Access the full text.
Sign up today, get DeepDyve free for 14 days.
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
Analyst information intermediation during the COVID-19 pandemic Pawel Bilinski Cass Business School, City, University of London firstname.lastname@example.org This draft 11.01.2021 ABSTRACT We use the COVID-19 pandemic to examine how this tail risk event affected analysts research production and their information intermediation role. Analysts markedly increase their research activity in the initial months of the pandemic: the number of quarterly earnings forecasts increases by 72%, revenue forecasts by 80%, cash flow by 59%, dividend by 11%, target prices by 154% and stock recommendations by 88% in March 2020 compared to the same pre-pandemic month. Forecasts issued during the pandemic associate with significantly higher errors compared to a comparable pre-pandemic period. Analysts aggressively revise their forecasts during the pandemic compared to the pre-pandemic period: the average absolute revisions range between 142% for revenue forecasts and 9% for dividend estimates. Price reactions to revisions in analyst forecasts and stock recommendation are incrementally higher during the pandemic though lower for target prices. This effect is magnified in periods where investors actively search for information about the pandemic and the stock market as captured by google searches. Investors value more analyst private information discovery role than their role in interpreting public information during the pandemic. Keywords: COVID-19; Coronavirus; forecast accuracy; price reactions; information discovery; information intermediation JEL: G01, G14, G32, F14 This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed 1. Introduction The COVID-19 outbreak is a textbook example of a tail-risk shock to the stock market. The quick spread of the virus coupled with the uncertainty on how consumers, governments, and firms will respond to the pandemic resulted in a sudden and market-wide increase in uncertainty. To illustrate, between February 19 and March 23 of 2020, the S&P 500 stock market index lost 33.7%, then surged by 29% between March 24 and April 17. High volatility continued throughout 2020. Higher uncertainty increases investor demand for information that helps assess firm fundamentals and value (Grossman and Stiglitz 1980, Pastor and Veronesi 2012, Amiram, Landsman, Owens and Stubben 2018). We use the COVID-19 pandemic to examine how this tail-risk event affected analysts research production and their information intermediation role in the market. This setting has important econometric benefits. As we focus on firms with analyst coverage before and during the pandemic, the endogeneity concern related to selectivity in analyst coverage is of limited concern. Further, because the pandemic is not caused by firm or market fundamentals, changes in properties of analyst forecasts cannot be explained by potentially omitted correlated variables. This makes the COVID-19 pandemic a natural laboratory to study the impact a tail-risk event has on analyst behavior and informativeness of their research forecasts. We focus on sell-side analysts because previous research documents they play a key role in facilitating information transmission in capital markets (Healy and Palepu, 2001). However, how analysts contribute to market efficiency in tail-risk events remains unclear due to the scarcity of such events. Baker, Bloom, Davis, Kost, Sammon and Viratyosin (2020) document that “[N]o previous infectious disease outbreak, including the Spanish Flu, has impacted the stock market as forcefully as the COVID-19 pandemic” and no other crisis had such a sudden and market-wide impact. Though rare, tail-risk events have profound impact on companies and investors and lead to an exogenous shock to investors’ information demand. Analysts seem well-placed to address the demand shock, This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed however, it is not clear how their research activity will change during the pandemic and how investors will react to potential changes in analyst research supply. On the one hand, higher uncertainty during the pandemic is likely to negatively affect forecast accuracy and investors may discount noisy signals, which in turn reduces analysts’ incentives to supply more research. Further, investors may find it hard to identify relevant information among the flood of COVID-19 news, e.g. th on 20 March 2020 alone around 20,500 news articles about COVID-19 were published online (Wang and Xing 2020). Analysts may also find it challenging to incorporate general market information into their forecasts (Hann, Ogneva and Sapriza 2012 and Hugon, Kumar and Lin2016). On the other hand, analysts may play important private information discovery role in the noisy and information-scarce environment. Wang and Xing (2020) document that corporate SEC disclosure in the first quarter of 2020 “are general and lack specifics”. Analysts may fill this information void. Thus, we empirically examine how analysts change the supply and breadth of their forecast, how the pandemic affected accuracy of their research, and how investors assess the information content of analyst research. We collect a sample of 428,297 quarterly earnings-per-share (EPS) forecasts issued between January 2018 and November 2020 for 4,889 unique firms. We classify forecasts issued from January 2020 as pandemic forecasts because (1) Baker et al. (2020) highlight that the “the COVID-19 volatility surge began in the fourth week of January” and (2) the Q1 results for 2020 will be affected by the pandemic forcing analysts to incorporate the effect of COVID-19 into their forecasts. We consider forecasts issued between January 2018 and December 2019 as pre-pandemic forecasts. To understand how the pandemic affected the breadth of analyst research, we also collect quarterly revenue estimates (SAL), cash flow-per-share forecasts (CPS), and dividend-per-share estimates (DPS) issued jointly with the EPS forecasts. We look at revenue forecasts following the evidence in Our conclusions are the same if we designate the start of the pandemic in the beginning of March 2020. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Ertimur, Mayew and Stubben (2011) that investors use revenue estimates to disaggregate earnings forecasts into revenue and costs estimates and attach more weight to the more persistent revenue component. Cash flow forecasts allows investors to disaggregate earnings estimates into accrual and cash flow estimates allowing to gauge earnings persistence and the likelihood of financial distress (DeFond and Hung 2003, Givoly, Hayn and Lehavy 2009). Dividend forecasts contain incremental information compared to earnings, revenue and cash flow estimates and help investors assess persistence of earnings (Bilinski and Bradshaw 2020). In testing the informativeness of forecast revisions, we also look at analyst target prices and stock recommendations. We first examine changes in the supply of analyst forecasts. We find that compared to the same pre-pandemic months, the number of quarterly EPS estimates is similar in January and February 2020, increases by 72% in March 2020, and remains higher at around 22% between April and July 2020. The number of earnings forecasts is comparable to pre-pandemic levels in August 2020 and reduces from September to November 2020. Similar patterns are evident for other forecasts: the number of revenue forecasts increases in March by 80%, cash flow forecasts by 59% and dividend estimates by 11%, and declines to pre-pandemic levels towards the end of 2020. The number of target prices is 154% and stock recommendations is 88% higher in March 2020 compared to same month before the pandemic. Thus, the analyst initial response to pandemic- induced market uncertainty is to increase their provision of information. Turning to forecast errors, EPS forecast errors increases by 76.8% for Q1 2020 results compared to same quarter before the pandemic and reduce gradually to a 37.5% higher error in quarter 4. Compared to the pre-pandemic period, revenue forecast error is on average 62.1% higher during the pandemic, cash flow forecast error is up by 14.9%, and divined forecast errors are on average 17.9% higher. Thus, research production during the pandemic associates with lower average precision of estimates. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Next, we turn to investor assessment of the informativeness of analyst research as measured by price reaction regressions. Compared to the pre-pandemic period, we observe significant downward revisions in all analyst forecasts during the pandemic: the absolute magnitude of analyst revisions are 112% higher for EPS forecasts, 142% for revenue estimates, 64% for cash flow forecasts and 106% for dividend estimates. We also observe 9% more recommendation revisions and average 57% price target revisions. Revisions in analyst forecasts associate with an average 18% stronger absolute price reaction on the forecast announcement date. Using firm-fixed effects regressions, we confirm that investors react more strongly during the pandemic to revisions in analyst forecasts (EPS, revenue, cash flow and dividend) and in stock recommendations compared to the pre-pandemic period. The economic effects are large: on average, price reactions to revisions are between 42% to over 2 times larger during pandemic compared to pre-pandemic period. Price reactions to target price revisions are 22% lower during the COVID-19 outbreak. Thus, despite lower average forecast accuracy, investors consider analyst forecasts to convey valuable new information during the pandemic. To explain why investors put more weight on comparatively less precise forecasts, we first examine the role analyst research plays in resolving uncertainty during periods of increased investor demand for information. We capture information demand by the intensity of google searches for pandemic and stock market information. We find that price reactions to analyst forecast revisions are incrementally higher during periods of increased google searches. This result is consistent with analyst forecasts responding to investor information demand shock. See Da, Engelbberg and Gao (2011) for tests validating google searches as information demand measure. Bento, Nguyen, Wing, Lozano-Rojas, Ahn, and Simon (2020) document a 36% spike in google searches for information following public announcements of COVID-19 cases. Costola Iacopini and Santagiustina (2020) report that google searches associate with stock price volatility in a cross-section of six countries. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Next, we examine whether during the pandemic investors value analyst private information discovery role more than their role in interpreting corporate information. Chen, Cheng and Lo (2010) document that information discovery dominates in the weeks before firms announce their earnings results and information interpretation is more important in the weeks after earnings announcements. We follow Chen et al. (2010) and focus on EPS forecasts in a 10-day window around earnings announcements excluding a three-day window centered on the earnings announcement day to avoid confounding effects. We find that before the pandemic, price reactions are similar in magnitude before and after earnings announcements, a result consistent with Francis, Schipper and Vincent (2002) and Frankel, Kothari and Weber (2006) who document that investors value both analysts information discovery and interpretative functions. However, during the pandemic, investors value more analyst private information discovery role, a result consistent with greater demand for private information discovery. This study contributes to the accounting literature that examines the capital markets consequences of analyst research. This literature examined the accuracy and price impact of analyst forecasts, and the importance of analyst information discovery role compared to the information interpretative function (Dempsey 1989, Shores 1990, Womack 1996, Loh and Stulz 2011, Ivkovic and Jegadeesh, 2004, Chen, Cheng and Lo 2010). We document how the COVID-19 pandemic, a tail-risk event, affected analysts research production, accuracy of their forecasts and investor assessment of analyst research information content. We find that despite significantly lower accuracy, investors consider analyst research informative, particularly in periods of high demand for information, and when research reveals private information rather than interprets corporate disclosures. Our evidence questions the common assumption in accounting literature that lower accuracy reduces analyst incentives to supply research as investor demand for low precision forecasts is likely to be low (e.g. Ertimur et al. 2011, Givoly et. 2009). Our evidence suggests that in periods of This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed high market uncertainty, lower accuracy does not necessarily reduce the supply nor informativeness of research. Further, the results suggest significant value of analyst information intermediation role during periods of high market uncertainty. This evidence contrasts the view on the declining importance of sell-side analysts in the market stemming from regulatory changes, such as Markets In Financial Instruments Directive II in Europe (Fang, Hope, Huang and Moldovan 2020), declining research budgets, and an increasing shift to passive ownership (Appel, Gormley and Keim 2016). The study also contributes to the growing literature on the impact COVID-19 had on financial markets. Du (2020) uses analyst forecasts issued in March 2020 to examine the timeliness of forecasts by female compared to male analysts. Landier and Thesmar (2020) use earnings forecasts to infer the implied discount rates for largest NYSE, Nasdaq, or Amex stocks during the COVID-19 crisis. Cox, Greenwald and Ludvigson (2020) estimate a dynamic asset pricing model to capture fluctuations in the pricing of stock market risk during the pandemic. Ding, Levine, Lin and Xie (2020) study firm characteristics that predict the magnitude of share price drop in response to COVID-19 outbreak. Baker et al. (2020) document the dynamics of news about the disease between February 2020 and April 2020 and their correlation with the stock market volatility. Ramelli and Wagner (2020) examine the magnitude of price declines to the pandemic. Li, Liu, Mai and Zhang (2021) report that firms with a strong corporate culture outperform their peers with a weak culture during the pandemic. Cejnek, Randl and Zechner (2020) study the effect COVID-19 had on corporate dividend policy, Anginer, Donmez, Seyhun and Zhang (2020) on insider trades, and Tkachenko and Bataeva (2020) on share repurchases. Fahlenbrach, Rageth and Stulz (2020) study the effect financial flexibility has on share price reaction to COVID-19 outbreak. We add to this Bloomberg article highlights that “Research is the niche that’s been buffeted most violently by the forces crashing into the finance industry: technology, regulation and the demands of the marketplace itself.” And “Research spending by the buyside has dropped between 20% and 30% since the new rules [MiFID II] came in, U.K.” , see https://www.bloomberg.com/news/articles/2019-12-19/analyst-jobs-vanish-as-a-perfect-storm-hits-wall-street- research. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed literature by documenting the behavior of analysts in response to the COVID-19 market shock and how investors value analyst research during this period. 2. Data We collect analyst individual quarterly EPS forecasts and contemporaneously issued revenue, cash flow and dividend estimates, target prices and stock recommendations from I/B/E/S over the period January 2018 to November 2020. I/B/E/S imposes a four-month gap between when the data is available for academic compared to commercial research, which determines the end of our sample period. We require that the forecasts have the actual value to calculate forecast errors and share price information on CRSP. Our final sample includes 428,297 EPS forecasts issued for 4,889 unique firms by 3,809 unique analysts employed by 332 unique brokers. Table 1 presents the annual number of forecasts between 2018 and 2020. The fraction of revenue forecasts issued with EPS estimates is 69.3%, a result consistent with Ertimur et al. (2011) that since 2001, almost all analysts produce revenue estimates. We find that 13.6% of cash flow forecasts are issued jointly with EPS estimates, which is twice the fraction of joint EPS and cash flow forecasts reported in DeFond and Hung (2003) for their period 1993-1999 and 9.3% in Bilinski (2014) over the period 2000-2008. Around 4.1% of EPS estimates are issued jointly with a dividend forecast, an evidence consistent with Bilinski and Bradshaw (2020) that dividend forecasts are rare in the U.S.. Target prices are issued with 44.5% of earnings forecasts, and stock recommendations with around 6.9% of EPS estimates. [Table 1] 3. Empirical results 3.1 The changes in the supply and accuracy of analyst forecasts during the pandemic The first test examines changes in the monthly number of analyst forecasts during the pandemic This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed compared to similar periods before the COVID-19 outbreak. This test is useful to understand the analyst supply response to the outbreak of the pandemic. Figure 1 plots the monthly number of EPS, revenue, cash flow and dividend forecasts, target prices and stock recommendations in the pre- pandemic years 2018 and 2019 and during 2020. We identify two main results. First, the number of forecasts is markedly similar in 2018 and 2019, which suggests a certain routine in analyst research production. Second, there is a visible increase in the number of forecasts in March 2020, the month that observed the most dramatic volatility in market indices during the pandemic. Compared to the same pre-pandemic month, the number of quarterly EPS forecasts increases by 72.4% in March 2020, revenue forecasts by 79.5%, cash flow forecasts by 58.7%, dividend estimates by 11%, target prices by 154.3%, and stock recommendations by 88.2%. After this initial increase, we observe a gradual reduction in the number of forecasts and a convergence to pre-pandemic levels around August 2020, with a slight decline below pre-pandemic numbers in the last sample months. Figure 1 evidence is consistent with analysts promptly responding to the onset of market uncertainty caused by the pandemic shock by increasing their information provision. [Figure 1] Next, we examine the accuracy of analyst forecasts. We calculate the forecast error at the analyst level j for firm i for quarter q of fiscal year t, Ferror, as the absolute difference between the actual and forecasted values on day d, scaled by 1 plus the absolute value of the actual: |𝑙𝐴𝑐𝑡𝑢𝑎 −𝑜𝑟𝑡𝐹𝑒𝑠𝑐𝑎 | 𝑖 ,𝑞 ,𝑡 +1 𝑖 ,𝑗 ,𝑑 ,𝑞 ,𝑡 +1 𝐹𝑒𝑟𝑟𝑟𝑜 = . (1) 𝑖 ,𝑗 ,𝑑 ,𝑞 ,𝑡 +1 1+|𝑙𝐴𝑐𝑡𝑢𝑎 | 𝑖 ,𝑞 ,𝑡 +1 This construct is analogous to standard measure of EPS forecast accuracy (Das et al., 1998; Hope, 2003; Richardson, Teoh, and Wysocki 2004). We winsorize forecast errors at the 1% and 99%. Figure 2 presents the mean quarterly forecast error for analyst earnings, revenue, cash flow We cannot reject the null that the number of forecasts each month is similar between 2018 and 2019. Raw data for 2020 used to create Figure 1 is presented in Appendix A. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed 6 and dividend estimates. As with the number of forecasts, we observe comparable forecast errors across all measures in 2018 and 2019 (we cannot reject the null hypothesis that the average forecast error is the same in 2018 and 2019). We observe a significant increase in the average forecast error in Q1 of 2020 compared to the average forecasts error in Q1 for 2018 and 2019. Forecast errors peak in Q2 of 2020 as firm fundamentals start to fully reflect the impact of the pandemic, including state lockdowns. Forecasts errors gradually decline in Q3 and Q4. [Figure 2] We formalize the analysis from Figure 2 in Table 2, which reports the mean quarterly forecast errors for Q1 to Q4 split between pre- and pandemic periods. For each forecast measure, we also calculate the percentage difference between pre- and pandemic forecast errors and the corresponding t-test. We find that during the first two quarters of 2020, all forecast errors were significantly higher compared to the pre-pandemic period. This confirms the graphical evidence from Figure 2. Further, EPS and revenue forecast errors continue to be higher in quarters Q3 and Q4 with the cash flow and dividend forecast errors converging to pre-pandemic levels by quarter Q4. Lower forecast errors in the later quarters are consistent with a gradual relaxation of restrictions in the second half of 2020, which allowed firms to resume pre-pandemic activity. Lower dividend forecast errors towards the end of 2020 also reflect higher resilience of dividends following the initial uncertainty on the extent of dividend cuts. [Table 2] Because target prices have a 12-month forecast horizon, we do not have actual prices to compute target price forecast errors. Median dividend forecast error is zero consistent with the evidence in Bilinski and Bradshaw (2020). The Janus Henderson Global Dividend Index 2021 Report highlights that “the annual review of dividends in the dominant US market was significantly more benign than expected” and that “several companies restored payments that they had suspended earlier in the year” with the total dividend changes in 2020 vs. 2019 of only 2.1% in the U.S.. See https://cdn.janushenderson.com/webdocs/Janus+Henderson+Global+Dividend+Index.pdf This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed The last columns of Table 2 examine the changes in forecasting accuracy over the course of the fiscal year, by comparing average forecasts error in Q4 vs. Q1, for the pre- and pandemic years. The evidence for the pre-pandemic period is mixed: we observe both increases in forecast error over the course of the year, as captured by higher average error in Q4 compared to Q1 for earnings and dividend forecasts, and a decline in forecast error, as captured by lower Q4 vs. Q1 errors for revenue and cash flow forecasts. For the pandemic period, we observe an improvement in forecast errors across all measures, as captured by declining forecast errors between Q4 and Q1, and relative to the pre-pandemic period. To illustrate, the reduction in forecast error between Q4 and Q1 is 5.526 larger for EPS forecasts during the pandemic than compared to the same change before the pandemic. These results likely reflect (i) that analysts are better able to assess the impact the pandemic has on firm fundamentals over time and (ii) the improvement in market conditions towards the end of 2020. 3.2 Price reaction to analyst forecast revisions Our next test looks at price reactions to analyst forecast revisions to assess the informativeness of analyst outputs. We calculate the forecast revision, Δ𝑡𝑎𝑠𝑐𝐹𝑜𝑟𝑒 , as the difference between the analyst current and previous forecast issued for the same fiscal quarter and same firm scaled by the absolute value of the previous forecast, 𝑡𝑐𝑎𝑠𝑒𝑟𝐹𝑜 −𝑡𝑐𝑎𝑠𝑒𝑟𝐹𝑜 𝑖 ,𝑗 ,𝑑 +1,𝑞 ,𝑡 +1 𝑖 ,𝑗 ,𝑑 ,𝑞 ,𝑡 +1 Δ𝑡𝐹𝑜𝑟𝑒𝑎𝑠𝑐 = . (2) 𝑖 ,𝑗 ,𝑑 +1,𝑞 ,𝑡 +1 |𝑡𝑐𝑎𝑠𝑒𝑟𝐹𝑜 | 𝑖 ,𝑗 ,𝑑 ,𝑞 ,𝑡 +1 Using percentage revisions makes forecasts expressed on a per-share basis, e.g. EPS estimates, more comparable with forecasts on non-per-share basis, such as revenue, compared to scaling the forecasts by the share price or market capitalization. We winsorize revisions at 1% and 99%. Quarterly forecasts in Q1 are for the first fiscal quarter and forecasts in Q4 are for the last fiscal quarter, thus the forecasts do not capture the same actual value. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Figure 3 reports the monthly average revisions for the pre-pandemic years 2018 and 2019 and for 2020. We observe very similar magnitudes of revisions across months before the pandemic. EPS revisions tend to be negative, which reflects that analysts tend to start at a high level and firms walk-down forecasts to beatable levels (Richardson et al. 2004, Graham, Harvey and Rajgopal 2005). The picture during the pandemic is markedly different: analysts revise downwards all forecasts starting in March up till June 2020 and revisions become smaller in magnitude towards the end of [Figure 3] To test if forecast revisions associate with significant price reactions, we regress them on a three-day absolute cumulative abnormal return, ACAR, centered on the forecast revision date: | | | | | | | | | | 𝑅𝐴𝐶𝐴 = 𝛼 + 𝛼 Δ𝑃𝑆𝐹𝐸 + 𝛼 Δ𝐹𝑆𝐴𝐿 + 𝛼 Δ𝐹𝐶𝑃𝑆 + 𝛼 Δ𝐹 𝐷𝑃𝑆 + 𝛼 Δ𝐹 𝐶𝑅𝐸 𝑑 0 1 𝑑 2 𝑑 3 𝑑 4 𝑑 5 𝑑 + 𝛼 |Δ𝐹𝑇𝑃 | + 𝛼 |Δ𝑃𝑆𝐹𝐸 | × 𝐶𝑜𝑣𝑖𝑑 + 𝛼 |Δ𝐹𝑆𝐴𝐿 | × 𝐶𝑜𝑣𝑖𝑑 6 𝑑 7 𝑑 8 𝑑 (3) | | | | | | + 𝛼 Δ𝐹𝐶𝑃𝑆 × 𝐶𝑜𝑣𝑖𝑑 + 𝛼 Δ𝐹 𝐷𝑃𝑆 × 𝐶𝑜𝑣𝑖𝑑 + 𝛼 Δ𝐹 𝐶𝑅𝐸 × 𝐶𝑜𝑣𝑖𝑑 9 𝑑 10 𝑑 11 𝑑 | | + 𝛼 Δ𝐹𝑇𝑃 × 𝐶𝑜𝑣𝑖𝑑 + 𝑟𝑚𝐹𝑖 /𝑒𝑎𝑟𝑌 /𝑒𝑟𝑄𝑢𝑎𝑟𝑡 𝑠𝑡𝑒𝑐𝑓𝑓𝑒 + 𝜉 . 12 𝑑 We omit other subscripts for brevity. We calculate abnormal returns for ACAR using the Carhart (1997) model as the normal return benchmark using daily data over 150 trading days before the forecast announcement and allowing a 15-day gap between the forecast announcement date and the end of the estimation period. ΔFEPS is the EPS forecast revision, ΔFSAL is the revenue forecast revision, ΔFCPS the cash flow forecast revision, ΔFDPS is the dividend forecast revision, ΔFREC the stock recommendation revisions, and ΔFTP the target price revision. Because we are interested in magnitudes of price reaction rather than direction, we use absolute values of the revisions. To capture incremental price effects during the pandemic, we interact the revisions with an indicator variable, Covid, that takes a value of one during 2020 and zero otherwise. Similar to earlier research, As we control for year effects, we do not include Covid variable in the regression. The results are the same when we define Covid starting from March 2020. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed e.g. Keung (2010), we assume a zero revision for a forecast not revised jointly with the earnings estimate on day d. The regression controls for firm-, year- and quarte-fixed effects and 𝜉 is the error term. To avoid confounding effects, we exclude a three-day window centered on the quarterly earnings announcements. Panel A of Table 3 reports descriptive statistics for equation (3). The mean absolute price reaction to analyst forecast revisions is 0.5% in 2018 and 0.51% in 2019, but 0.6% in 2020, an 18% increase (t-test for the difference in the average ACARs during and before the pandemic is 10.87). This preliminary evidence suggests stronger investor reactions to analyst forecasts issued during the pandemic, despite significantly lower accuracy of these forecasts. Compared to pre-pandemic, we observe significant revisions in all analyst forecasts during the pandemic ranging between 9% for stock recommendations to 142% for sales revenue, a result consistent with Figure 3. Panel B reports Pearson correlations between the variables, which are generally of the expected sign. The magnitudes of correlations between revisions are of moderate strength, which suggests revisions in each measure are likely to contains a unique signal. [Table 3] Table 4 reports price reaction regression results for equation (3). Consistent with earlier research, we find positive associations between price reactions and revisions in analyst earnings forecasts, sales, target prices and stock recommendations before the pandemic (Gleason and Lee 2003, Womack 1996, Jegadeesh and Livnat 2006, Keung 2010, Bradshaw, Brown and Huang 2013). The evidence on insignificant price reactions to joint earnings and cash flow forecast revisions is consistent with Bilinski (2014). [Table 4] We observe incrementally higher price reactions to revisions in all forecasts during the pandemic, but for target prices. The economic magnitudes of higher investor reactions are This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed significant, for example, price reactions to analyst EPS forecast revisions are 82% stronger during the pandemic compared to the pre-pandemic period. Price reactions are 42% higher for revenue forecast revisions, 2072% higher for cash flow forecast revisions, 559% stronger for dividend forecast revisions, 137% higher for stock recommendation changes and 22% lower for target price revisions. Thus, despite lower accuracy, analyst forecasts during the pandemic convey incremental signals to investors. In untabulated results, we perform several additional tests. First, we examine price reactions to analyst revisions in the early period of the pandemic, March to April 2020, compared to the latter period May-November 2020. Price reactions are on average incrementally higher during the early period of the pandemic consistent with higher investor demand for information in that period. Second, we augment equation (3) with control variables that include the book-to-market ratio to capture firm’s growth opportunities, the price-to-sales ratio as a measure of the relative valuation of a firm, debt-to-assets ratio to capture financial leverage, firm’s return on assets to capture profitability, R&D-to-sales to capture innovation, and advertising-to-sales to capture product visibility to investors. Including these controls reduces the sample size, but our conclusions remain unchanged. Further, the results are unchanged when we re-estimate equation (3) using (i) market-adjusted ACAR as the dependent variable, and (ii) we keep only analyst-firm pairs that are present in the sample over the entire period. 3.3 Cross-sectional variation in price reactions The evidence in Table 4 suggests analyst forecasts convey incrementally valuable signals to investors during the pandemic. To sharpen this analysis, we next examine when during the pandemic investors We calculate this as the sum of coefficients on|∆EPS|×Covid and|∆EPS| dividend by the latter. The price-to-sales ratio is more useful in valuation for loss-making firms than the price-to-earnings ratio. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed find analyst research particularly valuable. First, we propose that analyst signals are more useful in periods of increased information demand. To capture investor information search activity, we follow the approach from Da et al. (2011) and use the Google aggregate search frequency for information about COVID-19 and its impact on the stock market. We focus on Google, which accounts for close to 90% of internet searches in the U.S.. Specifically, we create a variable Google, which is the sum of weekly Google searchers for the terms “Covid19”, “Covid”, “Coronavirus”, “SP500” and “stock market” over the period January 2020 to December 2020. Each weekly google search term is returned scaled by the average search volume over the search period. Figure 4 presents the time- series distribution of the Google measure and it shows a clear spike in search activity at the start of the pandemic, March 2020, and later levelling of internet searches over the reminder of 2020. We then interact Google with revisions in analyst forecasts over the pandemic period. [Figure 4] The last columns of Table 4 report equation (3) results augmented with the interaction terms between analyst forecast revisions and the google search measure. We observe that all interaction terms are positive, which suggests that investors find analyst forecasts particularly useful when their information demand, as captured by their online search activity, is high. Several studies examine the role analysts play in discovering private information compared to their role in interpreting public information (e.g. Ivkovic and Jegadeesh, 2004, Asquith et al. 2005). Francis et al. (2002) and Frankel et al. (2006) report that both functions are important to investors. Chen et al. (2010) document that analyst information discovery role dominates in the weeks before firms announce their earnings and information interpretation is more important in the weeks after See https://gs.statcounter.com/search-engine-market-share/all/united-states-of-america We do not interact the Google measure with revisions during the pre-pandemic period as there are no searches for COVID-19 related terms during that period. Using only searches for “stock market” and “SP500” over the pre- pandemic period to build the Google measure and then interacting it with pre-pandemic revisions has no impact on our conclusions. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed earnings announcements. We use this insight to examine the weight investors attach to these two roles. Specifically, we select earnings forecasts issued in a 10-day window before and after quarterly earnings announcements and create a variable Pre_EA, which takes a value one for analyst EPS forecast revisions issued in a 10-day period before quarterly earnings announcements and zero otherwise. We then interact this variable with revisions in analyst earnings forecasts and estimate the following model | | | | | | 𝑅𝐴𝐶𝐴 = 𝛽 + 𝛽 Δ𝑃𝑆𝐹𝐸 × 𝑃𝑟𝑒 _ + 𝛽 Δ𝑃𝑆𝐹𝐸 + 𝛽 Δ𝑃𝑆𝐹𝐸 × 𝑃𝑟𝑒 _ × 𝐶𝑜𝑣𝑖𝑑 𝑑 0 1 𝑑 2 𝑑 3 𝑑 + 𝛽 |Δ𝑃𝑆𝐹𝐸 | × 𝐶𝑜𝑣𝑖𝑑 + 𝛽 𝑟𝑒𝑃 _ + 𝛽 𝑒𝑃𝑟 _ × 𝐶𝑜𝑣𝑖𝑑 (4) 4 𝑑 5 6 + 𝑟𝑚𝐹𝑖 /𝑒𝑎𝑟𝑌 /𝑒𝑟𝑄𝑢𝑎𝑟𝑡 𝑠𝑒𝑡𝑒𝑐𝑓𝑓 + 𝑢 where 𝛽 and 𝛽 capture incremental price reactions to analyst earnings forecast revision in the short 1 3 window before earnings announcements, compared to revisions after earnings announcements, before and during the pandemic respectively. As with equation (3), we exclude EPS forecasts issued in a three-day window around earnings announcements to avoid the confounding effect of earnings announcements. Table 5 results confirm the evidence in Francis et al. (2002) and Frankel et al. (2006) that, in the pre-pandemic period, both analyst interpretation and information discovery roles are important to investors. However, during the pandemic, investors attach more weight to analyst information discovery than interpretation functions. This result is consistent with higher investor information demand for new information that helps assess firm performance during the pandemic. 4. Conclusions This study examines how a tail-risk event, as captured by the COVID-19 pandemic, affected analysts research production and analyst information intermediation role in the market. We document that We focus on earnings forecasts as revisions in other estimates tend to be less frequent in the short window around earnings announcements, which leaves relatively few observations. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed 𝐸𝐴 𝐸𝐴 𝐸𝐴 𝐸𝐴 analysts increase their research activity in the initial months of the pandemic compared to similar months before the COVID-19 outbreak. Forecasts issued during the pandemic associate with significantly lower accuracy, however, investors react incrementally higher to revisions in analyst estimates compared to the pre-pandemic years. This effect is magnified in periods of increased information demand as captured by googles searches for coronavirus and stock market information. We attribute this result to increased investor demand for information that helps assess firm value induced by the COVID-19 outbreak. Further tests reveal that analyst private information discovery role is more important to investors during pandemic compared to the information intermediation role. Overall, the study adds important evidence to the debate on the usefulness of analysts as information intermediaries in capital markets. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed References Amiram, D., W. Landsman, E. Owens, and S. Stubben. 2018. How are analysts’ forecasts affected by high uncertainty? Journal of Business Finance and Accounting 45, 295-318. Anginer, D., A. Donmez, H. Seyhun, H., and R. Zhang. 2020. Global Economic Impact of COVID-19: Evidence from Insider Trades. Working paper Simon Fraser University. Appel, I., T. Gormley and D. Keim. 2016. Passive investors, not passive owners. Journal of Financial Economics 2016, vol. 121, issue 1, 111-141. Asquith, P., M. Mikhail and A. Au 2005. Information Content of Equity Analyst Reports. Journal of Financial Economics 75(2):245-282 Baker, S., N. Bloom, S. Davis, K. Kost, M. Sammon, and T. Viratyosin. 2020. The Unprecedented Stock Market Impact of Covid-19. NBER Working Paper No. w26945. Ben-Rephael, A, Z. Da, and R. Israelsen. 2017. It depends on where you search: Institutional investor attention and underreaction to news. Review of Financial Studies 30, 3009–3047 Bento, A., T. Nguyen, C. Wing, F. Lozano-Rojas, Y. Ahn, and K. Simon. 2020. Evidence from internet search data shows information-seeking responses to news of local COVID-19 cases. Proceedings of the National Academy of Sciences 117, 11220–11222. Bilinski P. and M. Bradshaw. 2020. Analyst Dividend forecasts and their usefulness to investors. Working paper, Boston College. Bilinski, P. 2014. Do analysts disclose cash flow forecasts with earnings estimates when earnings quality is low? Journal of Business Finance and Accounting 41, 401–434 Bradshaw, M., L. Brown and K. Huang. 2013. Do sell-side analysts exhibit differential target price forecasting ability? Review of Accounting Studies 18, 930–955(2013 Carhart, M. 1997. On Persistence in Mutual Fund Performance. Journal of Finance 52, 57-82. Cejnek, G., O. Randl,and J. Zechner. 2020. The COVID-19 Pandemic and Corporate Dividend Policy. Working paper ZZ Vermögensverwaltung. Chen, X., Q. Cheng and K. Lo. 2010. On the relationship between analyst reports and corporate disclosures: Exploring the roles of information discovery and interpretation. Journal of Accounting and Economics 49, 206–226. Costola, M., M. Iacopini, and C. Santagiustina. 2020. Google search volumes and the financial markets during the COVID-19 outbreak. Finance Research Letters, forthcoming. Cox, J., D. Greenwald, and S. Ludvigson 2020. What Explains the Covid-19 Stock Market? NBER Working Paper No. w27784 Da, Z., J. Engelbberg, and P. Gao. 2011. In search of attention. Journal of Finance 66, 1461-1499 This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Das, S., C. Levine, and K. Sivaramakrishnan. 1998. Earnings predictability and bias in analysts' earnings forecasts. The Accounting Review 73, 277–294. DeFond, M., and M. Hung. 2003. An empirical analysis of analysts’ cash flow forecasts. Journal of Accounting and Economics 35, 73–100. Dempsey, S., 1989. Predisclosure information search incentives, analyst following, and earnings announcement price response. The Accounting Review 64 (4), 748–757 Ding, W., R. Levine, Ch. Lin, and W. Xie. 2020. Corporate Immunity to the COVID-19 Pandemic. Journal of Financial Economics, forthcoming Du, M. 2020. Locked-in at Home: Female Analysts' Attention at Work during the COVID-19 Pandemic. Working paper, University of Mannheim. Ertimur, Y., W. Mayew and S. Stubben. 2011. Analyst reputation and the issuance of disaggregated earnings forecasts to I/B/E/S. Review of Accounting Studies 16, 29–58 Fahlenbrach, R., K. Rageth, and R. Stulz. 2020. How Valuable is Financial Flexibility When Revenue Stops? Evidence from the COVID-19 Crisis. Fisher College of Business Fang, B., O-K. Hope, Z. Huang and R. Moldovan. 2020. The effects of MiFID II on sell-side analysts, buy-side analysts, and firms. Review of Accounting Studies 25, 855-902. Francis, J., Schipper, K., Vincent, L., 2002. Earnings announcements and competing information. Journal of Accounting and Economics 33 (3), 313–342. Frankel, R., Kothari, S.P., Weber, J.P., 2006. Determinants of the informativeness of analyst research. Journal of Accounting and Economics 41 (1–2), 29–54 Givoly, D., C. Hayn, and R. Lehavy. 2009. The quality of analysts’ cash flow forecasts. The Accounting Review 84, 1877–911. Gleason C. and Ch. Lee. 2003. Analyst Forecast Revisions and Market Price Discovery. The Accounting Review 78, 193-225 Graham, J., C. Harvey, and S. Rajgopal. 2005. The economic implications of corporate financial reporting. Journal of Accounting and Economics 40, 3-73. Grossman, S., and J. Stiglitz. 1980. On the impossibility of informationally efficient markets, American Economic Review 70, 393–408. Hann, R., M. Ognev and H. Sapriza. 2012. Forecasting the macroeconomy: Analysts versus economists. Working paper. Healy, P, and K. Palepu. 2001 Information asymmetry, corporate disclosure, and the capital markets: A review of the empirical disclosure literature. Journal of Accounting and Economics 31, 405-44 Hope, O. 2003. Disclosure practices, enforcement of accounting standards and analysts’ forecast accuracy: An international study. Journal of Accounting Research 41, 235–272. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Hugon, A., A. Kumar, and A. Lin. 2016. Analysts, macroeconomic news, and the benefit of active in-house economist. The Accounting Review 91, 513–534. Ivkovic, Z., Jegadeesh, N., 2004. The timing and value of forecast and recommendation revisions. Journal of Financial Economics 73 (3), 433–463. Jegadeesh, N., and J. Livnat. 2006. Revenue surprises and stock returns. Journal of Accounting and Economics 41, 147-171 Keung, E. 2010. Do supplementary sales forecasts increase the credibility of financial analysts’ earnings forecasts? The Accounting Review 85, 2047–74. Landier, A. and D. Thesmar. 2020. Earnings Expectations in the COVID Crisis. HEC Paris Research Paper No. FIN-2020-1377. Li, K., X. Liu, F. Mai, and T. Zhang. 2020. The Role of Corporate Culture in Bad Times: Evidence from the COVID-19 Pandemic. European Corporate Governance Institute – Finance Working Paper No. 726/2021. Loh, R., and R, Stulz. 2011. When Are Analyst Recommendation Changes Influential? The Review of Financial Studies 24, 593–627. Pastor, L., and P. Veronesi. 2012. Uncertainty about goverment policy and stock prices. Journal of Finance 67, 1219–1263 Ramelli, S. and A. Wagner. 2020. Feverish Stock Price Reactions to COVID-19. Review of Corporate Finance Studies, forthcoming. Richardson, S., S. Teoh, and P. Wysocki. 2004. The walk-down to beatable analyst forecasts: the role of equity issuance and insider-trading incentives. Contemporary Accounting Research 21, 885–924. Shores, D., 1990. The association between interim information and security returns surrounding earnings announcements. Journal of Accounting Research 28 (1), 164–181. Tkachenko, I. and B. Bataeva. 2020. Share repurchases, coronaeconomy and stakeholders’ interests. Working paper. Wang, V. and B. Xing 2020. Talk about the Coronavirus Pandemic: Initial Evidence from Corporate Disclosures. Working paper University of Waterloo Womack, K. 1996. Do Brokerage Analysts' Recommendations Have Investment Value? Journal of Finance 51, 137-167. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Appendix A. Raw data for Figure 1 EPS SAL CPS DPS pre- Pre- Pre- Pre- covid 2020 Covid 2020 Covid 2020 Covid 2020 Number Number Number Number Number Number Number Number of of % of of % of of % of of % forecasts forecasts difference forecasts forecasts difference forecasts forecasts difference forecasts forecasts difference January 13684 12601 -7.9% 8849 8467 -4.3% 2070 1829 -11.6% 846 937 10.8% February 15694 14837 -5.5% 11562 11199 -3.1% 2184 1653 -24.3% 944 858 -9.1% March 9117 15717 72.4% 6104 10957 79.5% 1265 2007 58.7% 409 454 11.0% April 15550 18616 19.7% 10433 13337 27.8% 2438 2333 -4.3% 743 744 0.2% May 16220 18385 13.4% 11479 13407 16.8% 2274 2098 -7.7% 485 735 51.5% June 5233 7544 44.2% 3514 5331 51.7% 718 789 9.9% 249 250 0.6% July 16002 17598 10.0% 11136 12436 11.7% 2445 2166 -11.4% 744 497 -33.2% August 15081 14082 -6.6% 10503 9986 -4.9% 2104 1345 -36.1% 435 295 -32.1% September 4858 4299 -11.5% 3146 2843 -9.6% 601 383 -36.2% 177 95 -46.2% October 17291 8068 -53.3% 11973 4887 -59.2% 2550 1146 -55.1% 645 149 -76.9% November 14329 379 -97.4% 10021 269 -97.3% 1810 38 -97.9% 428 7 -98.4% TP REC Pre-Covid 2020 Pre-Covid 2020 Number of Number of Number of Number of forecasts forecasts % difference forecasts forecasts % difference January 6664 5584 -16.2% 1013 1033 2.0% February 6964 6428 -7.7% 786 715 -9.0% March 3803 9669 154.3% 781 1469 88.2% April 6072 9418 55.1% 890 1172 31.8% May 6321 8443 33.6% 840 845 0.6% June 2420 3918 61.9% 852 888 4.2% July 6812 8419 23.6% 996 981 -1.5% August 6430 6870 6.8% 886 646 -27.0% September 2379 2070 -13.0% 715 688 -3.8% October 7478 3650 -51.2% 1127 616 -45.3% November 6191 163 -97.4% 852 32 -96.2% Notes: The table reports the monthly number of analyst earnings forecasts, revenue, cash flow, dividend, target price, stock recommendations issued be tween January 2018 and November 2020. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Figure 1 The monthly number of forecasts EPS forecasts Revenue forecasts 20,000 2018 2019 2020 15,000 15,000 10,000 10,000 5,000 5,000 2018 2019 2020 Cash flow forecasts Dividend forecasts 2018 2019 2020 2018 2019 2020 3,000 1,500 2,000 1,000 1,000 - - Target Prices Recommendations 15,000 2,000 2019 2020 2020 2018 2019 2020 1,500 10,000 1,000 5,000 The figure reports the monthly number of analyst earnings forecasts, revenue, cash flow, dividend, target price, stock recommendations issued between January 2018 and November 2020. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Figure 2 Quarterly forecast error Earnings forecast Cash flow forecasts 0.35 0.18 0.16 0.30 0.14 0.25 0.12 0.10 0.20 0.08 0.15 0.06 0.04 0.10 0.02 0.05 0.00 0.00 Median Mean Median Mean Revenue forecast Dividend forecasts 0.12 0.03 0.10 0.03 0.08 0.02 0.06 0.02 0.04 0.01 0.02 0.01 0.00 0.00 Median Mean Median Mean The figure reports the average quarterly percentage forecast error for analyst earnings forecasts, revenue, cash flow, dividend, target price, stock recommendations. Forecast error is calculated as absolute difference between the actual and forecasted values, scaled by 1 plus the absolute value of the actual. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Figure 3 Monthly average revisions in analyst forecasts Revisions in analyst forecasts: avearge of 2018 and 2019 0.30 0.25 0.20 0.15 0.10 0.05 0.00 -0.05 -0.10 -0.15 -0.20 January February March April May June July August September October November EPS SAL CPS DPS TP REC Revisions in analyst forecasts in 2020 0.20 0.10 0.00 -0.10 -0.20 -0.30 -0.40 EPS SAL CPS DPS TP REC -0.50 January February March April May June July August September October November The figure presents the monthly average revisions in analyst quarterly earnings-per-share forecasts (EPS), revenue forecasts (SAL), cash flow-per-share forecasts (CPS), dividend-per-share forecasts (DPS), target prices (TP) and stock recommendations. We calculate a revision as the difference between the analyst current and previous forecasts issued for the same fiscal quarter scaled by the absolute value of the previous forecast. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed Figure 4 Monthly google searches from January to December 2020 The figure reports the average monthly cumulative value of weekly Google searchers for the terms “Covid-19”, “Covid”, “Covid19”, “Coronavirus”, “SP500” and “stock market”. Each google weekly search term is scaled by the average search volume over the search period January 2020 to December 2020. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed TABLE 1 The annual distribution of analyst forecasts 2018 2019 2020 Total Earnings forecasts (EPS) 150122 146049 132126 428297 Revenue forecasts (SAL) 102430 101212 93119 296761 % of EPS forecasts 68.23% 69.30% 70.48% 69.3% Cash flow forecasts (CPS) 22443 19948 15787 58178 % of EPS forecasts 14.95% 13.66% 11.95% 13.6% Dividends forecasts (DPS) 6446 6107 5021 17574 % of EPS forecasts 4.29% 4.18% 3.80% 4.1% Target prices (TP) 65380 60680 64632 190692 % of EPS forecasts 43.55% 41.55% 48.92% 44.5% Stock recommendations (REC) 10781 9489 9085 29355 % of EPS forecasts 7.18% 6.50% 6.88% 6.9% Notes: The table reports the annual number of analyst individual quarterly earnings-per-share forecasts (EPS), revenue forecasts (SAL), cash flow-per-share forecasts (CPS), dividend-per-share forecasts (DPS), target prices (TP) and stock recommendations. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed TABLE 2 The average forecasts error before and during the COVID-19 pandemic Q1 Q2 Q3 Q4 Q4−Q1 t-test EPS I. Average pre-pandemic 0.082 0.078 0.078 0.088 0.005 -8.14 II. 2020 0.145 0.160 0.139 0.121 -0.025 10.87 (II−I)/absI 0.768 1.065 0.781 0.375 -5.526 -2321.09 t-test I − II -67.350 -94.240 -70.780 -20.890 I. Average pre-pandemic 0.063 0.058 0.057 0.061 -0.002 -2.43 SAL II. 2020 0.089 0.114 0.100 0.082 -0.007 -12.30 (II−I)/absI 0.419 0.951 0.769 0.344 -5.076 3333.90 t-test I − II -28.610 -58.780 -45.500 -12.340 I. Average pre-pandemic 0.263 0.245 0.213 0.208 -0.055 12.51 CPS II. 2020 0.280 0.307 0.265 0.215 -0.066 5.53 (II−I)/absI 0.065 0.253 0.247 0.031 -2.193 15.28 t-test I − II -2.830 -10.650 -8.970 -0.750 DPS I. Average pre-pandemic 0.017 0.016 0.014 0.020 0.003 2.04 II. 2020 0.022 0.026 0.015 0.016 -0.006 -0.97 (II−I)/absI 0.289 0.585 0.074 -0.232 -3.033 -444.44 t-test I − II -2.710 -4.660 -0.520 0.820 Notes: The table reports average forecast error for each quarter calculated for the pre-pandemic years 2018 and 2019, I. Average pre-pandemic, and the for the pandemic year 2020, II. 2020. (II−I)/absI reports the difference between the average quarterly forecast error during the pandemic scaled by the absolute value of the pre-pandemic error and t-test I − II reports the corresponding t-test. Column Q4−Q1 reports the difference in average forecasts for Q4 vs. Q1 and t-test reports the corresponding t-test. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed TABLE 3 Descriptive statistics for price reaction regression variables 2018 2019 2020 % COVID /pre-COVID Mean STD p Mean STD p Mean STD p Panel A: Descriptive statistics ACAR 0.050 0.055 0.000 0.051 0.057 0.000 0.060 0.061 0.000 18% |∆FEPS| 0.169 0.445 0.000 0.185 0.491 0.000 0.375 0.732 0.000 112% |∆FSAL| 0.025 0.056 0.000 0.026 0.057 0.000 0.061 0.101 0.000 142% |∆FCPS| 0.024 0.160 0.000 0.022 0.153 0.000 0.038 0.223 0.000 64% |∆FDPS| 0.002 0.033 0.000 0.001 0.027 0.000 0.003 0.049 0.000 106% |∆FREC| 0.011 0.107 0.000 0.012 0.121 0.000 0.013 0.115 0.000 9% |∆FTP| 0.064 0.097 0.000 0.064 0.101 0.000 0.101 0.141 0.000 57% ACAR |∆EPS| |∆SAL| |∆CPS| |∆DPS| |∆REC| Panel B: Pearson correlations coefficients |∆FEPS| 0.114 0.000 |∆FSAL| 0.110 0.366 0.000 0.000 |∆FCPS| 0.029 0.112 0.106 0.000 0.000 0.000 |∆FDPS| -0.003 0.023 0.037 0.002 0.037 0.000 0.000 0.303 |∆FREC| 0.031 0.006 0.005 0.008 -0.002 0.000 0.000 0.004 0.000 0.341 |∆FTP| 0.262 0.085 0.103 0.026 0.001 0.071 0.000 0.000 0.000 0.000 0.705 0.000 Notes: Panel A reports descriptive statistics for the price reaction regression variables in equation (3). ACAR is the absolute cumulative abnormal return estimated using the Carhart (1997) model as the expected return benchmark. ΔFEPS is the EPS forecast revision, ΔFSAL is the revenue forecast revision, ΔFCPS the cash flow forecast revision, ΔFDPS the dividend forecast revision, ΔFREC the stock recommendation revision, and ΔFTP the target price revision. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed TABLE 4 Price reaction regression result Estimate p Estimate p |∆EPS| 0.001 0.000 0.001 0.000 |∆SAL| 0.013 0.000 0.011 0.000 |∆CPS| 0.000 0.690 0.000 0.687 |∆DPS| 0.002 0.467 0.003 0.421 |∆REC| 0.004 0.000 0.004 0.000 |∆TP| 0.084 0.000 0.083 0.000 |∆EPS|*Covid 0.001 0.006 -0.002 0.000 |∆SAL|*Covid 0.005 0.024 -0.017 0.000 |∆CPS|*Covid 0.005 0.000 0.000 0.707 |∆DPS|*Covid 0.013 0.003 -0.005 0.443 |∆REC|*Covid 0.005 0.001 -0.002 0.423 |∆TP|*Covid -0.018 0.000 -0.055 0.000 |∆EPS|*Covid*Google 0.064 0.000 |∆SAL|*Covid*Google 0.115 0.018 |∆CPS|*Covid*Google 0.121 0.000 |∆DPS|*Covid*Google 0.000 0.000 |∆REC|*Covid*Google 0.124 0.001 |∆TP|*Covid*Google 0.341 0.000 No 0.0003 0.000 Google Year effects Yes Yes Yes Yes Quarter effect Firm effects Yes Yes 445572 442426 R2 27.93% 29.30% Notes: The table reports regression results for equation (3), which examines price reactions to analyst quarterly forecast revisions. Covid is an indicator variable equal to one for year 2020 and zero otherwise. Google is the sum of weekly Google searchers for the terms “Covid-19”, “Covid”, “Covid19”, “Coronavirus”, “SP500”, “stock market”. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed TABLE 5 Analyst information discovery vs. interpretation role Estimate p |∆EPS|*Pre_EA -0.001 0.293 |∆EPS| 0.002 0.000 |∆EPS|*Pre_EA*Covid 0.003 0.070 |∆EPS|*Covid -0.002 0.000 Pre_EA -0.025 0.000 PRE_EA*Covid 0.004 0.003 Year effects Yes Quarter effect Yes Firm effects Yes R2 32.99% Notes: The table reports regression results for equation (4) which examines price reactions to analyst quarterly forecast revisions before compared to after quarterly earnings announcements. Pre_EA equals one for analyst EPS forecast revisions in a 10-day period before earnings announcements and zero otherwise. Covid is an indicator variable equal to one for year 2020 and zero otherwise. This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3807974 Preprint not peer reviewed
ARN Conferences & Meetings – SSRN
Published: Jan 1, 2021
Keywords: COVID-19; Coronavirus; forecast accuracy; price reactions; information discovery; information intermediation
Access the full text.
Sign up today, get DeepDyve free for 14 days.