of assets that ended up doing well during the pandemic. The Southeast portion of Figure 3a shows those assets that are harmed by oil price falls and that are also damaged by increases in COVID cases. Many of these assets such as CGG, Vallourec, oil equipment and services stocks, Total, and Maurel & Prom are closely related to the oil industry. The Northwest portion shows those assets that are not harmed by oil price falls and that benefit from increases in COVID cases. These assets include medical services stocks and also the stocks of Biomerieux, a maker of medical equipment. News of the medical emergency benefited these stocks. Figure 3b shows that the bank Société Générale in the Northeast of Figure 3b benefits from an increase in cases. The figure also indicates that Société

Générale gains from euro appreciations. This is partly because it has borrowed from abroad. As an increase in the number of cases was associated with a euro appreciation, it benefited Société Générale. On the other hand, the bank Natixis is harmed by increases in the number of cases.

Morris and Smith (2020) reported that Natixis had extended many loans to companies related to the oil sector. As the pandemic caused oil prices to crash, it damaged Natixis. Figure 3b also shows that the banking sector overall is unaffected by increases in the number of cases.

France's luxury brands, including LVMH, Hermès, Kering, and Christian Dior, always have positive coefficients in Table 1 and in several cases the coefficients are statistically significant. Unlike many French firms, these company's stocks are either not harmed or only slightly harmed by appreciations. These firms have pricing power (see, e.g., Goldstein 2021).

J o u r n a l P r e -p r o o f Thorbecke (2021) also reported that these stocks performed well during the pandemic. Investors thus discounted the impact of the health crisis on these stocks.

The coronavirus crisis caused stock prices in Europe and around the world to collapse between the middle of February and the middle of March 2020. found that increases in COVID-19 cases in early 2020 lowered Eurozone stock prices and raised Eurozone sovereign bond spreads over German bond yields. They reported that the ECB's intervention on 12 March 2020 broke the link between news of COVID-19 cases and turmoil in European stock and bond markets.

This paper investigates why coronavirus news roiled the French stock market before the ECB's actions on 12 March 2020. The results indicate that investors responded to increases in COVID-19 cases in France by bidding down the prices of assets that benefit from stronger oil prices, a weaker euro, and contractionary monetary policy.

Couppey-Soubeyran, Perego, and Tripier (2020) observed that the coronavirus crisis could lead to non-performing loans and threaten bank solvency, leading to sovereign debt and banking crises. There is no evidence that these fears drove the fall in French stock prices in early 2020. Banks such as Natixis that had loaned to firms exposed to oil price falls were harmed by increases in the number of cases. On the other hand, banks such as Société Générale that gain from euro appreciations benefited from increases in the number of cases. The overall banking sector exhibited no exposure to the number of cases. Table 2 indicates that less than 24 percent of the variance in asset prices due to French COVID-19 cases can be explained by plausible macroeconomic variables. It could be that the COVID crisis itself functioned as a systematic state variable that helped explain the cross section J o u r n a l P r e -p r o o f of asset returns in early 2020. This is difficult to investigate because France has not experienced many pandemics recently and because the sample period is short. Nevertheless future research should investigate whether assets exposed to COVID-19 had to pay increments to their required returns and whether the ECB intervention on 12 March 2020 suppressed these risk premia.

J o u r n a l P r e -p r o o f Table 1 as independent variables. Column (2) presents the results using the coefficients from the first difference of the contemporaneous number of cases as the regressand and column (4) presents the coefficients on the level of the contemporaneous number of cases as the regressand. The first six regressors listed in column (1) are the regression coefficients from regressions of the returns on the 174 assets listed in Table 1 on the return on the aggregate French stock market, the return on the world stock market, the change in the log of the spot price for Brent crude oil, the change in the log of the dollar/euro exchange rate, Altavilla et al's (2019) measures of the changes in 2-year French sovereign yields driven by European Central Bank press conferences, and a dummy variable that equals one on 26 July 2012 when ECB President Draghi said that he would do whatever it takes to save the euro and zero otherwise. The sample period used to obtain the coefficients on the right-hand side variables that are used as regressors extends from 22 January 2001 to 19 January 2021. There are 5,216 observations. One additional regressor in column 1 is the actual change in the return on the asset between 1 January 2020 and 19 January 2021. This last regressor could be informative if investors could foresee how the pandemic would affect different firms and sectors. S.E. in columns (3) and (5) Note: The figure presents the scatter plot of assets' exposures to the first difference of COVID-19 cases to assets' exposures to Brent crude oil prices. Assets' exposures to news of the first difference of contemporaneous cases come from a regression of daily returns on the 174 assets listed in Table 1 on the first difference of the contemporaneous number of new cases, the change in the dollar/euro nominal exchange rate, the return on the aggregate French stock market, the return on the world stock market, the change in the log of the spot price for Brent crude oil, and Altavilla et al's (2019) measures of the changes in 2-year French sovereign yields driven by European Central Bank press conferences. The sample period extends from 1 January 2020 to 13 March 2020. Assets' exposures to Brent crude oil prices come from regressions of the returns on the 174 assets on the return on the aggregate French stock market, the return on the world stock market, the change in the log of the spot price for Brent crude oil, the change in the log of the dollar/euro exchange rate, Altavilla et al's (2019) Note: The figure presents the scatter plot of assets' exposures to the first difference of COVID-19 cases to assets' exposures to the dollar/euro exchange rate. Assets' exposures to the first difference of contemporaneous cases come from a regression of daily returns on the 174 assets listed in Table 1 on the first difference of the contemporaneous number of new cases, the change in the dollar/euro nominal exchange rate, the return on the aggregate French stock market, the return on the world stock market, the change in the log of the spot price for Brent crude oil, and Altavilla et al's (2019) measures of the changes in 2-year French sovereign yields driven by European Central Bank press conferences. The sample period extends from 1 January 2020 to 13 March 2020. Assets' exposures to the U.S. dollar/euro exchange rate come from regressions of the returns on the 174 assets on the return on the aggregate French stock market, the return on the world stock market, the change in the log of the spot price for Brent crude oil, the change in the log of the dollar/euro exchange rate, Altavilla et al's (2019) measures of the changes in 2-year French sovereign yields driven by European Central Bank press conferences, and a dummy variable that equals one on 26 July 2012 when ECB President Draghi said that he would do whatever it takes to save the euro and zero otherwise. The sample period extends from 22 January 2001 to 19 January 2021.

There are 53 observations. S.E. in columns (3) and (6) are heteroscedasticity and autocorrelation consistent standard errors

The COVID-19 Pandemic Haunting the Transmission of the Quantitative Easing to the Exchange Rate

Measuring Euro Area Monetary Policy

Lagarde Triggers Investor Jitters as ECB Launches Virus Response

Measuring Economic Policy Uncertainty

Return Connectedness Across Asset Classes Around the COVID-19 Outbreak

Measuring Security Price Performance

Using Daily Stock Returns: The Case of Event Studies

Measuring Geopolitical Risk. International Finance Discussion Paper No. 1222

Economic Forces and the Stock Market

and the Opportunity Insights Team. 2020. How Did COVID-19 and Stabilization Policies Affect Spending and Employment? A New Real Time Economic Tracker Based on Private Sector Data. Opportunity Insights Working Paper. Cambridge: Opportunity Insights

European Banks and the Covid-19 Crash Test

An Intertemporal General Equilibrium Model of Asset Prices

The Variation of Economic Risk Premia

Macroeconomics and Finance: The Role of the Stock Market

The Bubble Contagion Effect of COVID-19 Outbreak: Evidence from Crude Oil and Gold Markets

Here are the European Stocks with Pricing Power to Benefit When Pandemic Ends, According to Citigroup

The Global Interdependence among Oil-equity Nexuses

The Impact of COVID-19 on G7 Stock Markets Volatility: Evidence from a ST-HAR Model

A Dangerous Slip-up from Lagarde

Estimation and Inference of Vector Autoregressions by Local Projections

The Exchange Rate Exposure of US Multinationals

Coronavirus Scare Highlights Euro Role in New 'Carry Trade'. Financial Times

COVID-Induced Sovereign Risk in the Euro Area: When Did the ECB Stop the Contagion?

Quand la BCE A-t-elle Stoppé la Contagion de la COVID-19

Aux Marchés Financiers?

Exchange Rate Risk Exposure and the Value of European Firms

Financial Bubble Implosion and Reverse Regression

The New Palgrave Dictionary of Money and Finance

The Arbitrage Theory of Capital Asset Pricing

COVID-19 Pandemic, Oil Prices, Stock Market, Geopolitical Risk and Policy Uncertainty Nexus in the U.S. Economy: Fresh Evidence from the Wavelet Based Approach

Global Supply Chains in a Post-Pandemic World

The Exposure of French and South Korean Firms to Exchange Rates and the COVID-19 Pandemic: Evidence from the Stock Market

Monetary Policy, Time-Varying Risk, and the Bond Market Debacle of 1994

Acknowledgments: I thank two anonymous referees for excellent comments. This paper was written while I was the 2020 FFJ/Banque de France Fellow. Any errors are my responsibility.