Geographic connections to China and insider trading at the start of the COVID-19 pandemic

Abstract

The sudden and exogenous nature of the COVID-19 crash provides a unique identification opportunity to study insiders’ informational advantages. We find that the sales of insiders at firms with connections to China were significantly more profitable during the COVID-19 crisis than the sales of insiders at firms without connections to China. Consistent with greater attentiveness to public information about the COVID-19 pandemic, this result is driven by China connected insiders executing larger (smaller) sales in the early (late) COVID-19 period than non–China connected insiders. We find our results are driven by trades that are not preplanned under Rule 10b5–1 and are consistent with anticipation of the systematic market effects of COVID-19 on an insider’s firm as opposed to firm-specific effects. Aggregate China connected insider trades also predict market returns during the COVID-19 period. Our study contributes to the insider trading literature by introducing geographic connection to market-wide information as a source of public information advantage and to regulatory efforts to investigate and understand corporate insider behavior related to the COVID-19 pandemic.

1 Introduction

The trades of corporate insiders often generate significant abnormal returns, attracting intense public and media scrutiny, allegations of criminal behavior, and the attention of regulatory bodies. Both the Securities and Exchange Commission (SEC) and the Department of Justice (DOJ) have signaled their intention to investigate potential illegal insider trades made during the COVID-19 pandemic.¹ Importantly though, only trades based on material nonpublic information are illegal under the Securities and Exchange Acts of 1933 and 1934, making the source of information upon which an insider’s trades are based a crucial aspect of any investigation. Most studies that focus on the source of information precipitating insider trades examine information that is private in nature, arising from an insider’s position within their own firm or from their social, industry, or political networks (e.g., Ahern 2017; Goergen et al. 2019; Jagolinzer et al. 2020). Much less, however, is known about insiders’ ability to process and profit from publicly available information ahead of market participants. We introduce geographic connection to the origin of an exogenous stock market crash as a potential public information advantage for insiders.

The COVID-19 pandemic, believed to have originated in Wuhan, China, in late 2019, precipitated a sudden and unexpected global stock market crash. Initial reports from China suggested that COVID-19 was unlikely to affect areas outside of Wuhan’s borders. Despite a growing number of confirmed cases and deaths in China in the early weeks of February 2020, reports of stabilized infection rates and optimism among US investors yielded record high closings of the NASDAQ Composite and the S&P 500 indices on February 19, 2020. Five days later, with announcements of COVID-19 outbreaks in South Korea, Italy, and Iran, the S&P 500 fell 3.3%. By mid-March 2020, COVID-19 was deemed a pandemic, and much of the world was placed on, or faced the prospect of, near total lockdown. The halt to global economic activity resulted in a 30% market decline, the fastest fall in worldwide financial markets in history.

We examine whether corporate insiders with a geographic connection to the pandemic’s early stages were better able to anticipate the effects of COVID-19 on their firm’s stock price. The COVID-19 outbreak was initially characterized by a lack of transparency from the Chinese government until the announcement, in late January 2020, that the virus could be passed between people. Wuhan and several other cities within the province were subsequently placed under immediate lockdown.² Although US financial markets reacted to the virus’s effects on global supply and demand chains, the reaction did not occur until roughly one month after Wuhan’s lockdown and public announcements about the virus’s existence.

We posit that insiders of firms with operations, production, or supply-chain activities in China (“China connected insiders”) were better able to incorporate and act upon public information related to the COVID-19 outbreak ahead of other equity market participants without geographic connection to the pandemic’s origins (“non–China connected insiders”). Our expectation is based on the view that China connected insiders were more attuned to the ramifications of COVID-19 on future firm performance and operational disruptions, given that they had relatively more China-specific knowledge and expertise and increased exposure to public information disseminated within China about the virus. Importantly, we do not suggest that China connected insiders possessed and illegally traded upon material nonpublic information about COVID-19. Rather, we posit that China connected insiders were more attentive to and better understood the implications of COVID-19’s spread due to the geographic connections of their firm to the initial outbreak.³

China connected insiders could observe the virus’s toll on municipal and corporate infrastructure, operations, and supply and demand before its eventual spread outside of China. Further, China connected insiders were also better positioned to understand the broader economic implications of the complete lockdown of Wuhan, a transportation hub so large and vital that its disruption significantly affected supply chains originating anywhere in China and ending in the US. The shutdown of Wuhan is analogous to a sudden and complete closure of FedEx and UPS domestic operations, which would yield unfathomable supply chain and economic disruptions both in the US and globally.⁴ We confirm this expectation by first documenting that COVID-related disclosures by China connected firms occurred much sooner in the pandemic than disclosures by non–China connected firms and that these disclosures precede media and public interest in COVID-19 by approximately one month.

We identify a firm’s China connection in three ways, two of which use text-based data developed in Hoberg and Moon (2017) to capture whether a firm purchases inputs from, or sells outputs to, China. We first consider an insider to be China connected if the Hoberg and Moon (2017) data indicates an above median number of mentions of output or input words in proximity to the word “China” in their firm’s 10-K. Second, we use the number of mentions of outputs to or inputs from China to capture the intensity of an insider’s geographic connection to China. Finally, we consider an insider to have a China connection if their firm discloses a segment in China.

We expect that China connected insiders’ awareness of the potential economic severity of COVID-19 began as early as mid-January 2020. To examine whether insiders profited from this awareness, we compare the 30-trading-day raw stock returns following insider sales and sales volume during the COVID-19 crisis to those in a pre-COVID-19 period for both China and non–China connected insiders.⁵ We focus only on sales in anticipation of the stock market crash because of its sudden, exogenous nature.⁶ During the pre-COVID-19 period, we find the pattern of returns around insider sales is similar for China and non–China connected insiders, with both exhibiting sales that follow price increases and precede periods of relatively flat returns. In the COVID-19 period, however, non–China connected insiders appear to sell after their firm’s share price begins to decline, while the sales of China connected insiders are better timed to precede larger forthcoming price declines.

In a multivariate difference-in-difference framework, we find robust evidence that sales made by China connected insiders during the COVID-19 period were three to five percentage points more profitable, as they avoided significantly larger price decreases in the 30 trading days following their sales. The economic significance of our result is large, as we estimate the mean profit per sale for China connected insiders increased from a $14,000 loss in the pre-COVID-19 period to a $251,000 profit in the COVID-19 period. By contrast, the mean profit per sale for non–China connected insiders increased from an $8,000 loss in the pre-COVID-19 period to a $160,000 profit. Thus, the profitability of China connected insider sales increased by approximately 58% more than that of non–China connected insiders.

We also examine the extent and timing of insider trading to understand the nature of insiders’ underlying selling behavior. We find that the overall change in the number and dollar volume of weekly insider sales at the firm level from the pre-COVID-19 period to the COVID-19 period is not statistically different for China and non–China connected insiders. However, when we partition the COVID-19 period into “early COVID-19” and “late COVID-19” periods, we find that China connected insiders executed a relatively lower (higher) number and dollar volume of sales in the late (early) COVID-19 period than non–China connected insiders. This finding is consistent with the incremental profitability of China connected insider sales being driven by an intertemporal shift of sales within the COVID-19 period, where China connected insider sales better preempt price declines than the sales by non–China connected insiders. In economic terms, we find that the China connected insiders in our sample avoided approximately $375 million in aggregate losses by selling stock before the initial stages of the COVID-19-induced market decline.

Insider sales often occur because the insider expects a decline in their own firm’s stock price resulting from any factor that may affect the stock price, whether firm-specific or macroeconomic in nature (Seyhun 1988; Brochet 2018). Our primary results, therefore, have two possible explanations. First, it is possible that COVID-19 had an idiosyncratic effect on China connected firms and that their insiders anticipated this effect. Alternatively, the trades of China insiders anticipated a systematic COVID-19 effect on their firm.⁷ We therefore compare the cumulative returns of firms with and without China connections and find they are nearly identical in the COVID-19 period, providing ex post evidence that COVID-19 did not differentially affect the stock prices of China connected firms. Next, we separate insiders’ raw 30-day buy-and-hold returns into their systematic (i.e., market returns) and idiosyncratic (i.e., market-adjusted returns) components and find that our main result is concentrated in the systematic component of returns. We also perform a market-level analysis where we regress future market returns on aggregate China connected insider sales and find that aggregate sales by China connected insiders are associated with negative market-wide stock performance two, three, and four weeks later. Our results are consistent with China connected insiders anticipating declines in their firm’s stock that were driven by the systematic effects of COVID-19 ahead of other market participants.⁸

In further analysis, we partition our sample on whether a trade was executed pursuant to an SEC Rule 10b5–1 trading plan, because these trades are preplanned and less likely to be due to an information advantage at the time of the trade. Consistent with strategic informed trading on the part of China connected insiders, we find that our profitability results are found only in the subset of non-10b5–1 trades. Finally, we confirm the robustness of our results to alternative specifications, such as matching on the 30-day pre-trade return of the insider’s firm, dropping sales at firms in the medical supply industry, examining purchases, controlling for recent earnings announcements, altering the length and dates of the COVID-19 and pre-COVID-19 periods, understanding whether insider trades precede or follow analysts’ revisions, and using month and year fixed effects.

Our study has four key contributions. First, we inform the SEC’s current efforts to investigate and understand corporate insider behavior during the COVID-19 pandemic. By documenting that geographic connection to the source of market volatility is one way in which corporate insiders gain an information advantage over other market participants, we provide important insight into how to focus regulatory resources when investigating insider trading. While geographic connections have previously been examined in academic literature as a source of information advantage for institutional investors (Baik et al. 2010; Kang et al. 2021), mutual funds (Coval and Moskowitz 2001), or retail investors (Chi and Shanthikumar 2017; Ivkovic and Weisbenner 2005), geography has not been documented as a source of an insider’s information advantage. Our results are consistent with China connected insiders anticipating declines in their firm’s stock that were driven by the systematic effects of COVID-19 ahead of other market participants, indicating that China connected insiders were not acting on inherently private information but were more attentive to and better able to process public information about the overall market effects of COVID-19.

Second, our finding that Rule 10b5–1 trading plans limited the ability of insiders to opportunistically profit from their trades should be of interest to the SEC as it works to revise the rules governing Rule 10b5–1 plans. The SEC is currently working to modify such plans based on extant research suggesting that some executives use Rule 10b5–1 plans to opportunistically time their insider trades (e.g., Larcker et al. 2021).⁹ This result should be of interest to regulators because, to our knowledge, we are one of the first studies to empirically document that Rule 10b5–1 trading plans are effective mechanisms to limit opportunistic insider trading in certain circumstances. However, we acknowledge that our conclusions may not generalize to trades pursuant to Rule 10b5–1 trading plans during normal market conditions.

Third, the idea that corporate insiders can attain macroeconomic information advantages is relatively scant in the insider trading literature. Only Seyhun (1988, 1992) and Brochet (2018) examine the possibility that insider trades can predict market returns. However, these studies examine normal conditions that lack significant, identifiable events that would create the conditions necessary for increased insider trades. We document that the trades made by geographically connected insiders can serve as a leading indicator of overall market returns. This result should be of use to economic forecasters and investors seeking to time market movements.

Finally, we contribute to the literature on the extent to which market participants can anticipate market crashes. Adebambo et al. (2015) and Fahlenbrach and Stulz (2011) examine the financial crisis of 2007–2008 and find no evidence that managers of financial firms anticipated it, despite their insider knowledge of the events affecting their industry. However, these studies only examine the potential for an insider to anticipate his or her own firm’s returns and not overall market returns. The COVID-19-induced stock market crash provides conditions that differ from the financial crisis along several dimensions. First, the stock market decline accompanied by the financial crisis was gradual, occurring over a much longer period of 18 months, making it difficult to identify a particular point in time at which insiders should have anticipated its effects ahead of other market participants. By contrast, the COVID-19-induced stock market crash occurred over a period of 1.5 months, the fastest fall into a bear market in history. Second, the stock market decline associated with the financial crisis was endogenous to the financial system because of its origin in the subprime mortgage crisis and the trading of financial derivatives, while COVID-19 represents an exogenous shock to the market, akin to that of a natural disaster. As a result, COVID-19 provides a much stronger identification opportunity to examine insiders’ ability to anticipate market-wide effects.

2 Insider trading, geographic connection, and COVID-19

COVID-19 provides an ideal setting in which to examine insider trading behavior both in response to publicly available information and ahead of market-wide declines, because there is a relatively short and clear period over which the information flow about COVID-19 out of China was restricted (providing an information advantage for China connected insiders). Thus, the events of COVID-19 lend themselves to the relatively straightforward identification of a short timeframe in which information was publicly available but not acutely in the frame of public consciousness. A detailed timeline of relevant dates in the COVID-19 period is provided in Appendix A.

Although infections first occurred in late 2019 and quickly spread beyond Wuhan, very little information was provided to, or reported by, US news sources during the early stages of the pandemic.¹⁰ Chinese institutions were ordered not to publish any information about COVID-19, and the official report from the Wuhan Health Commission, released in early January 2020, indicated that no new COVID-19 cases had been identified. By late January, Wuhan was placed on mandatory lockdown, suggesting that the severity of the disease had been significantly understated. Appendix A illustrates that nearly seven weeks passed between the earliest public indications of the virus in Wuhan and the point when formal action was taken by both the Chinese and US governments to contain its spread. During this period, the NASDAQ Composite and S&P 500 indices reached new highs. The stock market decline began only after more information about the virus became widely available in the US.

Our expectation that corporate insiders with geographic connection to the unfolding COVID-19 crisis in China were better able to anticipate its future stock price effects is based on two distinct lines of research: one examining insider trading and one focused on local informational advantages. It is well documented that insiders possess private information about their firm’s prospects and that they often use this information to earn abnormal returns on their trades (e.g., Aboody and Lev 2002; Ke et al. 2003; McVay et al. 2006). This private information can come from an insider’s social or sociopolitical connections. For example, Goergen et al. (2019) find that trades of directors with better networks have higher abnormal returns, and Jagolinzer et al. (2020) find that the political connections of board members at certain financial institutions provided an information advantage about distributions pursuant to the Troubled Asset Relief Program (TARP). However, only one study suggests that corporate insiders can also be more attentive to public information affecting their own firm’s performance and profit from it when other investors are inattentive. Alldredge and Cicero (2015) provide evidence that insiders of suppliers sell more of their own-firm stock when recent public disclosures made by their primary customers indicate poor future performance (i.e., when the supplier’s primary customers are likely to make fewer future purchases from them).

While most insider trading studies, including Alldredge and Cicero (2015), identify idiosyncratic firm information as the primary information advantage of corporate insiders, there is some evidence that insider trades can be made based upon broader information related to industry or economy-wide factors. Ben-David et al. (2019) study a proprietary dataset of trades in insiders’ personal portfolios and find that insiders are more likely to earn abnormal returns from trading the stock of firms in their own industry due to their greater familiarity with it, as opposed to having private, nonpublic information about these firms. Seyhun (1988, 1992) and Brochet (2018) introduce the possibility that insider trades can predict market returns, consistent with insiders trading based on the expected effect of market-wide events on their firm. However, other evidence suggests that corporate insiders are not well attuned to the effects of market-wide movements on their own firm’s stock. Hutton et al. (2012) find that analysts are better able to predict the effects of macroeconomic factors on firm earnings than a firm’s own managers. Further, Adebambo et al. (2015) and Fahlenbrach and Stulz (2011) find no evidence that managers of financial firms anticipated the financial crisis of 2007–2008 despite industry-specific insights.

We expect that insiders at firms with operational connections to China were more attuned to the potential effects of COVID-19 because of their firms’ operational and supply chain connections to the country and could profit from this greater attentiveness during the initial stages of the pandemic. Geographic proximity is often cited in the accounting and finance literatures as a source of information advantage (Chi and Shanthikumar 2017). Baik et al. (2010) examine the effect of local and non-local institutional investors’ stock trades on future stock returns and find that within-state proximity to their investments results in informed trading by institutional investors that predicts future stock returns. Several studies also show that individual investors earn higher returns when trading in firms local to them (e.g., Ivkovic and Weisbenner 2005; Berry and Gamble 2013), a result that also extends to mutual funds (Coval and Moskowitz 2001). However, there is no evidence that corporate insiders benefit from geographic connections outside their existence within social networks (Ahern 2017).¹¹

COVID-19’s spread was publicly reported in Chinese and US news sources from the beginning stages of the virus’s known existence at the time, so we do not assert that China connected insiders traded illegally on private information about their firm.¹² Instead, our expectation is similar to Alldredge and Cicero (2015) in that the insider trades we expect to observe are in response to publicly available information. Unlike Alldredge and Cicero (2015), we expect that geographic connections, as opposed to economic dependence, result in insider attentiveness to public information. Specifically, we expect that China connected insiders were more attentive to early, publicly available COVID-19-related information and events than non–China connected insiders due to their firsthand exposure to a municipal lockdown, the virus’s effects within their own firm’s operations and supply chains, and relatively more China-specific knowledge and expertise. We view the week that Wuhan was placed under quarantine as a key period for China connected insiders to gain insights ahead of non–China connected insiders. Even if COVID-19 did not spread beyond Wuhan, it is likely that the stock prices of US firms would be adversely affected (albeit, to a lesser extent), as nearly 21% of US imports come from China, and Wuhan serves as one of China’s primary transportation hubs.¹³

Figure 1 collectively indicates that China connected insiders were attentive to COVID-19 earlier than both the public and non–China connected insiders, as their disclosures about it pre-empted those of non–China connected insiders and large-scale media and public interest in the virus. Figure 1a shows that extensive newspaper coverage in the US does not occur until the end of February 2020. Similarly, a plot of the Google Trends index over time for the word “coronavirus” or “COVID” in Fig. 1b suggests that public consciousness was not acutely focused on the pandemic until mid to late March, corresponding with the rapid decline in the US stock market and the implementation of citywide and statewide lockdowns within the US. Figure 1c graphs the cumulative percentage of China connected and non–China connected firms in our sample with at least one SEC filing mentioning “coronavirus” or “COVID” and shows that some firms began to include discussions of COVID-19 in their disclosures as early as January 2020, well in advance of broader public awareness. This figure also shows that China connected firms disclosed COVID-19-related information earlier than non–China connected firms, consistent with our expectation of earlier awareness of COVID-19’s effects, and suggests that China connected insider sales were not necessarily the result of trading on undisclosed firm-specific information.¹⁴

Fig. 1

A Number of “Coronavirus” or “COVID” articles published in US newspapers. B Google trends index for “Coronavirus” or “COVID” over time. C: Percentage of firms with COVID-related disclosures. This figure illustrates the timing of both public and corporate consciousness of the COVID-19 pandemic during the first four months of 2020. a plots the number of US news articles containing the word “coronavirus” or “COVID” per ProQuest. b plots the Google Trends index for the word “coronavirus” or “COVID.” c plots the percentage of China and non-China firms with at least one filing on EDGAR mentioning “coronavirus” or “COVID”

In concurrent work, Anginer et al. (2020) use a machine learning technique that analyzes corporate insiders’ names and classifies them into ethnicities to determine whether insiders with cultural affiliation to east Asia profited from the COVID-19 pandemic. While Anginer et al. (2020) focus on personal and cultural connections as their primary theoretical construct, we are the first, to our knowledge, to document geographic connection as a source of insiders’ stock market information advantage. Anginer et al. (2020) also argue that cultural connections result in private information advantages for individuals assumed to have east Asian ethnicity. We do not argue that insiders gained a private information advantage in advance of the COVID-19 stock market crash, because information about the unfolding pandemic in China was public as early as January 2020. Instead, we posit that geographic connections made insiders of China connected firms more attentive to the already public information about the developing COVID-19 pandemic. Finally, Anginer et al. (2020) focus primarily on insider purchases after the COVID-19 stock market crash to gauge insiders’ expectations about the market’s recovery period following the crash. By contrast, we focus on the ability of insiders with China exposure to utilize their geographic information advantage to anticipate the COVID-19 stock market crash. Thus, our study takes an ex ante approach, while the Anginer et al. study takes an ex post approach to insider trading.

3 Profitability of insider sales

3.1 Sample and data

The SEC requires corporate insiders to file a Form 4 with information on their stock transactions within two business days of the transaction. Prior literature has found that Form 4 s contain useful information and are a significant component of investors’ information sets (e.g., Brochet 2010; Rogers et al. 2016, 2017). We collect data from SeekEdgar on insiders’ open market stock sales from Form 4 filings on EDGAR between January 2018 and April 2020. We require the insider to be either an officer or director and that the shares sold were directly owned by the insider (i.e., we exclude trades made by spouses or dependents and trades connected with options and other derivatives).

We focus on sales because COVID-19 precipitated a market decline. We require information on the date, transaction price, and the number of shares sold.¹⁵ We also require the firm to be included in Compustat, CRSP, and the Hoberg and Moon (2017) “coverage” file.¹⁶ Consistent with Shon and Veliotis (2013), we exclude firms in regulated industries and those incorporated outside of the US. Our final sample comprises 51,858 insider-trade observations from January 2018 through March 2020.¹⁷

We first determine whether a firm is China-connected and then consider all firm insiders to be China connected insiders.¹⁸ We proxy for geographic connection to China in three ways. The first two rely on the Hoberg and Moon (2017) database on offshoring activities, which employs textual analysis to identify mentions of a specific country name and any abbreviations or variations of the country name within 25 words of “output” words (e.g., sell, sales, revenue, consumers, store, export) or “input” words (e.g., supplier, vendor, import, subsidiary, plant, venture, factory).¹⁹ The dataset contains the specific number of output mentions, input mentions, and, separately, input mentions with or without the ownership of assets in a given country in the firm’s 2017 10-K.

Our first measure of China connection is an insider firm’s “China Exposure,” which is equal to one if an insider’s firm has above median mentions of inputs from or outputs to China in their 2017 Form 10-K compared to all firms that mention China (the median number of mentions is equal to three in our sample) and zero otherwise.²⁰ Our second measure of China connection is an insider firm’s “China Intensity,” equal to the natural log of one plus the number of mentions of inputs from or outputs to China from the 2017 Form 10-K. We confirm that the China and non-China classification is sticky from year to year prior to 2017.²¹ A firm is considered non–China connected if Hoberg and Moon manually analyze the firm’s 10-K and find no mention of an input or output connection to China. Firms that are not covered in the Hoberg and Moon dataset are dropped from our analysis.

The Hoberg and Moon (2017) dataset allows us to identify China connected firms with significantly less error than the Compustat segment data, which is not well populated and is subject to significant reporting discretion. For example, a firm may report geographic segments that do not necessarily correspond to a specific country or report segments based on product lines as opposed to geography. The Hoberg and Moon data is also superior to Exhibit 21 data, which only requires the reporting of “major” subsidiaries. Major subsidiaries, however, are only one aspect of a firm’s operational presence in a country. For example, Walmart does not have a Chinese segment reported in the Compustat segment data, nor does it list a major subsidiary in China in Exhibit 21. However, Walmart’s website indicates it has been operating in China for over 20 years and has substantial purchasing activity from China manufacturers.²² As such, Walmart is categorized as a China firm using the Hoberg and Moon data. However, to help ensure that our results are not simply an artifact of the Hoberg and Moon data, our third measure, “China Segment,” is an indicator equal to one if an insider firm has a China or Asia segment reported in the Compustat segment database.

We begin the COVID-19 period on January 19, 2020. As reported in the timeline presented in Appendix A, Wuhan officials implemented screening measures for travelers leaving the city through passenger terminals on January 14, 2020, and the US began screening travelers arriving from Wuhan at three airports on January 17, 2020. We begin the COVID-19 period on the following Sunday (i.e., beginning of the following week) to align the COVID-19 period both for our profitability tests, which rely on insider-trade-day data, and for our trade volume tests (presented in Section 4), which rely on firm-week level data. As discussed above, we view the week beginning on January 19 as a key period for China connected insiders to gain insight ahead of non–China connected insiders because it includes the week of the Wuhan lockdown.

3.2 Descriptive statistics

In Table 1, we present descriptive statistics for our sample of insider sales. Approximately one-third of sales are made by China connected insiders using our China Exposure or China Segment variables. The mean insider sale experiences a profit of 0.3% over the following 30 days, and 48.6% of all insider sales have a positive 30-day profit. To help alleviate concerns that our results are due to differential stock price performance between China and non–China connected firms, we compare the stock price performance of China and non–China connected firms using our two indicator proxies for China connections during the COVID-19 period in Fig. 2. We find that the cumulative equal-weighted portfolio returns of China connected firms are nearly indistinguishable from those of non–China connected firms in the COVID-19 period, never diverging by more than 2.4%.

Table 1 Descriptive statistics – insider sale sample

Fig. 2

Portfolio returns for China and non-China connected firms. Figure 2 compares the performance of equal weighted stock portfolios of China connected and non–China connected firms in the COVID-19 period by plotting the weekly cumulative equal weighted portfolio returns during the first four months of 2020

Figure 3 suggests that the patterns of returns before and after China and non–China connected insider sales in the pre-COVID-19 period are almost identical for either definition of an insider’s China connection. Returns for firms of both insider groups exhibit a systematic run-up prior to a sale, followed by relatively stable prices, similar to the return patterns documented by Jagolinzer (2009). However, the return patterns diverge for China and non–China connected insiders in the COVID-19 period. Figure 3a and b both suggest that non–China connected insiders sell in response to an initial price decline and before a continued price drop (i.e., “selling on the way down”), while China connected insiders appear to sell earlier in the decline, preempting a more significant decline in stock price over the following 30 days. We find that the post-trade returns of China and non-China firms are nearly indistinguishable in the days immediately following a trade by a China connected insider. The returns following China and non-China insider sales only begin to deviate in the latter half of the time window, indicating that investors did not initially learn from and respond to China connected insider sales.²³

Fig. 3

a Returns around insider sales – China exposure. b Returns around insider sales – China segment. Figure 3 presents cumulative buy-and-hold returns of China and non-China firms pre and post insider sales. The insider sale date is at day 0 on the x-axis, and negative (positive) values on the x-axis represent dates prior to (following) the insider sale. Firm returns are equal to the cross-sectional mean of each insider trade day’s cumulative raw return relative to the sale transaction date

In Table 2, we present univariate evidence in a two-by-two matrix comparing differences in sale profitability and the percentage of sales that are profitable. Panel A presents the results for our China Exposure variable, while Panel B presents results for our China Segment variable. Profit is equal to the mean 30-day raw buy-and-hold return following all insider sales, multiplied by negative one so that a positive value indicates a profitable sale (in that it avoids a price decline). In the pre-COVID-19 period, we find no statistical difference between the profitability of China connected insider and non–China connected insider sales. In the COVID-19 period, the average profitability of China connected insider sales significantly exceeds that of non–China connected insider sales by 5.18 (2.51) percentage points when the China connection is measured as China Exposure (China Segment).

Table 2 Profitability for China connected and non–China connected sales

While the profitability of sales increased for both China and non–China connected insiders from the pre-COVID-19 to the COVID-19 period, the increase was more pronounced for China connected insiders. We find similar patterns when looking at the percentage of profitable sales, with approximately 46% of all insider sales being profitable in the pre-COVID-19 period. While all sales were more likely to be profitable in the COVID-19 period, the increase in profitability from the pre- to the COVID-19 period was significantly larger (between 7 and 13%) for China connected insiders. In dollar terms, our data indicates that the 1520 (1453) trades executed by China Exposure (China Segment) insiders during the COVID-19 period represent $2.8 ($2.0) billion in total trading volume, an average trade size of $1.8 ($1.4) million, and an average profit per trade of $251,000 ($258,000). The average incremental profit of China insider trades relative to non–China insider trades is equal to approximately $90,000. Thus, the volume of all insider trades and the profits earned by China connected insiders are economically significant, particularly when compared to the dollar value of insider trades and profitability surrounding other events documented in prior research. For example, the total trading volume of politically connected insiders during the TARP period equaled $324 million, and the average market adjusted profits earned by those insiders were $22,000 per trade (Jagolinzer et al. 2020). As we note, however, our profit figures are based on raw returns and, therefore, represent an upper bound of the economic significance of the profitability of insider trades during the COVID-19 period.

3.3 Multivariate analysis

We empirically test whether China connected insider sales are more profitable than non-China connected insider sales during COVID-19 in a multivariate framework as follows:

$$ {\displaystyle \begin{array}{c}{Profit}_{i,t}={\upalpha}_0+{\beta}_1{China}_{i,t}+{\beta}_2{COVID}_{i,t}+{\beta}_3 China\ast {COVID}_{i,t}+{\beta}_4{lnMVE}_{i,t}+{\beta}_5{MTB}_{i,t}+\\ {}{\beta}_6{ROA}_{i,t}+{\beta}_7 Ret\_{Qtr}_{i,t}+{\beta}_8{RecentEA}_{i,t}\ \gamma k+{\varepsilon}_{i,t+1}\end{array}} $$

(1)

Our dependent variable, Profit, equals the raw buy-and-hold return over the 30 trading days following the insider sale, multiplied by negative one. Our primary measure of profitability uses raw returns, as opposed to abnormal returns, because COVID-19’s effect on a firm’s return could be idiosyncratic or systematic in nature. Further, the use of market-adjusted returns is often based on the implicit assumption that the insider sells their firm’s stock for diversification purposes and reinvests the proceeds into the market, which is unlikely to be the case in the presence of a market-wide disruption. However, because it is possible that some insiders reinvested the proceeds from the sale of their own-firm stock into the market, the results of estimating Eq. 1 may be interpreted as an upper bound of the incremental profits earned by China connected insiders on sales of own-firm stock ahead of the COVID-19 stock market crash. In additional analyses we bifurcate a firm’s post-trade return into its market and market-adjusted components.

China is equal to one of our three China connection proxies (China Exposure, China Intensity, or China Segment), and COVID is equal to one for trading days on or after January 19, 2020. Our coefficient of interest is β₃ (China*COVID), and we expect it to be positive if China connected insiders profited from their geographic connections and attentiveness to the unfolding COVID-19 pandemic. We include, as control variables, the natural log of a firm’s market value of equity (LnMVE), market to book ratio (MTB) and return on assets (ROA), all measured as of the most recent quarter-end to the insider sale. We also include a firm’s prior three-month buy-and-hold return (Ret_Qtr_t), and we control for the existence of a recent earnings announcement (RecentEA) by including an indicator—equal to one if the firm released an earnings announcement in the two weeks preceding the sale—and industry fixed effects.²⁴ We include these variables to ensure that our results are driven by trades related to attentiveness to COVID-19-related information and not trades related to sales following positive firm-specific information.

In Table 3, we present the results of estimating Eq. 1. Across all specifications, we find that the profitability of insider sales is significantly larger in the COVID-19 period than in the pre-COVID-19 period. We do not find the pre-COVID-19 sales of China connected insiders to be more profitable than their non-China peers. However, when China is measured as one of our two indicator proxies, the estimated value of β₃ indicates that the increase in profitability of China connected insider sales from the pre-COVID-19 to the COVID-19 period is approximately three to five percentage points greater than the change in profitability of non–China connected insider sales. In economic terms, China connected insiders exhibit a 20%–40% relative increase in the profitability of their trades. Further, we find that a one standard deviation increase in China Intensity is associated with an approximately 2% higher profitability of sales in the COVID-19 period. Overall, the results indicate that, while all insider sales were more profitable in the COVID-19 period, the presence and intensity of geographic connections to China are associated with a statistically and economically larger profit.

Table 3 Profitability of insider sales for China and non–China connected insiders

4 Frequency of insider sales

4.1 Sample and data

The incrementally larger profitability of China connected insider sales is consistent with the anticipation of a COVID-19-precipitated decline in an insider firm’s share price. To earn these incremental profits, China connected insiders could either intertemporally shift regular sales to preempt an expected stock price decline or hold their trading behavior constant prior to the stock market decline but cancel planned sales scheduled for later in the pandemic. Either scenario would indicate an information advantage for China connected insiders relative to non–China connected insiders. In this section we provide context to our overall profitability result by testing the volume and timing of observed insider sales.

We utilize a sample of 208,964 firm-week observations subject to the same data constraints as our trade-level sample and examine two dependent variables that capture the extent of insider trading activity: the number of sales made by insiders in a week (#Sales), and the natural log of the dollar value of sales made in a week (Sales$Value). The frequency analysis we perform does not require post-sale returns, so our firm-week sample ends with the last full week of April 2020. Table 4 provides descriptive statistics for our firm-week sample. Similar to our sample of sales, we find that 27%–36% of observations are China connected firm-weeks.

Table 4 Descriptive statistics – firm-week sample

4.2 Multivariate analysis

In our multivariate analysis of trading frequency, we estimate Eq. 1 using #Sales and Ln(Sales$Value) as dependent variables to examine the change in frequency for China and non-China connected insiders from the pre-COVID-19 to the COVID-19 period. We estimate the sales volume tests separately when China is based on China Exposure, China Intensity, and China Segment. In addition, we augment Eq. 1 by evenly splitting the COVID-19 period into an “early” and a “late” period to examine the timing of insider sales. We define an indicator, EarlyCOVID, as equal to one for the seven weeks between January 19 and March 8, and another, LateCOVID, as equal to one for the seven weeks after March 8. In this specification, the baseline for each post-COVID-19 period is the pre-COVID-19 period. If China connected insiders intertemporally shifted sales in anticipation of a decline in their firm’s stock price, we will observe more insider sales in the early COVID-19 period (which preceded the beginnings of the stock market crash) and fewer sales in the late COVID-19 period (during and after the stock market crash).

In Table 5, we present the results from a Poisson regression when #Sales is the dependent variable and an OLS regression when ln(Sales$Value) is the dependent variable. Panel A compares sales volume for China connected and non–China connected insiders in the pre-COVID-19 period to the COVID-19 period, and Panel B compares sales volume in the early and late COVID-19 periods to the pre-COVID-19 period. In Panel A, we find that the change in frequency of China insider sales from the pre-COVID-19 period to the COVID-19 period is not statistically different than the change in non-China insider sales across the two periods (i.e., the coefficient on China*COVID is insignificant). This suggests that China connected insiders did not sell more, or less, on average, than non-China connected insiders throughout the entire COVID-19 period. Further, the insignificant coefficient on China*COVID indicates that the incremental profitability of China connected insider sales is not due to differences in the total volume or extent of sale activity during the COVID-19 period.

Table 5 Sales volume for China connected and non–China connected insiders

In Panel B, for both our China Exposure and China Segment proxies, we find evidence consistent with China connected insiders executing incrementally more sales and a larger increase in dollar volume than non–China connected insiders in the early COVID-19 period. We note that the sum of the EarlyCOVID coefficient and the China*EarlyCOVID coefficient is near zero. This suggests that the sales of China insiders did not economically differ from the sample mean, but that non–China insiders executed fewer sales in the early COVID-19 period. Further, for both our China Exposure and China Segment proxies, we find that China connected insiders executed incrementally fewer sales and a larger decrease in dollar volume than non–China connected insiders in the late COVID-19 period. Specifically, China connected firms experienced a 10%–15% incremental increase in dollar volume of sales in the early COVID-19 period, followed by a 7%–14% incremental decrease in dollar volume in the late COVID-19 period, compared to the sample mean.²⁵

This analysis provides some evidence that the higher profitability we observe for China insider sales in the COVID-19 period is at least partially driven by an earlier incorporation of information about the impending effects of COVID-19 and increased dollar sales ahead of stock price declines and is consistent with insiders’ geographic connections allowing them to intertemporally shift trades to profit from expected stock price declines. However, we interpret these results with caution, given that the sign of the estimated coefficient for China*EarlyCOVID, though directionally consistent across all specifications, is not consistently statistically significant when the China connection is measured using the intensity of an insider’s China connection.

5 Additional analyses

5.1 Firm-specific vs. economy-wide factors

Our primary focus is on whether China connected insiders’ sales of their own firm’s stock anticipated future stock price declines due to their geographic connection to China and their resulting increased attention to COVID-19 information. What is unclear from our primary tests is whether the insider trading behavior we document reflects an idiosyncratic effect of COVID-19 on China connected firms or market-wide effects that were not yet reflected in all firms’ share prices (Seyhun 1992). In other words, insiders could anticipate a decline in their firm’s share price due either to factors that affect only their firm or factors that are macroeconomic in nature and have broader effects.

We perform several additional analyses to better understand why China connected insiders’ trades predict future returns in the COVID-19 period. Specifically, we test whether the main results we document in Sections 3 and 4 are consistent with an idiosyncratic effect of COVID-19 on China connected firms or with insider sales in advance of a systematic effect of COVID-19 on all firms. Importantly, we do not suggest that it is possible to identify whether the insider knows the information is idiosyncratic or systematic in nature before trading. Rather, the analyses presented in this section are intended to identify whether the insider trading behavior we document is based on information not yet incorporated into the returns of only China connected firms or into the overall market return.²⁶ In addition to providing context to our primary results, this analysis allows us to directly contribute to the insider trading literature by examining whether insiders appear to anticipate a general stock market decline as opposed to idiosyncratic events occurring within their firm.

To formally test whether our results are consistent with market-wide versus firm-specific effects, we decompose Profit into Profit_Market, equal to the 30-day market return following an insider sale, and Profit_Firm, equal to the firm’s market-adjusted 30-day buy-and-hold return following the insider sale. We present the results of estimating Eq. 1 with Profit_Firm and Profit_Market as separate dependent variables for each of the three definitions of China in Table 6. In Columns 1 through 3, we find no evidence that China connected insiders were able to earn incremental market-adjusted abnormal returns in the COVID-19 period, as the estimated coefficient on China*COVID is insignificant in each specification. By contrast, we find that our overall profitability result is concentrated in the market-wide component of China connected insiders’ incremental returns, as the estimated coefficient on China*COVID is positive and significant in Columns 4 through 6 with Profit_Market as the dependent variable.

Table 6 Firm-specific vs. market-wide insider sale profitability

The finding that our profitability results are largely driven by the market component of returns and not by abnormal returns provides some evidence that the incremental profitability of China connected insiders is due to their earlier incorporation of information about COVID-19’s market-wide effects that was not currently reflected in their firm’s share price. However, given that a stock market crash is characterized by a high correlation among firm returns within the overall stock market, we cannot rule out that the results in this section are being driven by power issues related to the majority of raw returns capturing market-wide effects in this period. As a result, we also test whether aggregate China connected insider trading is associated with future market returns in the COVID-19 period.

Our market-level tests are similar to those in both Seyhun (1988) and Brochet (2018), who examine the information content of aggregate insider trades for monthly market returns. Because our COVID-19 period contains only 3.5 months, we perform our aggregate analysis at the weekly level. We compute aggregate China connected insider trades as the ratio of all China connected insider sales in a week to all insider sales occurring in the week (PctChina_Week_t).²⁷ We then examine whether aggregate China connected insider trades in a week are associated with weekly market returns up to four weeks ahead and, separately, market returns in the following month. We include an interaction between PctChina_Week and the COVID-19 period indicator (PctChina_Week_t*COVID). We also include, as control variables, contemporaneous weekly and prior monthly market returns. Our aggregate sample consists of 121 trading weeks.

The results of regressing future market returns on aggregate China connected insider trades are reported in Table 7. Panel A (B) reports the results when China connected insider sales are determined using China Exposure (China Segment). In each panel, we find that the estimated coefficient on PctChina_Week_t is insignificant for all future market return horizons, suggesting that aggregate China connected insider sales do not predict market returns in the pre-COVID-19 period. We also find that aggregate China connected insider sales during the COVID-19 period are negatively and significantly associated with future market returns for all future return horizons beyond one week when China connection is based on China Exposure and beyond two weeks when China connection is based on China Segment. This result is consistent with the aggregate sales of China connected insiders in the COVID-19 period reflecting information not yet incorporated into market prices and with the insider trading behavior of China connected insiders during the COVID-19 period preceding the macroeconomic effect of COVID-19 on their firm, as opposed to a differential idiosyncratic effect of COVID-19 on China connected firms.

Table 7 Aggregate China connected insider sales and future market returns

5.2 Rule 10b5–1 trading plans

SEC Rule 10b5–1 was enacted in October 2000 to reduce opportunistic trading and provide clarity with respect to whether trades are made based on material nonpublic information. The rule was intended to deter illegal insider trading by allowing insiders a safe harbor for preplanned trades ahead of when they might possess material nonpublic information (Jagolinzer 2009). Typically, trades pursuant to Rule 10b5–1 plans are passive in nature, as insiders must not exert subsequent influence over whether specific planned trades are executed. Trades pursuant to Rule 10b5–1 plans therefore represent trades entered into with some degree of advance notice (although there is no uniformly specified time requirement). However, Rule 10b5–1 trading plans have recently faced increased scrutiny due, in part, to suspicions about their abuse during the COVID-19 pandemic.²⁸

Because of how rapidly the COVID-19 crisis unfolded and the nature of Rule 10b5–1 trades, we do not expect trades executed pursuant to a Rule 10b5–1 plan to occur based on attentiveness to COVID-19 related information. As a result, we expect non-Rule 10b5–1 trades to be more representative of trades arising from a sudden and short-lived informational advantage. However, this expectation is not without tension, as Jagolinzer (2009) finds evidence that trades pursuant to Rule 10b5–1 plans are not, on average, solely due to uninformed trading. We use textual analysis to parse the footnotes of Form 4 to identify Rule 10b5–1 trading plans and partition our profitability analyses on whether trades were executed under a 10b5–1 trading plan.²⁹

The results of our analysis are provided in Table 8. We find that the profitability of trades made both pursuant to and not pursuant to a Rule 10b5–1 plan are not significantly more profitable for China connected insiders in the pre-COVID-19 period. We find that only non-preplanned sales made outside of Rule 10b5–1 trading plans are incrementally more profitable for China connected insiders in the COVID-19 period using all three measures of an insider’s China connection. We perform Chi-squared tests comparing the coefficients on China*COVID for Rule 10b5–1 trades and non-Rule 10b5–1 trades and find they are significantly different across all three China proxies. These results shed light on the efficacy of Rule 10b5–1 in deterring opportunistic trading by corporate insiders. While we believe the particular information advantage we study is related to public information, our evidence is broadly consistent with managers not manipulating Rule 10b5–1 trading plans to exploit short-lived information advantages.

Table 8 Profitability of preplanned rule 10b5–1 trades

5.3 Other robustness tests

A potential alternative explanation for our results is that the incremental profitability we document for China connected insiders in the COVID-19 period is due to differences in pre-sale price runups for China versus non–China connected firms. While our findings that China connected insider sales precede the macroeconomic effects of COVID-19 on their firm and predict market returns in aggregate help reduce this concern, we perform a matched sample analysis to further address this issue. In Table 9, we estimate our profitability analysis matching each insider sale at a China connected firm to an insider sale at a non–China connected firm with the closest 30-day pre-trade return.³⁰ In each matched sample specification, we find results consistent with our non-matched sample estimations. Specifically, β₃ indicates that the increase in profitability of China insider sales from the pre-COVID-19 to the COVID-19 period is between 1.7 and 3.5 percentage points greater than the change in profitability of non–China insider sales, and the results in Columns 3 through 6 indicate that this was driven by the overall market return and not idiosyncratic firm returns. In untabulated analysis, we find that our results examining the volume of insider sales are similarly unchanged using a matched sample.

Table 9 Profitability of insider sales – matched sample

We further address this concern by dropping insiders of firms in the medical supply industry.³¹ Manufacturers and suppliers of general medical supplies and protective equipment for medical use experienced a significant surge in demand for their products very early in the pandemic.³² Thus, firms in this industry likely experienced positive returns during the months captured by our COVID-19 indicator, and it is possible that insiders of these firms profited by selling shares in their own firm stock following announcements of good news related to positive demand shocks. However, our results are substantially unchanged when we drop insiders of an industry that experienced good news during the COVID-19 period.

Our primary arguments focus on China insiders’ anticipation of stock market declines because of their increased attentiveness to early COVID-related information. As a result, we expect and find that corporate insiders profited from executing sales in advance of the COVID-19 market crash. An alternative expectation would be that China connected insiders did not execute more, or even executed fewer, purchases based on their geographic information advantage. In untabulated analysis, we run Eq. 1 but instead examine returns following insider purchases; we find no evidence that China connected insiders earned significantly greater profits than non–China connected insiders on their purchases. Further, we also examine purchase volume and find no consistent evidence of China connected insiders having different purchase behavior. Our collective results therefore suggest that the information advantage of China connected insiders resulted in profitable sales in anticipation of the stock market decline.

We perform several untabulated additional analyses to ensure the robustness of our results. First, we use detailed information available in the Hoberg and Moon dataset to examine an insider’s geographic connection to China with increased granularity. Specifically, we separately identify three categories of insiders with geographic connections to China based on whether, according to the Hoberg and Moon methodology, their firm 1) sells output to China, 2) purchases inputs from China with ownership of assets within China, or 3) purchases inputs from China without ownership of assets within China. Ex ante, we expect the weakest geographic connections to exist for insiders of firms that purchase inputs from China but whose public financial statements do not indicate ownership of assets within China. A lack of physical presence in China or direct relationship with a customer indicates a lack of a “boots on the ground” presence in China, which is our primary mechanism for the attentiveness gained from a geographic connection. The connection to China for this subset of firms is likely weakest because their physical connection to China is low; these firms do not have productive assets in China, nor do their revenues depend on China buyers. We find the geographic connections of insiders of firms that either output or input to China are associated with significantly higher profitability in the COVID-19 period than insiders of firms without geographic connections to China. Consistent with our expectations, we find that insiders whose geographic connections arise only from inputs from China (without ownership of assets) do not earn abnormal profits on their sales in the COVID-19 period.

To examine whether insider profits are driven by trades made before or after analyst revisions, we re-estimate our main profit analysis on subsamples of firms with and without an analyst forecast revision in the prior week We find, in untabulated analysis, that the interaction between COVID-19 and China connections is only significant for the subsample of trades made in the week after an analyst revision. This result is consistent with insiders not front-running information that causes analysts to revise their forecasts but instead trading in response to public news not already incorporated into their firm’s stock price during the COVID-19 period. This result further corroborates our expectation that insiders did not trade on material, non-public information, but instead better understood the implications of already public COVID-19-related information.

We re-estimate all our results using year and month fixed effects and find that they are substantively unchanged, suggesting our results are not driven by the cyclicality of insider sales. We use a two-year pre-COVID-19 period in our primary analyses but find that our results are not sensitive to the use of a one-year or a three-year pre-COVID-19 period. Finally, we vary the dates used to define both the COVID-19 and the early/late COVID-19 periods by one week in either direction and find that our results are substantially unchanged.

6 Conclusion

Baker et al. (2020) empirically document that no disease outbreak has affected the stock market more than COVID-19. In the aftermath of such significant market swings, public consciousness and media analyses typically focus on identifying the ex post winners and losers, often in the context of guiding future trading and investing opportunities.³³ In contrast to this ex post attention, we find that the stock sales of China connected insiders were more profitable ex ante than those of non–China connected insiders during the COVID-19 pandemic and that China connected insiders executed their sales earlier in COVID-19 pandemic than non–China connected insiders. Our results suggest that China connected insiders were better able to pre-empt stock price declines precipitated by COVID-19 because of their greater awareness of, and attentiveness to, public information about the virus and its potential effects on the supply chains and operations of firms. Further, we find that this result is predominantly driven by trades occurring outside of Rule 10b5–1 trading plans.

The results of this study contribute to the insider trading literature, specifically studies identifying the type of information acted upon by insiders in legal insider trades and the sources of information used by insiders. We extend this literature by documenting a geographic component to insiders’ information advantage and by showing that the geographic connections of insiders to the early stages of catastrophic events may be used to glean insights about the events’ impending effects. Our study is the first, to our knowledge, to find results consistent with a broad sample of insiders acting upon their private insights into an upcoming market crash in a predictable manner.

We also contribute to concurrent research examining firm and market behavior during the COVID-19 pandemic. These studies collectively describe that while firms did not anticipate a pandemic-related risk while preparing their 2019 annual reports (Loughran and McDonald 2020), firms were proactive in providing information to investors about the pandemic in the first quarter of 2020 (Wang and Xing 2020a, 2020b), going so far as to articulate specific pandemic-related risks they faced as a result of COVID-19 in their corporate disclosures (Hassan et al. 2020). Our results show that corporate insiders were better positioned to profit from their insights into this publicly disclosed information.

Appendix A

COVID-19 timeline through declaration of US national emergency

1. Sources: BusinessInsider.comhttps://www.businessinsider.com/coronavirus-pandemic-timeline-history-major-events-2020-3, Axios.comhttps://www.axios.com/timeline-the-early-days-of-chinas-coronavirus-outbreak-and-cover-up-ee65211a-afb6-4641-97b8-353718a5faab.html, the New York Times https://www.nytimes.com/article/coronavirus-timeline.html, CNN https://www.cnn.com/2020/02/06/health/wuhan-coronavirus-timeline-fast-facts/index.html, National Review https://www.nationalreview.com/the-morning-jolt/chinas-devastating-lies/, Wikipedia, and The World Health Organization https://www.who.int/news-room/detail/08-04-2020-who-timeline%2D%2D-covid-19. Solid white boxes represent dates in the pre-COVID-19 period, while shaded boxes represent dates falling in the COVID-19 period