Analyst Earnings Forecast Optimism during the COVID-19 Pandemic: Evidence from China

Abstract

Analysts are important participants in the capital market, but there are relatively few studies on the impact of the COVID-19 outbreak on analysts’ behaviors. This article examines the impact of the COVID-19 outbreak on the analysts’ earnings forecast optimism by using a sample of visits to Chinese listed firms during 2019–2020. We find that the analysts’ earnings forecasts become less optimistic and show pessimism after the outbreak of COVID-19. This result is consistent with past research findings that major natural disasters lead to analysts’ forecasts pessimism. However, we also find that the earnings forecasts issued by analysts with on-site visits are more optimistic after the COVID-19 outbreak. The increase in optimism is associated with accounting information transparency, the proportion of tangible assets, and the revenue geographical concentration of the visited companies. Further analysis shows that higher optimism in visiting analysts’ earnings forecasts after the COVID-19 outbreak leads to a positive market response, suggesting that optimism in visiting analysts’ forecasts misleads the market’s resource allocation. We also find that the higher level of optimism in visiting analysts’ earnings forecasts disappeared after the COVID-19 outbreak was well controlled. Overall, our study enriches the study of the impact of COVID-19 on capital markets from the perspective of analysts’ forecast optimism. Investors in other countries should also be aware of the impact of similar phenomena.

1. Introduction

The outbreak of COVID-19 in 2020 is one of the largest pandemics in human history, which brings great uncertainty to global economic growth and business development, and also affects the operation of capital markets. However, there are relatively few studies on the impact of the COVID-19 outbreak on capital markets [1,2]. Therefore, our paper aims to fill this gap by exploring the impact of the COVID-19 outbreak on capital markets from the perspective of analysts’ earnings forecast optimism.

Analysts are important participants in the capital market and play an important role in information gathering, analysis, and transmission. There is no precedent for the impact of a global pandemic on a company’s performance before the COVID-19 outbreak. It is a worthwhile question to explore how analysts forecast firm performance in such a scenario and whether this leads to conservative behavior in analysts’ earnings forecasting. Psychology studies also show that individuals who focus too much on particular information can lead to misevaluate the decision weight of that information and lead to biased judgments [3,4]. Therefore, the question of whether the private information obtained by analysts through on-site visits is assigned a higher certainty premium in the face of the uncertainty brought by the pandemic, leading to the visiting analysts’ earnings forecasts optimism and influencing capital market pricing, is also a topic of interest in this paper.

In this paper, based on a sample of visits to Chinese listed firms during 2019–2020, we adopt a difference-in-differences (DID) research design to evaluate the effect to investigate whether the COVID-19 outbreak leads to analysts’ pessimism and ultimately affects the analysts’ earnings forecast optimism, and explore the moderating effect of private information produced by on-site visits on analysts’ earnings forecast optimism. The DID method is one of the most commonly used non-experimental methods for assessing the impact of exogenous events and can effectively capture the impact of exogenous shocks. The article finds that analysts’ earnings forecasts are less optimistic after the outbreak of COVID-19, showing a tendency of pessimism. Meanwhile, visiting analysts issue more optimistic earnings forecasts after the COVID-19 outbreak. A cross-sectional analysis of firm characteristics shows that lower accounting information transparency, higher proportion of tangible assets, and lower revenue geographic concentration of visited firms lead to higher levels of optimism in visiting analysts’ earnings forecasts after the COVID-19 outbreak. Further analysis shows that higher optimism in visiting analysts’ earnings forecasts after the COVID-19 outbreak leads to a positive market response, suggesting that visiting analysts’ optimism misleads the market’s resource allocation. We also find that the higher level of optimism of visiting analysts’ earnings forecasts disappears after the pandemic is well controlled in the year 2021.

We contribute to the related literature in the following ways. First, this paper enriches the study of the impact of COVID-19 on capital markets from the perspective of analysts’ forecast optimism. Previous studies mostly discuss the economic consequences of major natural disasters from the perspective of corporate finance [5,6]. In this paper, we discuss this topic from the perspective of sell-side analysts and find that the COVID-19 outbreak reduces the optimism of analysts’ earnings forecasts. Second, we contribute to the psychology and economics literature by presenting evidence on how heuristic bias affects analyst judgments after the COVID-19 outbreak. We find that private information obtained by visiting analysts is assigned a higher certainty premium due to the uncertainty associated with the pandemic, leading to a higher optimism in visiting analysts’ earnings forecasts. We argue that this phenomenon stems from analyst heuristic bias under uncertainty. Third, we also examine the effects of firm characteristics. We find that lower accounting information transparency, higher proportion of tangible assets, and lower revenue geographic concentration result in a higher degree of optimism visiting analysts’ earnings forecasts after the COVID-19 outbreak.

The rest of the paper is organized in the following manner. Section 2 details the literature review and hypothesis development. Section 3 outlines the description of data sources and research methods, Section 4 presents the empirical results, Section 5 presents further development and robustness checks, and Section 6 concludes the paper.

2. Literature Review and Hypothesis Development

Previous studies show that major natural disasters, such as pandemics, can lead to analysts’ earnings forecasts pessimism. This pessimism comes from both rational and irrational channels [2].

Regarding the rational channel, major natural disasters can lead to deterioration in a firm’s financial health, disruption of supply chains, shrinking market demand, and deteriorating financial performance [7]. Jordà et al. find that the economic sequelae of pandemics last for about 40 years, with asset returns decreasing over that period [7]. Using data from listed companies in 28 countries around the world, De Vito and Gomez find that outbreak of the COVID-19 caused a significant decline in sales, a sharp increase in current liabilities, and a tightening of corporate liquidity [8]. During pandemic outbreaks, employee absenteeism also rises, with a consequent decrease in firm labor productivity [9], and the firm’s human capital decreases with the increase in operating costs, which has a negative impact on firm value. This phenomenon is particularly significant for developing countries such as China. Loayza et al. argue that because of the lower level of capital and knowledge accumulation, developing countries are more vulnerable to natural disasters and have more serious negative consequences compared to developed countries [10]. At the beginning of the COVID-19 outbreak, the Chinese government took extremely severe pandemic control measures, which led to a large-scale suspension of operations and production of firms. In this scenario, firms have difficulties in capital turnover, credit risk increases, and business risks subsequently increase. Firms’ external environment affects firms’ willingness to take risky investments [11] and firms reduce future investments after the COVID-19 outbreak. Therefore, due to concerns that the company’s operations and future earnings will be affected by the pandemic, analysts will issue pessimistic earnings forecasts after the COVID-19 outbreak.

In terms of irrational channels, past studies show that major natural disasters can distract analysts, thereby affecting their forecasting ability. Dong and Heo provide evidence that analysts have limited attention when the region in which they live experiences flu epidemics. They believe this is due to distractions of the sickness of family members, relatives, and colleagues [12]. Cuculiza et al. argue that terrorist attacks affect the attention and emotions of local analysts, making analysts’ forecasts more pessimistic than market expectations [13]. Prior researchers also show that people pay excessive attention to events that cause serious loss. People reflect uncertainty and threats of such an event in their feelings of fear, dread, and anxiety, and overestimate the event’s influence on their decision-making process [14]. Individuals will overestimate the possible risks incurred and show pessimistic tendencies [2]. Therefore, we propose hypothesis 1:

Hypothesis 1 (H1):

Analysts will significantly exhibit pessimism for earnings forecasts of firms after the COVID-19 outbreak.

On-site visits refer to the visit of investors or financial market intermediaries to the target company. The on-site visits process often includes face-to-face meetings with company management, employees, and investors’ relations officers. Some visits may even involve entering factories and workshops to see the company’s production activities, operations, and production facilities [15]. Previous studies show that analysts can obtain soft information about the company through on-site visits [16], such as company culture, employee morale, company strategy, etc. Such information is often subjective and contextual, often depending on face-to-face interaction [17]. Previous studies also show that such soft information tends to have a higher explanatory effect on firms’ future performance [17]. Therefore, private information from on-site visits tends to have a higher degree of certainty for analysts.

Prospect theory suggests that decision makers’ judgments about the probability of events occurring are nonlinear; higher weights are assigned to the probability of tail events occurring [3]. Tversky and Kahneman found that people often make decisions based on heuristics with only a subset of available information [18]. After the COVID-19 outbreak, visiting analysts may rely too much on private information obtained from on-site visits and overweight it in earnings forecasts. Bordalo et al. also argue that individuals focusing too much on a particular signal can cause biased judgments and generate mispricing [4]. Therefore, we believe that the visiting analysts’ earnings forecasts are more optimistic after the outbreak of COVID-19. Accordingly, we propose hypothesis 2:

Hypothesis 2 (H2):

After the COVID-19 outbreak, visiting analysts’ earnings forecasts have a higher optimism level.

In the following, we further discuss the effects of accounting information transparency, the proportion of tangible assets, and the revenue geographical concentration.

When a company’s information is less transparent, investors face more difficulties in collecting and processing company-specific information [19]. Therefore, when common information is less transparent, analysts give more weight to private information [20]. A lower accounting information transparency implies lower transparency of the company’s common information. As a result, analysts will place more importance on private information obtained through on-site visits. Therefore, for companies with lower accounting information transparency, the optimism of visiting analysis earnings forecasts is higher after the COVID-19 outbreak. Accordingly, we propose hypothesis 3:

Hypothesis 3 (H3):

Ceteris paribus, after the COVID-19 outbreak, the lower the accounting information transparency of the visited firms, the higher the visiting analysts’ earnings forecast optimism.

Unlike other analyst activities, on-site visits provide the opportunity to observe a company’s production processes, operational assets, assembly lines, and employee morale. The effectiveness of on-site visits varies with the amount of information obtained from the visual cues from these observations. Chen et al. argue that firms with more tangible assets have more observable activities and assets [17]. Cheng argues that site visits are more useful for firms with more tangible assets [15]. Therefore, for visited firms with a higher proportion of tangible assets, analysts can obtain more information through on-site visits and will rely more on the private information obtained through on-site visits. Accordingly, we propose hypothesis 4:

Hypothesis 4 (H4):

Ceteris paribus, after the COVID-19 outbreak, the higher the proportion of tangible assets of the visited firms, the higher the visiting analysts’ earnings forecast optimism.

Within a short time after the COVID-19 outbreak, a total of 24 provinces activated the level 1 major public health emergency response. Due to the Chinese government’s severe control measures at the beginning of the outbreak, the extent of the impact of the outbreak varied greatly from region to region. Previous research studies show that companies can achieve risk reduction through geographic diversification [21,22]. If a company’s revenue is more widely distributed geographically, it is less likely to be impacted by the pandemic. Visiting analysts would believe that the COVID-19 outbreak has less impact on the visited firms with lower revenue geographic concentration and are more confident in the reliability of the private information obtained from on-site visits. Accordingly, we propose hypothesis 5:

Hypothesis 5 (H5):

Ceteris paribus, after the COVID-19 outbreak, the lower the revenue geographical concentration of the visited firms, the higher the visiting analysts’ earnings forecast optimism.

3. Research Design, Data Sources, and Variable Measurements

3.1. Data Sources and Sample Selection

In this paper, we select Chinese A-share listed companies from 2019 to 2020 as our research sample. The data of analyst on-site visits, analyst characteristics, analyst forecasts, and firm-related financial data are obtained from the CSMAR database (China Stock Market & Accounting Research Database). We remove ST companies, listed financial institutions, and companies with incomplete financial data. Finally, we obtain a pooled cross-sectional dataset which includes 45,152 analyst earnings forecast-firm-annual observations. Among them, 27,023 earnings forecasts are made in 2019 and 18,129 are made in 2020. Moreover, we perform 1% and 99% winsorized processing on all continuous variables to ensure that the results are not influenced by extreme values.

3.2. Variable Measurement and Model Set

We follow the existing literature [15,23] and built the following baseline OLS linear regression model to test hypothesis 1:

Optimism_i,j,t = δ₀ + δ₁COVID_i,j,t + ControlVariables>_i,j,t+ ε_i,j,t

(1)

where Optimism_i,j,t is the main explanatory variable of this paper. We follow the previous literature [23,24,25], and define the analyst forecast optimism as:

Optimism_i,j,t = (EPS_forecast_i,j,t – EPS_Actual_i,j,t)/ P_i,j,t-1

(2)

where EPS_forecast_i,j,t represents forecast earnings per share as issued by analysts and EPS_Actual_i,j,t represents actual earnings per share in the year t. P_i,j,t-1 is the closing price of the stock at the end of the year t-1. This indicator measures the optimism of analysts’ forecasts by comparing analysts’ earnings forecasts to a company’s actual earnings per share. A higher value of this indicator means that the analysts’ earnings forecast level is higher than the company’s actual earnings. Therefore, the higher the value, the more optimistic the analyst’s earnings forecast is.

In model (1), COVID_i,j,t is the explanatory variable. We follow the previous literature [26,27,28], and define COVID_i,j,t as a value of 1 if the analyst’s earnings forecast is made in 2020, and 0 if the analyst’s earnings forecast is made in 2019. If H1 is valid, we expect δ₁ to be negative and significant.

Based on model (1), we develop the following empirical model to test hypothesis 2:

Optimism_i,j,t = δ₀ + δ₁COVID_i,j,t + δ₂Visit_i,j,t + δ₃COVID_i,j,t × Visit _i,j,t + ControlVariables_i,j,t + ε_i,j,t

(3)

where Visit_i,j,t is whether the analyst conducts an on-site visit, which is defined as a value of 1 if the analyst conducts an on-site visit to the targeted firm 30 days before the issue of the earnings forecast, and 0 otherwise. If H2 is valid, we expect δ₃ to be negative and significant.

Furthermore, based on model (3), we examine the effect of different cross-sectional variations by grouping companies according to their characteristics. We develop the following empirical model to test hypotheses 3 through 5:

Optimism_i,j,t = δ₀ + δ₁COVID_i,j,t + δ₂Visit_i,j,t + δ₃Char_i,j,t + δ₄Char_i,j,t × COVID_i,j,t + δ₅Char_i,j,t × Visit_i,j,t + δ₆Char_i,j,t
× COVID_i,j,t × Visit_i,j,t + ControlVariables_i,j,t + ε_i,j,t

(4)

where Char_i,j,t refers to firms’ characteristics variables. This paper measures firms’ characteristics variables in three dimensions: accounting information transparency, the proportion of tangible assets, and revenue geographic concentration.

Drawing on the existing literature [29,30,31], we use the modified Jones model to calculate the accounting information transparency.

First, we estimated the following models each year using all firm-year observations in the same two-digit SIC code to measure discretionary accruals:

$$\frac{TA{C}_{i,t}}{T{A}_{i,t-1}}={\alpha }_1+{\alpha }_2\frac{1}{T{A}_{i,t-1}}+{\alpha }_3\frac{PP{E}_{i,t}}{T{A}_{i,t-1}}+{\alpha }_4\frac{\Delta SAL{E}_{i,t}-\Delta A{R}_{i,t}}{T{A}_{i,t-1}}+{\epsilon }_{i,t}$$

(5)

where, TAC_i,t is the difference between net income and operating cash flow for firm i in year t, TA _i,t-1 is total assets for firm i in year t-1, PPE _i,t is the total gross plant property and equipment for firm i in year t, ΔSALE _i,t is the change in sales for firm i in year t, and ΔAR _i,t is the change in accounts receivable for firm i in year t. Based on the regressions, we obtain the regression residuals ε _i,t, and take the absolute values to indicate the level of accounting information transparency of the firm. The higher this value, the less transparent the company’s accounting information is.

Based on this, we define Transparency_i,j,t as the value of 1 when the firm’s accounting information transparency level is higher than the industry average, and 0 otherwise. If hypothesis 3 is valid, we expect δ₆ to be positive and significant.

The proportion of tangible assets (Tangibility_i,j,t) is measured by taking 1 if the targeted firm’s total gross plant property and equipment to total assets ratio exceeds the industry median, and 0 otherwise. If hypothesis 4 is valid, we expect δ₆ to be positive and significant.

We measure revenue geographic concentration by taking the sum of the squares of revenue per region as a percentage of total revenue. A smaller value means a more diversified revenue geographic source. We define areaHHi_i,j,t as the value of 1 if the targeted firm’s revenue geographic concentration is lower than the industry median, and 0 otherwise. If hypothesis 5 is valid, we expect δ₆ to be positive and significant.

For the control variables, we follow the existing literature [15,23] and control for the basic characteristics of the analyst:

(1) Horizon_i,j,t, the length of time (in day) between the release of analysts’ forecasts and the release of the annual financial report, and take the natural logarithm.

(2) Star_i,j,t, taking 1 if this earnings forecast is issued by a star analyst, 0 otherwise.

(3) Team_i,j,t, the number of people on the team for this earnings forecast report.

(4) Forecast_i,j,t, the total number of analyst teams issuing earnings forecasts in year t, and take the natural logarithm.

(5) Follow_i,j,t, the total number of firms covered by this analyst in year t, and take the natural logarithm.

(6) Brokerstar_i,j,t, the total number of star analysts at the analyst’s brokerage firm, and take the natural logarithm.

(7) Brokersize_i,j,t, the number of analysts at the analyst’s brokerage firm in year t, and take the natural logarithm.

We also control for the firm characteristics that can affect the analysts’ earnings forecast optimism:

(8) Size_i,j,t, the total assets of the company in year t-1, and take the natural logarithm.

(9) BM_i,j,t, the company’s book-to-market ratio in year t.

(10) Growth_i,j,t, the growth rate of the company in year t.

(11) Ins_i,j,t, the percentage of shares held by institutional investors in year t

(12) Loss_i,j,t, whether the company incurred a loss in year t, if yes, take 1, and 0 otherwise.

(13) Return_i,j,t, accumulated stock returns of the company in year t.

(14) State_i,j,t, the ownership variable of the firm in year t, taking 1 if the ultimate controller is state-owned, and 0 otherwise.

We also include industry fixed effects in the regression model. The t values are based on standard errors adjusted for analyst level clustering.

4. Empirical Results

4.1. Descriptive Statistics of Main Variables

Descriptive statistics are shown in

Table 1. Descriptive statistics of main variables.

Variable	N	Mean	Std. Dev.	Q1	Median	Q3
Optimismi,j,t	45,152	0.007	0.022	−0.001	0.001	0.008
COVIDi,j,t	45,152	0.402	0.490	0	0	1
Visiti,j,t	45,152	0.018	0.133	0	0	0
Sizei,j,t	45,152	23.466	1.172	22.558	23.324	24.156
BM	45,152	0.507	0.478	0.189	0.339	0.652
Growthi,j,t	45,152	0.172	0.258	0.029	0.134	0.271
Insi,j,t	45,152	0.104	0.072	0.047	0.091	0.147
Lossi,j,t	45,152	0.029	0.167	0	0	0
Returni,j,t	45,152	0.503	0.601	0.094	0.363	0.726
Statei,j,t	45,152	0.295	0.456	0	0	1
Horizoni,j,t	45,152	5.148	0.847	4.942	5.407	5.793
Stari,j,t	45,152	0.192	0	0	0	1
Teami,j,t	45,152	1.743	1	1	2	2
Forecasti,j,t	45,152	2.885	0.889	2.303	2.996	3.497
Followi,j,t	45,152	2.394	0.719	1.946	2.485	2.890
Brokerstari,j,t	45,152	0.970	1.040	0.000	0.693	1.946
Brokersizei,j,t	45,152	4.017	0.590	3.784	4.190	4.407
Transpancyi,j,t	45,152	0.429	0.495	0	0	1
Tangibilityi,j,t	45,152	0.424	0.494	0	0	1
areaHHIi,j,t	45,152	0.834	0.372	1	1	1

. The mean value of Optimism_i,j,t is 0.007, the median value is 0.001, and the standard deviation is 0.022. Both the mean and median values are greater than 0. This is consistent with the notion that Chinese analysts are optimistically biased [25]. An amount of 40.2% of analyst earnings forecasts in the sample are issued after the COVID-19 outbreak. A total of 1.8% of these analysts’ earnings forecasts had on-site visits before issuing. For other variables, the mean value of Horizon_i,j,t is 5.148. A total of 19.2% of analysts’ forecasts in our sample are from star analysts.

4.2. Univariate Analysis

Firstly, we perform the Shapiro–Wilk W test to check the normality for Optimism_i,j,t. As we can see in Panel A of

Table 2. Univariate Analysis.

Panel A: Shapiro-Wilk W test for normality for Optimismi,j,t
Obs.	z	Prob > z
45,152	24.441	0.000
Panel B: Univariate Analysis of Analyst Optimism in Non-COVID vs. COVID Periods
(1)COVIDi,j,t =0	(2)COVIDi,j,t =1	Difference Tests t-Test (2)–(1)
0.009	0.004	19.282 ***
Panel C: Univariate Analysis of Analyst Optimism with non-on-site visits vs. on- site visits in Non-COVID Periods
(1)Visiti,j,t =0	(2)Visiti,j,t =1	Difference Tests t-Test (2)–(1)
0.009	0.006	2.628 ***
Panel D: Univariate Analysis of Analyst Optimism with non-on-site visits vs. on- site visits in COVID Periods
(1)Visiti,j,t =0	(2)Visiti,j,t =1	Difference Tests t-Test (2)–(1)
0.005	0.004	0.852

Notes: t-statistics are given in the parentheses and Z-statistics for Shapiro-Wilk W test. *** denotes the significance of two-tailed tests at the 1% significance level.

below, the p-value of the test is less than 0.001. Thus, the distribution of Optimism_i,j,t is normal.

Table 2 also reports the results of the univariate mean test. In Panel B of Table 2, we compare the means of Optimism_i,j,t for subsample of COVID_i,j,t =0 verse COVID_i,j,t =1, the mean of the COVID_i,j,t = 1 subsample is larger than those of COVID_i,j,t = 0 subsample. The results of the t-test for the means are significant at the 1% level, suggesting analyst optimism is greater before the COVID-19 period. In Panel C and D of Table 2, we compare the impact of visiting analysts’ earnings forecast optimism before and after COVID-19. In Panel C of Table 2, the mean of the Visit_i,j,t = 0 subsample is larger than those of Visit_i,j,t = 1 subsample. The results of the t-test for the means are significant at the 1% level, suggesting analyst optimism is greater when analysts do not conduct an on-site visit before issuing earnings forecasts. In Panel D of Table 2, there is no significant difference between the means of Optimism_i,j,t for subsample of Visit_i,j,t =0 verse Visit_i,j,t =1. This result suggests that the corrective effect of on-site visits on analysts’ forecasts optimism disappeared after COVID-19 occurred.

4.3. Analyst Earnings Forecast Optimism during the COVID-19 Pandemic

The previous analyses argue that the COVID-19 outbreak has a negative impact on the economy and also creates negative sentiment and uncertainty expectations. The literature on behavioral finance suggests that people will experience some endogenous changes in risk preferences and behavior after an extreme event, and on this basis, hypothesis 1 suggests that analysts will significantly exhibit pessimism for earnings forecasts of firms after the COVID-19 outbreak. We expected a significant negative correlation between the analyst optimism variable Optimism_i,j,t, and the post-epidemic variable COVID_i,j,t.

Table 3. Analyst earnings forecast optimism during the COVID-19 pandemic.

VARIABLES	Optimismi,j,t
COVIDi,j,t	−0.002 ***
	(−8.47)
Visiti,j,t	−0.004 ***
	(−3.97)
COVIDi,j,t×Visiti,j,t	0.003 ***
	(2.77)
Sizei,j,t	−0.001 ***
	(−5.58)
BMi,j,t	0.005 ***
	(9.12)
Growthi,j,t	−0.008 ***
	(−14.54)
Insi,j,t	−0.012 ***
	(−6.51)
Lossi,j,t	0.046 ***
	(22.92)
Returni,j,t	−0.004 ***
	(−16.87)
Statei,j,t	−0.002 ***
	(−7.54)
Horizoni,j,t	0.004 ***
	(28.21)
Stari,j,t	0.000
	(0.65)
Teami,j,t	−0.000 *
	(−1.76)
Forecasti,j,t	0.000
	(0.67)
Followi,j,t	−0.000
	(−0.43)
Brokerstari,j,t	−0.000
	(−0.06)
Brokersizei,j,t	0.001 ***
	(2.97)
Constant	−0.002
	(−0.73)
Industry	Yes
Observations	45,152
Adj.R2	0.257
Shapiro-Wilk W test of the normality of residuals	(23.032) ***

Notes: t-statistics are given in the parentheses and Z-statistics for Shapiro-Wilk W test, and standard errors are clustered by the analyst. * and *** denote the significance of two-tailed tests at the 10% and 1% level of significance, respectively.

shows the coefficients of the regression. The results show that the coefficient between Optimism_i,j,t and COVID_i,j,t is −0.002 and is significant at the 1% level. This result suggests that the overall optimism of analysts’ earnings forecasts decreases after the COVID-19 outbreak. In terms of economic significance, the mean value of Optimism_i,j,t before the COVID-19 outbreak is 0.009, and a decrease of 0.002 implies a 22.22% decrease in the optimism of analysts’ earnings forecasts, which implies that our results are statistically significant and economically significant. This result is consistent with the prediction of hypothesis 1. This result is also consistent with the previous research. Kong et al. argue that individual show pessimistic tendencies after the major natural disasters [2].

Hypothesis 2 argues that individuals focusing too much on a particular signal can bias their judgment. The weight of information with high certainty is magnified in the face of great uncertainty. Therefore, after the COVID-19 outbreak, visiting analysts’ earnings forecasts have a higher optimism level. We expect that there is a significant positive correlation between the analyst optimism variables Optimism_i,j,t and COVID_i,j,t×Visit_i,j,t. The results of the test are recorded in Table 3. The results show that the coefficient before COVID_i,j,t×Visit_i,j,t is 0.003 and is significant at the 1% level. This result suggests that, instead of serving the purpose of reducing analysts’ earnings forecast optimism, the on-site visits even led to higher optimism after the COVID-19 outbreak. This result is consistent with the prediction of hypothesis 2.

The regression results for the other variables show that firm size (Size_i,j,t), firm sales growth rate (Growth_i,j,t), and institutional investor ownership (Ins_i,j,t) are significantly and negatively correlated with Optimism_i,j,t, implying that larger firm size, higher growth rate, and higher institutional investor ownership curb analysts’ earnings forecast optimism. However, BM_i,j,t, Loss_i,j,t, and Horizon_i,j,t are significantly and positively correlated with Optimism_i,j,t. This suggests that a higher book-to-market ratio, the occurrence of a loss in the current year, or a longer gap until the release of the annual financial report causes an increase in analyst earnings optimism. We also perform the Shapiro–Wilk W test for normality of residuals. As we can see in the table below, the p-value of the test is less than 0.001.

Overall, the above results suggest that analysts’ earnings forecast optimism declined after the COVID-19 outbreak after controlling for other potential determinants. Meanwhile, we find that visiting analysts experience a larger increase in forecast optimism than no visiting analysts.

4.4. Cross-Sectional Analyses for H3 to H5

Cross-sectional analyses for H3 to H5 are shown in

Table 4. Cross-sectional analyses for H3 to H5.

	Optimismi,j,t
VARIABLES	(1) Chari,j,t = Transpancyi,j,t	(2) Chari,j,t = Tangibilityi,j,t	(3) Chari,j,t = areaHHIi,j,t
COVIDi,j,t	−0.001 ***	−0.002 ***	−0.003 ***
	(−3.37)	(−6.76)	(−5.73)
Visiti,j,t	−0.002 **	−0.002	0.002
	(−2.43)	(−1.34)	(0.58)
COVIDi,j,t × Visiti,j,t	0.001	0.001	−0.003
	(0.76)	(0.87)	(−0.98)
Chari,j,t	0.004 ***	−0.001 **	0.001 **
	(11.54)	(−2.13)	(2.54)
Chari,j,t × COVIDi,j,t	−0.003 ***	−0.001	0.000
	(−6.42)	(−1.27)	(0.65)
Chari,j,t × Visiti,j,t	−0.003	−0.005 **	−0.006 *
	(−1.40)	(−2.55)	(−1.94)
Chari,j,t × COVIDi,j,t × Visiti,j,t	0.005 **	0.004 *	0.007 **
	(2.02)	(1.82)	(2.08)
Sizei,j,t	−0.001 ***	−0.001 ***	−0.001 ***
	(−5.35)	(−5.84)	(−5.51)
BMi,j,t	0.006 ***	0.006 ***	0.005 ***
	(9.36)	(9.32)	(9.11)
Growthi,j,t	−0.008 ***	−0.008 ***	−0.008 ***
	(−15.37)	(−14.66)	(−14.59)
Insi,j,t	−0.011 ***	−0.011 ***	−0.012 ***
	(−6.28)	(−6.25)	(−6.62)
Lossi,j,t	0.046 ***	0.046 ***	0.046 ***
	(22.94)	(23.02)	(22.89)
Returni,j,t	−0.004 ***	−0.004 ***	−0.004 ***
	(−16.95)	(−16.28)	(−16.72)
Statei,j,t	−0.002 ***	−0.002 ***	−0.002 ***
	(−7.70)	(−7.50)	(−7.57)
Horizoni,j,t	0.004 ***	0.004 ***	0.004 ***
	(28.09)	(28.23)	(28.29)
Stari,j,t	0.000	0.000	0.000
	(0.65)	(0.65)	(0.64)
Teami,j,t	−0.000 *	−0.000 *	−0.000 *
	(−1.78)	(−1.79)	(−1.79)
Forecasti,j,t	0.000	0.000	0.000
	(0.68)	(0.75)	(0.64)
Followi,j,t	−0.000	−0.000	−0.000
	(−0.41)	(−0.53)	(−0.41)
Brokerstari,j,t	−0.000	−0.000	−0.000
	(−0.11)	(−0.08)	(−0.04)
Brokersizei,j,t	0.001 ***	0.001 ***	0.001 ***
	(2.89)	(2.91)	(2.96)
Constant	−0.004	−0.002	−0.004
	(−1.18)	(−0.56)	(−1.07)
Industry	Yes	Yes	Yes
Observations	45,152	45,152	45,152
Adj.R2	0.261	0.258	0.257
Shapiro-Wilk W test of the normality of residuals	(23.026) ***	(22.986) ***	(23.030) ***

Notes: t-statistics are given in the parentheses and Z-statistics for Shapiro–Wilk W test, and standard errors are clustered by analyst. *, ** and *** denote the significance of two-tailed tests at the 10%, 5%, and 1% level of significance, respectively.

.

The results of column (1) in Table 4 show the impact of accounting information transparency. The regression coefficient for Transpancy_i,j,t × COVID_i,j,t × Visit_i,j,t is 0.005, which is significant at the 5% level. This result suggests that when the targeted firm’s accounting information is less transparent, the visiting analysts will trust more in the information they see with their own eyes, thus resulting in a higher level of on-site visits analyst earnings forecast optimism after the COVID-19 outbreak. The above results are consistent with the expectations of hypothesis 3. This result is consistent with the findings of Zhao et al. They argue that transparency is negatively associated with analyst forecast optimism [32].

The results of column (2) in Table 4 show the impact of the proportion of tangible assets. The regression coefficient for Tangibility_i,j,t × COVID_i,j,t × Visit_i,j,t is 0.004, which is significant at the 10% level. This result suggests that for visited firms with a higher proportion of tangible assets, there is a higher degree of optimism in visiting analysts’ earnings forecasts after the COVID-19 outbreak. Cheng et al. argue that analysts on-site visit firms with more tangible assets would gain more information because these firms have more observable activities and assets [15]. Therefore, this information may be given greater decision weight, ultimately resulting in a higher level of on-site visits analysts’ earnings forecast optimism after the COVID-19 outbreak.

The empirical results of the effect of revenue geographic concentration are shown in column (3) of Table 4. The regression coefficient for areaHHI_i,j,t × COVID_i,j,t × Visit_i,j,t is 0.007, which is significant at the 5% level. This result suggests that the lower the revenue geographical concentration of the visited firms, the higher the visiting analysts’ earnings forecast optimism. This result is in line with the expectations of hypothesis 5. The more geographically diversity of the company’s revenue sources, the less the company’s business is affected by the pandemic. Thus, companies with low revenue geographic concentration have a higher certainty of post-COVID-19 performance, and information obtained from analysts’ on-site visits is given higher decision weighting. This result is also consistent with the findings of Chang et al. They argue that analysts routinely misjudge the earnings implications when firm’s revenue geographic concentration is low [33].

Overall, the above results suggest that, after the COVID-19 outbreak, the lower the accounting information transparency, the higher the proportion of tangible assets, and the lower the revenue geographical concentration of the visited firms lead to the higher the visiting analysts’ earnings forecast optimism.

5. Further Development and Robustness Checks

5.1. On-Site Visits Analyst Earnings Forecast Optimism and Market Reaction

We further analyze the economic consequences of the increase in the level of optimism of visiting analysts’ earnings forecasts after the COVID-19 outbreak. We use an event study approach to examine the short-term market reaction to the issuance of analyst earnings forecasts. The event study method examines the impact of an event by determining whether the stock price fluctuates when the event occurs and whether it generates “abnormal returns”. In this approach, we calculate each daily abnormal return by subtracting the daily stock return from the market return. Then the cumulative abnormal return (CAR_{[−i, j]}) is calculated with a distinct point of time before or after the event day, where the i days before and j days after the event day (in our case, the event day is the day analysts release their earnings forecast) are referred to as the event window, denoted as [−i, j].

The specific regression model is as follows:

CAR _{[−i, i]}= δ₀ + δ₁ Visit_i,j,t + δ₂ Optimism_i,j,t + δ₃Visit _i,j,t ×Optimism_i,j,t + ControlVariables_i,j,t+ ε_i,j,t

(6)

where CAR _{[−i, i]} is the abnormal returns. We use the analyst earnings forecast issuance date as the event date, select [0, 3] and [0, 5] as the event window, and use a market index-adjusted approach to estimate stock excess returns. Visit_i,j,t and Optimism_i,j,t are defined as the same as above. We divide the sample into pre- and post-COVID-19 groups and compared the differences of δ₃ across groups.

The regression results are shown in

Table 5. On-site visits analyst earnings forecast optimism and market reaction.

	CAR[0, 3]		CAR[0, 5]
VARIABLES	(1) Before COVID-19	(2) After COVID-19	(3) Before COVID-19	(4) After COVID-19
Visiti,j,t	0.001	0.002	0.000	0.005
	(0.35)	(0.58)	(0.16)	(0.99)
Optimismi,j,t	−0.064 ***	−0.094 ***	−0.113 ***	−0.112 ***
	(−3.77)	(−3.75)	(−5.83)	(−3.86)
Optimismi,j,t × Visiti,j,t	−0.008	0.460 **	0.014	0.955 ***
	(−0.10)	(1.98)	(0.17)	(3.04)
Sizei,j,t	−0.002 ***	−0.003 ***	−0.003 ***	−0.002 ***
	(−6.24)	(−5.21)	(−6.54)	(−3.15)
BMi,j,t	0.000	−0.001	−0.001	−0.002
	(0.40)	(−0.64)	(−0.48)	(−1.23)
Growthi,j,t	−0.001	0.000	−0.003	0.000
	(−0.44)	(0.22)	(−1.52)	(0.11)
Insi,j,t	0.004	0.013*	0.010	0.025 ***
	(0.67)	(1.79)	(1.58)	(2.99)
Lossi,j,t	0.001	−0.006*	0.002	−0.005
	(0.45)	(−1.82)	(0.71)	(−1.12)
Returni,j,t	0.017 ***	0.012 ***	0.022 ***	0.015 ***
	(16.57)	(11.69)	(18.23)	(12.36)
Statei,j,t	−0.000	0.000	0.001 *	0.001
	(−0.46)	(0.19)	(1.78)	(1.41)
Horizoni,j,t	0.000	0.001	0.001	0.002
	(0.40)	(0.50)	(0.79)	(1.41)
Stari,j,t	0.001	0.001	0.001	0.001
	(1.09)	(0.36)	(0.64)	(0.29)
Teami,j,t	−0.001	−0.001	−0.001 *	−0.000
	(−1.19)	(−1.34)	(−1.82)	(−0.47)
Forecasti,j,t	0.000	0.001	0.000	0.002
	(0.19)	(1.07)	(0.02)	(1.24)
Followi,j,t	−0.001	−0.002	−0.002	−0.003
	(−0.91)	(−1.04)	(−0.98)	(−1.31)
Brokerstari,j,t	0.000	−0.000	0.000	−0.000
	(0.33)	(−0.06)	(0.17)	(−0.57)
Brokersizei,j,t	0.001	0.002 *	0.001 *	0.002 **
	(1.37)	(1.95)	(1.80)	(1.97)
Constant	0.053 ***	0.057 ***	0.056 ***	0.029 *
	(5.38)	(4.16)	(4.94)	(1.74)
Industry	Yes	Yes	Yes	Yes
Observations	26,133	16,677	26,133	16,677
Adj.R2	0.034	0.032	0.043	0.036
Shapiro-Wilk W test of the normality of residuals	(17.487) ***	(17.559) ***	(17.064) ***	(17.280) ***
Optimismi,j,t × Visit Coeff. Difference (1)–(2)	−0.4684 *		−0.9401 ***
F value	(3.03)		(7.61)

Notes: t-statistics are given in the parentheses and F-statistics are given in the parentheses in the coeff. difference test and Z-statistics for Shapiro–Wilk W test, and standard errors are clustered by analyst. *, ** and *** denote the significance of two-tailed tests at the 10%, 5%, and 1% level of significance, respectively.

. The results show that the regression coefficients before Optimism_i,j,t × Visit_i,j,t are insignificant for both CAR_{[0, 3]} and CAR_{[0, 5]} before the COVID-19 outbreak. However, after COVID-19, the regression coefficients before Optimism_i,j,t × Visit_i,j,t are all significantly positive. The coefficient intergroup comparison test also indicates that there is a significant difference in the regression coefficients of Optimism_i,j,t × Visit_i,j,t pre-, and post-COVID-19. The results suggest that, after COVID-19, the optimistic bias of visiting analysts’ earnings forecasts mislead the market and results in a positive market reaction.

5.2. Extent of Pandemic Control and Analysts’ Earnings Forecast Optimism

The cumulative number of COVID-19 infections in China in 2020 is 87,071, but the number of infections in 2021 is only 15,243, significantly decreasing in both the number of infections involved and the areas infected compared to 2020. Therefore, the operating performance of Chinese companies in 2021 is minimally affected by the COVID-19 outbreak and the macro uncertainty is reduced. If the previous analysis is correct, we should observe an increase in the optimism of analysts’ earnings forecasts in 2021 and the disappearance of visiting analysts’ earnings forecasts optimism.

We perform the following test for this. We include analysts’ earnings forecast data in 2021 into our sample and design a new variable Year2021_i,j,t. Year2021_i,j,t is defined as 1 if the analysts’ earnings forecasts are issued in 2021, and 0 otherwise. We include Year2021_i,j,t as well as Year2021_i,j,t×Visit_i,j,t in model (1) and examine the coefficients of Year2021_i,j,t and Year2021_i,j,t×Visit_i,j,t.

The results are shown in

Table 6. Extent of pandemic control and analysts’ earnings forecast optimism.

VARIABLES	Optimismi,j,t
COVIDi,j,t	−0.002 ***
	(−8.27)
Year2021i,j,t	0.003 ***
	(8.16)
Visiti,j,t	−0.004 ***
	(−3.94)
COVIDi,j,t × Visiti,j,t	0.003 ***
	(2.70)
Year2021i,j,t × Visiti,j,t	−0.000
	(−0.13)
Sizei,j,t	−0.001 ***
	(−11.18)
BMi,j,t	0.006 ***
	(11.49)
Growthi,j,t	−0.008 ***
	(−19.05)
Insi,j,t	−0.009 ***
	(−6.00)
Lossi,j,t	0.046 ***
	(27.54)
Returni,j,t	−0.004 ***
	(−16.22)
Statei,j,t	−0.002 ***
	(−8.35)
Horizoni,j,t	0.003 ***
	(21.45)
Stari,j,t	0.000
	(0.03)
Teami,j,t	−0.000
	(−0.53)
Forecasti,j,t	0.000
	(0.93)
Followi,j,t	−0.000
	(−0.68)
Brokerstari,j,t	−0.000
	(−0.52)
Brokersizei,j,t	0.000
	(1.40)
Constant	0.011 ***
	(4.00)
Industry	Yes
Observations	62,551
Adj.R2	0.266
Shapiro-Wilk W test of the normality of residuals	(23.584) ***

Notes: t-statistics are given in the parentheses and Z-statistics for Shapiro–Wilk W test, and standard errors are clustered by analyst. *** denote the significance of two-tailed tests at the 1% level of significance, respectively.

. The results show that the coefficient of Year2021_i,j,t is 0.003, and is significant at the 1% level. Meanwhile, the coefficient of Year2021_i,j,t×Visit_i,j,t is −0.000 and insignificant. This indicates that there is no significant difference in the level of optimism of visiting analysts’ earnings forecasts in 2021 and before the COVID-19 outbreak. This result also confirms hypothesis 1 of this paper that the optimism of visiting analysts’ forecasts during the COVID-19 outbreak is due to the overweighting of deterministic information under the macro environment uncertainty.

5.3. Placebo Test

There is still a potential problem with the causal inference in this paper, assuming that the findings in this paper are correct and that the effect should not be observed at other time points. We perform a placebo test following the methodology of Bernile et al. [34]. In this test, we assume that the year in which the pandemic outbreak occurs is one year (i.e., we assume the COVID-19 outbreak in 2019) to four years (i.e., we assume the COVID-19 outbreak in 2016) earlier. We then test the coefficients of COVID_i,j,t × Visit_i,j,t.

The results are shown in

Table 7. Placebo test.

	COVID = 2019	COVID = 2018	COVID = 2017	COVID = 2016
VARIABLES	2017 < year < 2020	2016 < year < 2019	2015 < year < 2018	2014 < year < 2017
COVIDi,j,t	0.003 ***	−0.003 ***	0.005 ***	−0.007 ***
	(6.12)	(−7.76)	(13.77)	(−13.65)
Visiti,j,t	−0.002 **	0.001	−0.001	−0.000
	(−1.99)	(1.02)	(−0.83)	(−0.49)
COVIDi,j,t×Visiti,j,t	−0.002*	−0.002*	0.001	0.000
	(−1.67)	(−1.87)	(0.76)	(0.31)
Sizei,j,t	−0.001 ***	−0.001 ***	−0.001 ***	−0.000
	(−6.74)	(−3.82)	(−5.21)	(−0.59)
BMi,j,t	0.005 ***	0.004 ***	0.001 *	0.003 ***
	(8.92)	(7.35)	(1.85)	(3.47)
Growthi,j,t	−0.006 ***	−0.004 ***	−0.003 ***	−0.002 ***
	(−10.79)	(−7.99)	(−6.32)	(−5.36)
Insi,j,t	−0.011 ***	−0.001	−0.001	−0.010 ***
	(−5.74)	(−0.50)	(−0.69)	(−5.73)
Lossi,j,t	0.062 ***	0.067 ***	0.038 ***	0.032 ***
	(30.20)	(30.10)	(24.19)	(16.38)
Returni,j,t	−0.009 ***	−0.014 ***	−0.011 ***	−0.003 ***
	(−23.42)	(−27.46)	(−24.75)	(−10.96)
Statei,j,t	−0.003 ***	−0.003 ***	−0.002 ***	−0.003 ***
	(−7.94)	(−7.66)	(−6.75)	(−7.86)
Horizoni,j,t	0.006 ***	0.007 ***	0.005 ***	0.005 ***
	(38.57)	(38.29)	(32.61)	(33.28)
Stari,j,t	0.000	−0.000	0.000	0.000
	(0.37)	(−0.05)	(0.40)	(0.86)
Teami,j,t	−0.000	0.000	0.000	0.000
	(−0.26)	(0.86)	(0.66)	(1.48)
Forecasti,j,t	0.000	0.001	0.001	0.001
	(0.69)	(0.80)	(1.04)	(1.33)
Followi,j,t	−0.000	−0.000	−0.000	−0.001
	(−0.29)	(−0.32)	(−0.65)	(−1.06)
Brokerstari,j,t	0.000*	0.000	0.000	0.000
	(1.90)	(0.70)	(0.94)	(1.58)
Brokersizei,j,t	0.000	0.000	0.000	−0.000
	(0.42)	(0.33)	(0.43)	(−0.84)
Constant	−0.005	−0.014 ***	−0.001	−0.015 ***
	(−1.36)	(−2.89)	(−0.28)	(−2.90)
Industry	Yes	Yes	Yes	Yes
Observations	54,672	57,881	59,314	51,210
Adj.R2	0.303	0.260	0.167	0.185
Shapiro-Wilk W test ofthe normality of residuals	(25.902) ***	(25.843) ***	(26.367) ***	(26.777) ***

Notes: t-statistics are given in the parentheses and Z-statistics for Shapiro–Wilk W test, and standard errors are clustered by analyst. *, ** and *** denote the significance of two-tailed tests at the 10%, 5%, and 1% level of significance, respectively.

. The table shows that the coefficients before COVID_i,j,t × Visit_i,j,t are insignificant in all regressions. These results suggest that the conclusions of this paper are robust.

5.4. Results of Propensity Score Matching

Since the data we obtained are from an observational study and not from a randomized controlled experiment, randomized grouping is not used. The above regressions have the potential to produce systematic bias and cause potential endogeneity problems [35]. We use the Propensity Score Matching Method (PSM) to address the potential endogeneity problem [36,37]. First, we estimate the propensity score of whether the analyst conducts an on-site visit before issuing an earnings forecast. Then, for each visiting analyst’s earnings forecast, we find the earnings forecast with the closest propensity score, but without on-site visits before issuance, and treat it as a control group. After one-to-one matching, we end up with a total sample of 1,618. We regress model (1) and the results are shown in

Table 8. Results of propensity score matching.

VARIABLES	Optimismi,j,t
COVIDi,j,t	−0.004 **
	(−2.27)
Visiti,j,t	0.001
	(0.30)
COVIDi,j,t × Visiti,j,t	0.005 **
	(2.11)
Sizei,j,t	0.006 ***
	(7.92)
BMi,j,t	−0.002 ***
	(−2.96)
Growthi,j,t	0.010 ***
	(5.04)
Insi,j,t	−0.009 ***
	(−3.18)
Lossi,j,t	−0.005
	(−0.61)
Returni,j,t	0.039 ***
	(10.92)
Statei,j,t	−0.004 **
	(−2.39)
Horizoni,j,t	−0.002
	(−1.56)
Stari,j,t	0.003
	(1.50)
Teami,j,t	−0.001
	(−0.94)
Forecasti,j,t	0.001
	(0.37)
Followi,j,t	−0.000
	(−0.05)
Brokerstari,j,t	−0.000
	(−0.56)
Brokersizei,j,t	0.002*
	(1.69)
Constant	0.004
	(0.22)
Industry	Yes
Observations	1,618
Adj.R2	0.261
Shapiro-Wilk W test of the normality of residuals	(14.078) ***

Notes: t-statistics are given in the parentheses and Z-statistics for Shapiro–Wilk W test, and standard errors are clustered by analyst. *, ** and *** denote the significance of two-tailed tests at the 10%, 5%, and 1% level of significance, respectively.

, where the COVID_i,j,t × Visit_i,j,t coefficient is significantly positive and the main conclusion remains robust.

6. Conclusions and Discussion

6.1. Conclusion

The outbreak of COVID-19 in 2020 brings great uncertainty to global economy and affects the operation of capital markets. This article examines the impact of the COVID-19 outbreak on the analysts’ earnings forecast optimism by using a sample of visits to Chinese listed firms during 2019–2020. We find that post-COVID-19 analyst forecasts exhibit a pessimistic bias. This result is consistent with past research findings that major natural disasters lead to analysts’ forecasts pessimism. We also find that visiting analysts issue more optimistic earnings forecasts after the COVID-19 outbreak. Cross-sectional analyses show that lower accounting transparency, higher tangible asset ratios, and lower revenue source geographic concentrations of the visited firms lead to more optimistic on-site visits analyst earnings forecasts. Further analysis shows that higher optimism in visiting analysts’ forecasts after COVID-19 leads to a positive market reaction, suggesting that optimism in visiting analysts’ forecasts misleads the market’s resource allocation. When COVID-19 is well controlled, the phenomenon of visiting analysts’ earnings forecasts optimism disappears.

6.2. Policy Implication

This paper also provides insights for regulators and investors. Previous studies suggest that on-site visiting improves the accuracy of analysts’ forecasts, reduces the analysts’ forecast optimism, and increases the usefulness of analysts’ forecasts. However, our study finds that the earnings forecasts issued by analysts with on-site visits are more optimistic after the COVID-19 outbreak. The increase in optimism is associated with accounting information transparency, the proportion of tangible assets, and the revenue geographical concentration of the visited companies. Further analysis shows that higher optimism in visiting analysts’ earnings forecasts after the COVID-19 outbreak leads to a positive market response, suggesting that optimism in visiting analysts’ forecasts misleads the market’s resource allocation. The above results provide theoretical evidence on how investors can better understand the information content of visiting analysts’ forecasts after the COVID-19 outbreak. The results of this paper also provide insight into how regulators can guide visiting analysts to make more reasonable forecasts.

6.3. Limitations and Future Research

This paper has several limitations, which provide directions for future research. First, this paper did not examine the impact of the COVID-19 outbreak at the city level. Future research could consider examining the severity of the pandemic at the city level, and the impact of mobility restrictions on analysts’ forecasting behavior. Second, after the COVID-19 outbreak, many firms adopt virtual visiting as an alternative to on-site visits. Subsequent research could examine the effect of virtual visiting on analyst behavior. Finally, the paper discusses the moderating effect mainly at the level of the firm characteristics. Subsequent research could examine analysts’ gender, age, and other personal characteristics.