Political Incentives to Suppress Negative Information: Evidence from Chinese Listed Firms

ABSTRACT

This paper tests the proposition that politicians and their affiliated firms (i.e., firms operating in their province) temporarily suppress negative information in response to political incentives. We examine the stock price behavior of Chinese listed firms around two visible political events—meetings of the National Congress of the Chinese Communist Party and promotions of high‐level provincial politicians—that are expected to asymmetrically increase the costs of releasing bad news. The costs create an incentive for local politicians and their affiliated firms to temporarily restrict the flow of negative information about the companies. The result will be fewer stock price crashes for the affiliated firms during these event windows, followed by an increase in crashes after the event. Consistent with these predictions, we find that the affiliated firms experience a reduction (an increase) in negative stock return skewness before (after) the event. These effects are strongest in the three‐month period directly preceding the event, among firms that are more politically connected, and when the province is dominated by faction politics and cronyism. Additional tests document a significant reduction in published newspaper articles about affected firms in advance of these political events, suggestive of a link between our observed stock price behavior and temporary shifts in the listed firms’ information environment.

1. Introduction

In this paper, we examine the proposition that politicians and their affiliated firms (i.e., firms operating in the politician's province) suppress negative information in response to incentives created by political events that temporarily increase the cost of releasing bad news. Prior research argues that politicians and governments like to avoid negative news about their activities (e.g., Peltzman [1976], Watts and Zimmerman [1986]). To the extent that certain events asymmetrically increase the cost of releasing bad news, local politicians will have a heightened incentive to temporarily restrict the flow of negative information about firms under their control or influence. Similarly, self‐interested firms reliant on the support of local politicians will also respond to the incentives created by the political event. If these temporary incentives are sufficiently strong, the result will be a reduced flow of negative information in advance of the events, followed by an increase in the flow of negative information after the event as previously suppressed information is subsequently released. For publicly traded companies, the temporary suppression of negative information will generate a distinct shift in stock price crash behavior around the political event if the market is unable to fully unravel the suppression of bad news.

We test these arguments by examining the stock price behavior of Chinese listed firms around two highly anticipated political events—meetings of the National Congress of the Chinese Communist Party (CCP) and promotions of high‐level provincial politicians—over the period 1993 to 2011. We employ a stock price crash methodology specifically designed to capture the suppression and subsequent release of value‐relevant negative information (e.g., Chen, Hong, and Stein [2001], Jin and Myers [2006]). We focus on Chinese listed firms because their information environment is expected to be sensitive to political incentives for opacity. In China, government and party officials exert significant influence on listed firms through direct ownership arrangements and/or indirect channels (e.g., bureaucracy, regulation, licensing requirements, and informal political and social networks); this influence creates a conduit by which listed firms are expected to align their behavior with the preferences of local politicians. We focus on meetings of the National Congress and provincial‐level promotions as these two events temporarily increase the political costs of revealing bad news. From a career concerns perspective, local politicians incur a personal cost when firms under their control report poor performance, as adverse economic news limits their advancement within the party structure (e.g., Chen, Li, and Zhou [2005], Li and Zhou [2005]). More generally, authoritarian governments like China need to demonstrate strengths to stay in power (e.g., Tullock [1987]). The revelation of bad news about Chinese firms and their political supervisors reveals weaknesses, disrupts the government's narrative, and impacts social harmony. In both cases, the revelation of bad news is very costly to both local politicians and their affiliated firms during these event windows, generating a heightened incentive to suppress negative information. Finally, China's institutional arrangements, such as weak regulatory enforcement, limited private enforcement opportunities, concentrated ownership structures, and relationship‐based contracting, may be insufficient to support the necessary incentives for transparency during these periods.

However, there exist countervailing forces capable of sheltering listed firms from these transitory political pressures. First, the conversion of state enterprises to joint stock companies and their subsequent listing on stock exchanges expose state‐controlled firms to western governance mechanisms and market forces that demand transparency. Second, the Chinese government also set up the State Asset Supervision and Administration Commission to further decentralize control power from central and local governments to the listed firms. Qian [1996] and Fan, Wong, and Zhang [2013] argue that the Chinese government has established structural wedges between politicians and listed firms to credibly commit to not intervene in these firms’ day‐to‐day operations. Third, suppression of negative information is inherently difficult. Successful suppression requires politicians to have control over an array of information dissemination and production mechanisms that prevent self‐interested market participants and competing politicians from leaking negative information. Finally, the market could interpret silence during political periods as bad news; this unraveling would eliminate the expected political benefits of suppression. To the extent that these factors effectively mitigate the influence of local politicians’ incentives for opacity, we should not observe Chinese listed firms responding to transitory pressures to suppress bad news around political events. More generally, we expect firms most (least) exposed to strong market (political) forces, such as firms with weak political connections, non‐state‐owned firms, or firms operating in market‐oriented provinces or in provinces without powerful local politicians, to be less sensitive to transitory political incentives.

Consistent with the bad news suppression predictions, we find that affiliated listed firms are significantly less (more) likely to experience stock price crashes in advance of (after) our two political events. Mirroring the theoretical predictions of Chen, Hong, and Stein [2001], this intertemporal pattern is consistent with the initial suppression, and subsequent release, of adverse news in response to the heightened incentives of the local politicians and affiliated firms. These results are robust to alternative empirical specifications and various empirical measures of stock price crash behavior and information flow (including stock price synchronicity). Our evidence suggests that the incentives and ability of the politicians to suppress bad news outweigh the countervailing factors that promote transparency from decentralization of control of listed firms and the economic development of the provinces in China. The China market's inability to unravel the deferral of bad news is consistent with the evidence in Giglio and Shue [2014] indicating that a more developed U.S. market also underreacts to information contained in the absence of news. In additional analyses, we document the strongest reduction in stock price crashes in the three months preceding the political events, consistent with the heightened incentives to suppress bad news being most acute in the last few months leading up to the event. Further highlighting the role of political costs in this setting, our evidence shows that the incentive to suppress negative information varies based upon the level of political connections of the firms, and whether the local politicians belong to a political faction or are offspring of old CCP revolutionaries.

However, we fail to find evidence that privatization, capital market development activity across provinces, or the stronger regulatory and market environment associated with a Hong Kong listing mitigate political incentives to suppress bad news. Specifically, suppression effects are found in both state‐owned and non‐state‐owned firms and are unaffected by firm and provincial‐level proxies for capital market‐based pressures. Also contrary to expectations, we find that Hong Kong‐listed Chinese firms, arguably better regulated and subject to stronger market forces and more sophisticated investors, also experience a reduction in stock price crashes prior to both events. On the one hand, the robustness of the suppression effect across ownership types, capital market proxies, and markets where the firms are listed raises puzzles on how the government can orchestrate such extensive information control; on the other hand, this robustness highlights the power that Chinese politicians continue to wield over both the state and nonstate sectors despite China's state‐owned enterprise (SOE) reforms and the regulatory and market pressures found in both domestic and Hong Kong markets.

These results illustrate that complex forces are at work around these political events. Despite the absence of intervention from the government through direct ownership control, non‐state firms may voluntarily engage in bad news suppression in order to build the needed connections and trust with politicians that is enjoyed by their counterparts in the state sector. Although the pressures and channels to increase transparency greatly increase in more developed provinces and the Hong Kong market, there are also counterbalancing political forces to heighten the suppression incentives in these settings. Negative news about firms operating in more developed regions and/or specially chosen for a Hong Kong listing is likely to generate greater embarrassment to the politicians; this effect arises because these particular firms are visible and under greater scrutiny by a public expecting superior performance from them. Finally, notwithstanding our recognition of the difficulty in suppressing bad news among listed firms in China, we have documented evidence consistent with the conjecture that political incentives can influence newspaper reporting around these political events. More specifically, we document that the number of newspaper articles published about the listed firms in both official and financial newspapers temporarily decreases in advance of these political events, which is suggestive of the government's ability in controlling one channel of the listed firms’ information environment.

Our results highlight the important role that transitory political incentives play in shaping the information flow of listed firms in highly politicized environments. Despite the government's efforts to decentralize state‐controlled firms and develop the domestic stock market, Chinese listed firms are found to be sensitive to acute political incentives for opacity. Relative to the extant literature, our paper has several contributions. First, we examine a dimension of opacity—the suppression of negative information—that has not been explored in prior research yet maps strongly into the prevailing preferences of politicians. Second, we utilize an empirical methodology that captures both the suppression and the subsequent release of negative information; the documentation of this theoretically predicted intertemporal pattern is novel to the accounting literature and eliminates many of the alternative interpretations of the pre‐event period patterns alone. Third, through the joint use of intertemporal and cross‐sectional proxies for heightened political costs, we minimize misspecification concerns found in a time‐series or cross‐sectional analyses alone, thus strengthening the link between transitory political costs, information suppression incentives, and observed stock price behavior. Finally, by examining changes in a stock price–based measure of negative information flow—the negative skewness of excess returns (i.e., stock price crash statistic)—we capture information effects beyond those contained in accounting, audit choice, or disclosure data alone. By construction, this stock price measure captures the net flow of all value‐relevant, adverse information through all available information channels; as such, the measure reflects news from the full set of information production and dissemination mechanisms that the local politician must control to effectively suppress negative information. From a research design perspective, this approach allows us to determine whether aggregate changes in accounting, reporting, disclosure, and dissemination practices around these political events have a net impact on the overall information environment of listed firms (as captured by observable stock price behavior).

The documentation of a theoretically predicted association between political events and stock price behavior is an important first step in characterizing how the prevailing incentives of politicians and the agents can shape one aspect of corporate transparency. Although we are unable to document the exact mechanisms through which the suppression of negative information occurs, our tests have the ability to refute our information suppression hypotheses. Our results provide a useful starting point for future research seeking to identify the exact channels by which politicians and the listed firms temporarily suppress the net flow of negative information around politically sensitive events, such as a reduction in firm‐level disclosures, issuing biased financial reports, self‐censorship by the media, direct censorship by the politician, and the adoption of disclosure (e.g., bundling) and/or trading strategies to camouflage negative information, and how these channels of suppression actually work across ownership types and political connection levels of the firms, provinces where the firms operate, and the markets where the firms are listed. Given the increased opportunity for investors to purchase shares in both Chinese firms and other politically affiliated firms around the world, understanding the impact of short‐term political incentives on the underlying financial reporting and information environment of these listed companies is a first‐order concern for market participants.

2. Background and Motivation

2.1. POLITICAL INCENTIVES AND THE TEMPORARY SUPPRESSION OF NEGATIVE INFORMATION

Institutions associated with strong investor protections and economic outcomes are typically associated with more favorable financial reporting practices and better information environments (e.g., Ball, Kothari, and Robin [2000], Ball, Robin, and Wu [2003], Leuz, Nanda, and Wysocki [2003], Bushman, Piotroski, and Smith [2004]). Many developing economies, however, lack the institutional structure that creates incentives for good governance, high‐quality financial reporting practices, and transparent information environments. The presence of strong political forces is frequently a factor contributing to these weak institutional environments and outcomes.

Politicians, and the owners and managers of affiliated listed firms, face unique tradeoffs with respect to corporate transparency. Greater transparency benefits both the listed firm and the economy by improving the monitoring capabilities of both the state and outside investors, reducing information gathering costs, improving the efficiency of capital allocation and investment decisions, increasing foreign investment, lowering the firm's and country's cost of capital and, ultimately, increasing economic growth. However, because the incentives of local politicians and their affiliated firms are frequently not aligned with value or wealth maximization, transparency, especially with respect to the revelation of bad news, is costly. Transparency about state‐controlled entities highlights inefficiencies, political agendas, and economic disparities that exist within these firms (e.g., Shleifer and Vishny [1997], Rajan and Zingales [2003]).1 Similarly, transparency about both state‐controlled and non‐state‐owned firms under the local politician's supervision can bring unwanted scrutiny, especially with respect to policy failures, favoritism, and corruption, and limit the ability of politicians and executives to consume their private benefits of control by exposing weak governance practices, expropriation, and rent‐seeking behavior. The revelation of poor operating performance by affiliated firms also imposes personal costs upon local politicians and the firm's executives in the form of diminished reputational capital and damaged career prospects. Because of these costs, affiliated, self‐interested firms are expected to adopt information practices that minimize the expected political costs associated with reporting adverse news. More importantly, incentives to suppress negative information are expected to be temporarily heightened in advance of political events that incrementally increase the costs of revelation.

China offers a natural setting to examine how acute political incentives affect the incentives of listed firms to temporarily suppress bad news. In general, Chinese listed firms are associated with weak financial reporting practices (see Piotroski and Wong [2012] for a summary). The reporting outcomes are shaped by institutional arrangements (e.g., concentrated ownership, relationship‐based contracting, and weak investor protections) that minimize both the demand for and supply of high‐quality external reports. Despite these institutional arrangements, there are several reasons why Chinese listed firms may not respond to transitory incentives to suppress negative information. First, the decentralization of state‐owned companies, through government reforms and the stock exchange listing process, creates a structural wedge between politicians and local state‐controlled firms (e.g., Qian [1996], Fan, Wong, and Zhang [2013]). If effective, this operational autonomy, and resultant reliance on external (nonstate) financing and use of professional management, can potentially insulate state‐controlled firms from the day‐to‐day influence of local politicians. Second, the maturation of China's capital markets and emergence of non‐state‐owned enterprises have increased the relative importance of both market and reputational incentives; given that these benefits rely upon a credible commitment for transparency, they may act as a countervailing force against transitory political incentives. Robust capital markets are also associated with additional channels by which negative information could be leaked by self‐interested market participants or competing politicians. Third, sophisticated investors may unravel silence in advance of political events as bad news; this unraveling would eliminate the expected benefits of temporarily suppressing negative information. Finally, the suppression of negative news is inherently difficult, requiring politicians to directly or indirectly control a broad array of information channels and self‐interested individuals.

However, as outlined later, political forces can have a strong influence on the information environment of Chinese listed firms, especially with respect to the revelation of adverse information around major political events.2 First, local politicians incur a reputation cost when firms or provinces under their control report poor performance. Chen, Li, and Zhou [2005] and Li and Zhou [2005] find that provincial leaders’ promotions and demotions are significantly associated with the change in economic performance of their province. Shih, Adolph, and Liu [2012] argue that favored provincial politicians are strategically placed in growing regions prior to their elevation to a position of more power, thus bringing legitimacy to their political advancement. Regardless of the direction of causality between performance and political promotions, politicians and local firms have an incentive to suppress bad news when a key government promotion looms in the province.

Second, theories of authoritarian politics argue that, to maintain power, the government has to demonstrate strengths, for example, through economic power (Tullock [1987]). Releasing information that draws unwanted attention or public scrutiny to the firm, local politicians, or government is costly. Unexpected negative information also reveals weaknesses and results in a “loss of face” for both local politicians and the central government. In addition, Chinese culture values social conformity and the maintenance of relationships, with government policy during the latter half of our sample period explicitly stressing the importance of maintaining social harmony. The existence of these cultural norms and policies creates significant incentives for politicians and executives to not disrupt the government's narrative around visible political events. These norms and policies also create a disciplining mechanism by which informed individuals, such as rival politicians, do not voluntarily release negative information for personal gain.

Third, although the central and local government has decentralized much of the day‐to‐day operational decisions to the state‐controlled company managers, politicians can still exert control over the state‐controlled and non‐state‐owned companies through the application of regulation and other politicized channels (e.g., bureaucracy, licenses, capital raising activity). As a result, the personal incentives of central and local politicians are expected to influence the decisions of affiliated firms. Finally, with the limited development of China's legal system and investor protections and the absence of credible property rights protection, firms seek the protection and resultant benefits that accrue to maintaining strong relations with local politicians; this relationship‐based environment creates neither a demand for nor supply of timely information. More importantly, these relationships can become a conduit for mapping the politicians’ preferences into firm‐level decisions, especially for non‐state‐owned firms.3

These political forces and their effects on the cost of releasing bad news are expected to vary by firm and provincial characteristics. This variation would enable us to test whether such political forces are related to the occurrence and degree of bad news suppression in advance of political events. We have a number of ways to capture this variation. First, the activities of politically connected firms and executives are likely to be more scrutinized than firms and executives without political connections. We expect that these relationships, and the political benefits and repercussions it can bring, potentially raise the relative cost of releasing bad news for affected firms, politicians, and executives. Second, we expect that politicians promoted on the basis of faction politics or family relations (rather than meritocracy), and the central government officials sponsoring them, have stronger incentives to suppress any bad news leading up to the promotion announcement in order to bring legitimacy to their advancement. Additionally, these same, powerful politicians are more likely to be able to exercise influence over listed firms and information dissemination channels than politicians lacking this political clout. Third, state‐controlled firms, which ultimately report to the local politician, typically suffer from numerous inefficiencies and agency conflicts; as a result, politicians have both the ability and motive to suppress negative information about these companies.4 However, local politicians can also influence the reporting and disclosure incentives of non‐state‐owned firms through conferring or denying benefits to non‐state‐owned entities such as access to capital and other investment opportunities (e.g., Chen and Yuan [2004]). Thus, these nonstate firms may even have just as strong an incentive to suppress bad news as state firms in order to establish close ties with politicians. It remains an empirical question whether the political incentives to temporarily suppress bad news are sufficiently strong to influence both state and nonstate firms or are limited to state‐controlled firms. Finally, we expect that firms operating in provinces with greater capital market development, issued B‐shares that are owned by foreign investors, and listed in the Hong Kong market are more sheltered from political pressures to defer the release of bad news during the two political events.

2.2. PRIOR LITERATURE ON THE RELATION BETWEEN POLITICAL INCENTIVES AND FINANCIAL REPORTING PRACTICES

We view our paper as a natural complement to prior research examining how political forces influence financial reporting incentives and practices. In their seminal paper, Watts and Zimmerman [1978] argue that accounting choices are influenced by the expected political costs associated with given financial reporting outcomes. These political costs can range from heightened tax burdens (e.g., windfall profits tax) to the outright expropriation of the firms’ productive assets, and also include a host of indirect taxes such as heightened regulation or the threat of greater government intervention into the firm's business activities. Consistent with these arguments, existing research using international (i.e., non‐U.S.) data finds that firms tend to remain opaque in the presence of strong political economy forces, as measured by level of state ownership of assets, greater risk of expropriation, and the presence of political connections (e.g., Bushman, Piotroski, and Smith [2004], Bushman and Piotroski [2006], Gul [2006], Leuz and Oberholzer‐Gee [2006], Wang, Wong, and Xia [2008], Chaney, Faccio, and Parsley [2011], Durnev and Guriev [2011]). Moreover, these financial reporting incentives shift as the firm becomes more or less exposed to these political incentives over time, such as following privatization events (e.g., Guedhami and Pittman [2006], Wang, Wong, and Xia [2008], Guedhami, Pittman, and Saffar [2009]).

Prior research focusing on financial reporting incentives in specific political contexts typically focuses on earnings management behavior via discretionary accruals. These papers, including Jones [1991], Petroni [1992], Han and Wang [1998], and Ramanna and Roychowdhury [2010], document that firms alter reported net income to either garner expected political gains or minimize expected political costs arising in their setting. However, whether such earnings management behavior had a net impact on the firm's overall information environment (as proxied by stock prices) is unclear.

3. Research Design, Sample Construction, and Data Sources

3.1. DATA

We utilize a sample of domestic, publicly listed Chinese firms over the period 1993 to 2011. To be included in the sample, we require the firm to have sufficient accounting and stock price data available in the China Security Market and Accounting Research (CSMAR) data set to estimate our empirical models. Our final sample consists of 12,723 firm‐year observations, split between 8,608 state‐controlled firms and 4,115 non‐state‐owned firms, respectively.5 Firms are classified as state‐controlled if local government controls more than 20% of the outstanding shares of the company. Data on ownership was manually collected from annual reports.

3.2. RESEARCH DESIGN

3.2.1. Measure of Stock Price Crash Frequency

To measure firm‐level variation in the flow of negative information into stock prices, we utilize a crash statistic that reflects the presence of large, negative stock price movements. This statistic, Ncskew_i,t, is taken directly from Jin and Myers [2006] and Chen, Hong, and Stein [2001], and is measured as the third moment of each stock's daily residual returns, divided by the cubed standard deviation of daily residual returns, times negative one, in year t. By putting the negative sign in front of the standardized third moment, an increase in Ncskew_i,t corresponds to greater left (i.e., negative) skewness in the distribution of daily residual returns.

Ncskew_i,t is calculated using the residuals from annual, firm‐specific estimations of the following market model using firm i's daily returns in year t:

$$\begin{array}{ccc}\hfill {\mathit{Daily}\mathit{Return}}_{i,j,t}& =& {\alpha +\beta }_1{\mathit{Market}}_{China,j,t}+{\beta }_2{\mathit{Market}}_{China,j-1,t}\hfill \\ & & +{\beta }_3{\mathit{Market}}_{China,j-2,t}+{\beta }_4{\mathit{Market}}_{\mathit{US},j,t}\hfill \\ & & +{\beta }_5{\mathit{Market}}_{\mathit{US},j-1,t}+{\beta }_6{\mathit{Market}}_{US,j-2,t}+\varepsilon .\hfill \end{array}$$ (1)

In this model, Daily Return_i,j,t is the raw return for firm i on day j of year t, Market_China,j,t is the value‐weighted market return for the Chinese stock market on day j of year t, and Market_US,j,t is the value‐weighted market return for the U.S. market on day j of year t.6 Data on the China market return was gathered through CSMAR; data on the U.S. market return was gathered through CRSP.

Following Chen, Hong, and Stein [2001] and Jin and Myers [2006], among others, we interpret firms with greater negative skewness as being more crash prone, and equate the presence of a large stock price crash as indicative of the release of material negative information about the firm's financial performance.7 A key takeaway from these prior papers is that the temporary suppression of negative information will produce a distinctive time‐series pattern in stock returns—the period of active suppression will prevent large stock price crashes, and the subsequent release of the suppressed information will generate large stock price crashes. Given those theoretical relations, we use the observed pattern of stock price behavior to infer the suppression of negative information around events that create an ex ante incentive for such behavior. Specifically, a decline in negative skewness during an event window is interpreted as a temporary reduction in the flow of negative information about these firms. A subsequent increase in large negative stock price movements immediately following the event window (i.e., greater negative skewness) reflects the release of the previously suppressed negative information.8

3.2.2. Political Events Creating an Incentive to Suppress Negative Information

Our empirical tests search for the aforementioned pattern in stock returns of Chinese listed firms in the periods directly preceding and following two anticipated political events: the National Congress of the CCP and key provincial‐level political promotions. We hypothesize that these two highly visible political events asymmetrically heighten the costs of releasing adverse news for the local politicians and the affiliated firms.

National Congress of the CCP: The National Congress is held once every five years, and represents the most significant central government meeting in China. Over our sample period, this National Congress was held in calendar years 1997, 2002, and 2007. These meetings outline central government policy, identify party leaders, highlight key developments, and set major party objectives for the next five years. Releasing negative information during congress years inherently contradicts an objective of these meetings and would be expected to impose a large political cost on local politicians and affiliated firms. Additionally, these meetings draw considerable domestic and international attention. To the extent that investor, public, and political scrutiny are elevated around a National Congress, all firms would have an incentive to delay the release of bad news during this event window. Firm‐years associated with National Congress meetings are classified as being sensitive to these political events.

Promotions of high‐level provincial politicians: To create a sample of political promotion events capable of shifting economic incentives in a province, we identify all provincial turnover events involving the transfer, reassignment, or promotion of either the provincial party secretary or the provincial governor to a position with more political power. These promotion events are visible and anticipated by economic agents in the province. Because the party secretary and governor are the top two positions in a province, these promotions create a tournament‐like contest for personal advancement within the provincial political and communist party structure; thus, all politicized agents in the province have an incentive to window‐dress economic performance before these events. Moreover, the central government can exert pressure on local firms to improve the credentials of the identified provincial officials before their promotion. This top‐down pressure also prevents local political rivals from releasing bad news to undermine the credentials of the soon‐to‐be‐elevated provincial politician. Because of the planned nature of political transitions in China, politicians and affected firms have a well‐defined window (usually 12 to 18 months) over which these incentives arise (Shih, Adolph, and Liu [2012]). Prior research shows that political promotions in China are influenced by the recent performance of the region (Chen, Li, and Zhou [2005], Li and Zhou [2005]); as such, it is costly for local politicians and managers to release negative financial information about local state‐controlled firms in advance of the appointment. Additionally, non‐state‐owned firms also have heightened incentives in advance of these promotion events given the importance of maintaining strong relationships with local politicians, as such relations grant the firm preferential access to capital and investment opportunities and reduced bureaucracy. Given the duration of the evaluation process, we define our event window as the year preceding and containing the promotion event (years ‒1 to 0). Firms that operate in the same province as the promoted politician are considered to be those affected by these promotion events. Data on political promotion events are hand‐collected by searching information published in the “Chinese Personnel Database” and “China VIPs” from China Information Bank and supplemented by Google web searches.

3.2.3. Research Design: Empirical Model

Our primary tests search for an association between Ncskew_i,t and the political events hypothesized to temporarily increase the incentive to suppress negative information. Specifically, we estimate variations of the following cross‐sectional model:

$$\begin{array}{ccc}\hfill {\mathit{Ncskew}}_{i,t}& =& \alpha +\mathit{Year}+\mathit{Province}{+\beta }_1{\mathit{Political}}_{i,t}{+\beta }_2\mathit{Post}-{\mathit{Political}}_{i,t}\hfill \\ & & +{\beta }_3{\mathit{Logsize}}_{i,t}{+\beta }_4{\mathit{Growth}}_{i,t}+{\beta }_5{\mathit{Sigma}}_{i,t}{+\beta }_6{\mathit{Turnover}}_{i,t}\hfill \\ & & +{\beta }_7{\mathit{Turnover}}_{i,t-1}{+\beta }_8{\mathit{Beta}}_{i,t}{+\beta }_9{\mathit{Return}}_{i,t}{+\beta }_{10}{\mathit{Return}}_{i,t-1}{+\varepsilon }_{i,t}.\hfill \end{array}$$ (2)

In this model, Political_i,t is an indicator variable, which is equal to one if the firm‐year falls within the political event window, zero otherwise. If temporary incentives to suppress negative information are created, we expect a negative coefficient on the indicator variable Political_i,t. The indictor variable Post‐Political_i,t equals one if the firm‐year falls in the year following the political event, zero otherwise. To the extent that the firm subsequently releases negative information following the political event, we expect a positive coefficient on the indicator variable Post‐Political_i,t. This research design allows the indicator variables to compare stock return behavior across firm‐years affected by the event against firm‐years unaffected by the event (e.g., National Congress years versus non‐National Congress years; firm‐years associated with provincial promotion versus firm‐years without provincial promotion).9

The remaining independent variables are drawn from prior research examining the determinants of stock return skewness. We include these variables to capture firm characteristics that could potentially induce a stock return crash independent of the historical suppression of financial information (e.g., Chen, Hong, and Stein [2001], Harvey and Siddique [2000]). Firm size, measured as the log of the firm's market value of equity at the end of year t (LogSize_i,t), is expected to be negatively related to negative skewness, as larger firms are less likely to experience dramatic price declines due to their diverse set of operations and their rich information environment. The firm's beta (Beta_i,t) and the standard deviation of daily residual returns (Sigma_i,t) are included to capture the systematic and idiosyncratic risk characteristics of the firm during year t. The negative skewness of residual returns is expected to be decreasing in Beta_i,t and Sigma_i,t because the distribution of payoffs to risky projects is positively skewed.10 The inclusion of Sigma is also important in this setting because political risk is associated with stock return volatility (Boutchkova, Doshi, and Durnev [2012]). Similarly, we include firm growth, measured as the growth in sales in year t, to capture the potential effect of growth options and correlated activities (such as R&D) on the distribution of realized returns.

Two control variables, average weekly share turnover (Turnover_i,t) and the annual stock return (Return_i,t), are included in the model as both contemporaneous and lagged variables. The firm's contemporaneous annual stock return controls for current firm performance, allowing for the examination of the short‐term flow of information conditional upon the firm's average performance. Contemporaneous share turnover is included to capture liquidity effects and investor interest. To the extent that stocks with heightened trading volume are more prone to crashes (i.e., torpedoed), contemporaneous share turnover will be positively related to negative skewness (e.g., Chen, Hong, and Stein [2001]). The inclusion of lagged share turnover and lagged annual returns allows us to control for momentum and behavioral factors (i.e., heightened investor sentiment, irrational exuberance, and bubbles) that can produce price reversals in the subsequent year (e.g., Chen, Hong, and Stein [2001]). We also include year and provincial fixed effects in all promotion estimations to control for unobservable characteristics.11 All explanatory variables are defined in appendix A.12

3.3. DESCRIPTIVE STATISTICS

Table 1 presents descriptive evidence on the distribution of our firm‐year observations across time and provinces. The number of observations increases dramatically over our 19‐year sample period, corresponding to the underlying growth in China's capital markets (panel A). The composition of our sample is tilted toward state‐owned entities, with nearly 67.7% of our total firm‐year observations drawn from local, state‐controlled firms. However, that relative percentage declines over our sample period, as economic policies promoting greater market development and encouraging entrepreneurial activities lead to an increase in non‐state‐owned firms listing onto China's exchanges in the latter half of our sample period.

Table 1.

Composition of Sample and Political Events

Panel A: Distribution of firm‐year observations and political events across sample time period
		State‐	Non‐State‐	Political
	All	Controlled	Owned	Promotion
Year	Firms	Firms	Firms	Events
1993	8	8	–	1
1994	39	39	–	6
1995	144	135	9	6
1996	228	213	15	3
1997	249	229	20	4
1998	398	354	44	11
1999	536	469	67	5
2000	618	524	94	2
2001	679	559	120	6
2002	782	575	207	9
2003	850	600	250	8
2004	890	605	285	2
2005	926	615	311	2
2006	979	612	367	5
2007	956	593	363	14
2008	988	602	386	2
2009	1,095	628	467	1
2010	1,159	631	528	4
2011	1,199	617	582	5
Total	12,723	8,608	4,115	96

Panel B: Distribution of firm‐year observations and political events across China's provinces
		State‐	Non‐State‐	Political
	All	Controlled	Owned	Promotion
Province	Firms	Firms	Firms	Events
Beijing	476	331	145	5
Tianjin	219	184	35	1
Shanghai	1505	1,178	327	4
Chongqing	233	170	63	1
Hebei	281	226	55	2
Shanxi	226	196	30	4
Inner Mongolia	151	116	35	1
Liaoning	499	340	159	5
Jilin	307	200	107	5
Heilongjiang	230	134	96	3
Jiangsu	839	495	344	6
Zhejiang	910	407	503	2
Anhui	436	350	86	5
Fujian	484	315	169	6
Jiangxi	193	176	17	3
Shandong	714	537	177	5
Henan	323	252	71	2
Hubei	490	300	190	5
Hunan	356	257	99	6
Guangdong	1,597	979	618	3
Guangxi	213	149	64	1
Hainan	213	97	116	5
Sichuan	592	349	243	2
Guizhou	130	95	35	2
Yunnan	217	161	56	3
Shaanxi	231	174	57	2
Gansu	176	140	36	4
Qinghai	96	56	40	3
Ningxia	109	88	21	0
Xinjiang	180	99	81	0
Tibet	97	57	40	0
Total	12,723	8,608	4,115	96

This table presents the distribution of our samples of Chinese listed firms and political promotion events by fiscal year (panel A) and provincial region (panel B). The sample consists of two types of listed firms, those that are ultimately controlled by the local government (i.e., state‐controlled firms), and those that are not controlled by local government entities (i.e., non‐state‐owned firms), with sufficient financial accounting and stock price data available in the China Security Market and Accounting Research (CSMAR) database to implement our primary empirical tests. A political promotion event relates to the transfer, reassignment, or promotion of either the provincial party secretary or the provincial governor to a position of more political power.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

Geographically, our sample firms are drawn from all 31 of China's provinces, 4 of which are officially called direct‐controlled municipalities but have provincial status. Guongdong and Shanghai contain the largest fraction of firm‐year observations (12.6% and 11.8%, respectively; see panel B). In terms of sample composition, the distribution of firm‐years across provinces is fairly similar for both state‐controlled firms and private firms.

Table 1 also documents the distribution of our provincial‐level political promotion events across time (panel A) and provinces (panel B). Over our sample time period, there exist 96 political promotion events across 28 provinces. For purposes of our empirical tests, these promotion events are not strictly clustered in time or specific provinces; all but three of our provinces experience at least one promotion event, and there is at least one event per year over the entire sample period.

Table 2 presents descriptive statistics on our sample of Chinese listed firms. These firms tend to be large entities, with an average market capitalization of 2.19 billion RMB. Despite their large size, the firms display considerable variation in terms of economic performance (Return_i,t), sales growth (Growth_i,t), share liquidity, and investor interest (Turnover_i,t) and risk (Beta_i,t and Sigma_i,t). The descriptive statistics highlight substantial variation in our crash measure, Ncskew_i,t, across firm‐years, and corroborate prior findings that state‐owned entities have weaker information environments than non‐state‐owned firms in China (i.e., are more crash prone (on average) and have more synchronous prices).

Table 2.

Descriptive Statistics

	All Firms (N = 12,723)			State‐Controlled Firms (N = 8,608)			Non‐State‐Owned Firms (N = 4,115)
	Mean	Median	Std. Dev.	Mean	Median	Std. Dev.	Mean	Median	Std. Dev.
Ncskew	−0.630	−0.607	0.660	−0.650	−0.630	0.674	−0.588	−0.562	0.627
Synchronicity	−0.445	−0.454	0.722	−0.376	−0.398	0.723	−0.563	−0.551	0.705
LogSize	14.746	14.648	0.986	14.804	14.698	0.981	14.625	14.537	0.985
Growth	0.113	0.122	0.393	0.116	0.119	0.351	0.108	0.129	0.468
Sigma	0.023	0.022	0.007	0.022	0.021	0.007	0.024	0.023	0.007
Turnover	0.110	0.089	0.078	0.103	0.084	0.074	0.123	0.103	0.084
Turnovert− 1	0.114	0.095	0.080	0.107	0.090	0.074	0.129	0.110	0.088
Beta	1.016	1.038	0.222	1.017	1.036	0.213	1.013	1.041	0.239
Return	0.070	−0.026	0.534	0.055	−0.028	0.508	0.100	−0.022	0.583
Returnt− 1	0.079	−0.022	0.551	0.061	−0.026	0.521	0.118	−0.011	0.605
Ind_Size	27.276	27.538	1.523	27.115	27.429	1.497	27.554	27.844	1.529
Ind_Num	5.244	6.087	1.391	5.127	5.894	1.389	5.447	6.142	1.371
Std_ROA	0.036	0.024	0.033	0.033	0.022	0.030	0.040	0.026	0.036
Regulated	0.077	0.000	0.267	0.095	0.000	0.293	0.046	0.000	0.209

This table presents descriptive statistics on the sample of Chinese listed firms included in our study. The first set of columns presents descriptive financial statistics on our complete sample. The second (third) set of columns presents descriptive financial statistics on our sample of listed firms that are ultimately controlled by the local government (not controlled by local or central government entities). All financial accounting and stock price data are gathered from the China Security Market and Accounting Research (CSMAR) database. Ncskew is the firm's third moment of excess daily stock returns scaled by its cubed standard deviation times minus one. Synchronicity is log(R ²/(1‒R ²)), where R ² is from the estimation of market model of firm weekly return against the current and one week lagged domestic market return and U.S. market return. LogSize is the log of the firm's total market value of equity. Growth is the log of one plus the firm's annual sales growth rate. Sigma is the standard deviation of daily excess returns. Turnover is the average weekly turnover of the firm's shares, defined as the total value traded of tradable shares scaled by total value of tradable shares at the end of the week. Beta is the firm's equity market beta, estimated using daily returns. Return is the annual market‐adjusted stock return of the firm. Std_ROA is the standard deviation of the firm's return on assets over the past five years. Regulated is an indicator variable equal to one for companies operating in a regulated industry, zero otherwise. Ind_Num is the log of the total number of listed companies operating in the same two‐digit SIC industry at fiscal year end. Ind_Size is the log of total assets of all listed companies in the same two‐digit SIC industry at fiscal year end. All variables are also defined in appendix A. The sample of state‐controlled and non‐state‐owned firms with Synchronicity is reduced to 4,940 and 2,846 firm‐year observations, respectively, due to data limitations.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

Finally, table 3 presents a correlation matrix for our firms’ crash statistics and firm‐level characteristics. Our crash statistic, Ncskew_i,t, displays significant negative relations with several firm‐level characteristics, most notably, firm size, share turnover, current performance, beta, and stock return volatility. These correlations highlight the importance of explicitly controlling for firm‐specific attributes in our tests.

Table 3.

Correlation Matrix

		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
(1)	Ncskew	1.00
(2)	Synchronicity	−0.09***	1.00
(3)	LogSize	−0.11***	0.03***	1.00
(4)	Growth	0.03***	−0.06***	0.19***	1.00
(5)	Sigma	−0.08***	−0.41***	−0.03***	−0.04***	1.00
(6)	Turnover	−0.15***	−0.20***	0.04***	−0.03***	0.76***	1.00
(7)	Turnovert− 1	−0.10***	0.13***	0.07***	−0.01	0.39***	0.54***	1.00
(8)	Beta	−0.17***	0.55***	0.00	−0.07***	0.12***	0.24***	0.28***	1.00
(9)	Return	−0.08***	−0.11***	0.29***	0.16***	0.23***	0.20***	0.15***	0.00	1.00
(10)	Returnt− 1	−0.01*	0.03***	0.20***	0.11***	−0.02**	−0.01	0.21***	0.07***	−0.02**	1.00
(11)	SOE	−0.04***	0.13***	0.08***	0.01	−0.12***	−0.12***	−0.13***	0.01	−0.04***	−0.05***	1.00
(12)	Std_ROA	0.08***	−0.15***	−0.17***	−0.16***	0.13***	0.06***	0.06***	−0.04***	−0.06***	−0.04***	−0.11***	1.00
(13)	Regulated	−0.01	0.04***	0.12***	0.03***	−0.04***	−0.03***	−0.01*	−0.01	0.00	0.00	−0.07***	−0.03***	1.00
(14)	Ind_Num	−0.04***	−0.02**	0.06***	0.04***	0.12***	0.18***	0.21***	0.08***	0.08***	0.09***	−0.11***	0.01	−0.44***	1.00
(15)	Ind_Size	−0.05***	−0.01	0.16***	0.07***	0.13***	0.19***	0.22***	0.08***	0.09***	0.10***	−0.14***	−0.01	−0.20***	0.90***

This table presents Spearman correlations between the financial characteristics of our sample of Chinese listed firms and provincial institutional characteristics. All financial accounting and stock price data are gathered from the China Security Market and Accounting Research (CSMAR) database. Ncskew is the firm's third moment of excess daily stock returns scaled by its cubed standard deviation times minus one. Synchronicity is log(R ²/(1‒R ²)), where R ² is from the estimation of the market model of firm weekly return against the current and one week lagged domestic market return and U.S. market return. LogSize is the log of the firm's total market value of equity. Growth is the log of one plus the firm's annual sales growth rate. Sigma is the standard deviation of daily excess returns. Turnover is the average weekly turnover of the firm's shares, defined as the total value traded of tradable shares scaled by total value of tradable shares at the end of the week. Beta is the firm's equity market beta, estimated using daily returns. Return is the annual market‐adjusted stock return of the firm. Std_ROA is the standard deviation of the firm's return on assets over the past five years. Regulated is an indicator variable equal to one for companies operating in a regulated industry, zero otherwise. Ind_Num is the log of the total number of listed companies operating in the same two‐digit SIC industry at fiscal year end. Ind_Size is the log of total assets of all listed companies in the same two‐digit SIC industry at fiscal year end. Unemployment is the average annual unemployment rate over the sample period for the listed firm's province. SOE is an indicator variable equal to one if the firm is state‐controlled. All variables are also defined in appendix A. The sample of state‐controlled and non‐state‐owned firms with Synchronicity is reduced to 4,940 and 2,846 firm‐year observations, respectively, due to data limitations. The superscripts ***, **, and * denote two‐tailed significance at the 1%, 5%, and 10% level, respectively.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

4. Empirical Results

This section presents our primary set of empirical analyses. These analyses employ the negative skewness methodology outlined in section 3. utilizing the full sample of firms (section 4.1), conditional upon the degree of the firm's political connections (section 4.2), conditional upon the political affiliation of the firm's local politicians (section 4.3), conditional upon whether or not the firm is controlled by the state (section 4.4), and conditional upon whether Chinese firms listed on a foreign stock exchange listing or exposed to strong market‐oriented pressures are differentially affected by these political costs (sections 4.5 and 4.6).

4.1. IMPACT OF POLITICAL EVENTS ON THE SUPPRESSION OF NEGATIVE INFORMATION

Table 4 presents select coefficients (and t‐statistics using clustered standard errors by province) from various pooled cross‐sectional estimations of equation (2). The first and second pair of estimations presents coefficients from analyses focusing on National Congress meetings and provincial promotions events, respectively. For each pair of estimations, the first column presents select coefficients from an estimation of equation (2) and the second column presents coefficients after estimating equation (2) with an additional control variable for state ownership.13 These estimations reveal two key findings.

Table 4.

Influence of Political Events on the Incentive to Suppress Negative Financial Information

Political	National		Provincial‐Level
Event:	Congress		Political Promotion
	(1)	(2)	(1)	(2)
Politicalt	−0.241***	−0.228***	−0.058**	−0.058**
	(−8.37)	(−7.61)	(−2.10)	(−2.12)
Post‐Politicalt	0.054**	0.058**	0.122***	0.121***
	(2.25)	(2.34)	(3.75)	(3.75)
LogSizet	−0.043***	−0.041***	−0.055***	−0.052***
	(−7.31)	(−6.70)	(−6.46)	(−5.77)
Growtht	−0.019	−0.019	−0.013	−0.013
	(−1.68)	(−1.68)	(−1.09)	(−1.08)
Sigmat	−9.432***	−9.554***	−18.744***	−19.006***
	(−4.95)	(−4.94)	(−7.89)	(−7.84)
Turnovert	0.927***	0.924***	0.397***	0.396***
	(7.36)	(7.36)	(2.79)	(2.78)
Turnovert− 1	−0.416***	−0.434***	−0.158	−0.176
	(−4.12)	(−4.19)	(−0.87)	(−0.96)
Betat	−0.275***	−0.271***	−0.296***	−0.291***
	(−8.90)	(−8.70)	(−10.88)	(−10.67)
Returnt	−0.037***	−0.038***	0.020	0.019
	(−3.58)	(−3.70)	(1.62)	(1.53)
Returnt− 1	0.078***	0.077***	0.076***	0.075***
	(10.39)	(10.28)	(8.42)	(8.15)
SOEt		−0.030*		−0.037**
		(−2.00)		(−2.26)
Year Fixed Effects	Included	Included	Included	Included
Provincial Fixed Effects	Included	Included	Included	Included
Adjusted R  2	0.174	0.174	0.125	0.126

This table presents coefficients from various pooled, cross‐sectional estimations of the following model:

$$\begin{array}{ccc}\hfill \mathit{Nc}{\mathit{skew}}_{i,t}& =& \alpha +\mathit{Year}+\mathit{Province}+{\beta }_1{\mathit{Political}}_{i,t}+{\beta }_2\mathit{Post}-{\mathit{Political}}_{i,t}+{\beta }_3{\mathit{Logsize}}_{i,t}+{\beta }_4{\mathit{Growth}}_{i,t}\hfill \\ & & +{\beta }_{5}Sigm{a}_{i,t}+{\beta }_6{\mathit{Turnover}}_{i,t}+{\beta }_7{\mathit{Turnover}}_{i,t-1}+{\beta }_8{\mathit{Beta}}_{i,t}+{\beta }_9{\mathit{Return}}_{i,t}\hfill \\ & & +{\beta }_{10}{\mathit{Return}}_{i,t-1}+{\beta }_{11}{\mathit{SOE}}_{i,t}+{\varepsilon }_{i,t}.\hfill \end{array}$$

The dependent variable Ncskew is the firm's third moment of excess daily stock returns scaled by its cubed standard deviation times minus one. Political is an indicator variable equal to one if the firm‐year relates to a specific political event; Post‐Political is an indicator variable equal to one in the year immediately following the event. In the first set of estimations, Political is an indicator variable equal to one for the years that a National Congress of the CCP was held, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following each National Congress event, zero otherwise. In the second set of estimations, Political is an indicator variable equal to one for the year preceding and corresponding to a provincial‐level political promotion event, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following the provincial‐level promotion event, zero otherwise. All other variables are defined in table 1 and appendix A. t‐statistics derived using clustered standard errors by province are presented in parentheses. Models include annual and provincial fixed effects (Year and Province, respectively; coefficients not reported). The superscripts ***, **, and * indicate that the estimated coefficient is significantly different from zero at the 1%, 5%, and 10% level (two‐tailed test), respectively. N = 12,723.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

First, Chinese listed firms are significantly less likely to experience negative stock price crashes during National Congress years relative to non‐congress years. This decline is followed by the predicted increase in negative stock return skewness in the year immediately following the National Congress, as affected firms and politicians eventually release the temporarily suppressed negative information. Economically, the observed reduction equates to a one‐third standard deviation reduction in negative stock return skewness during National Congress years, suggesting that these events and related incentives generate a meaningful shift in the overall information environment of listed firms.

Second, we find that affected firms are less likely to experience negative stock price crashes in the two years preceding and including provincial political promotions relative to firm‐years not associated with provincial promotion events. This decline is followed by a temporary increase in negative skewness in the year following the promotion, consistent with the subsequent release of previously suppressed information. This subsequent increase in Ncskew_i,t alleviates concerns that the positive association between Ncskew_i,t and Political_i,t is an artifact of reverse causality, namely, provinces with strong economic conditions and stock price performance generate more political promotions. Economically, the observed reduction equates to a one‐tenth of a standard deviation reduction in negative stock return skewness in advance of political promotions. This decline is less severe than the reduction observed around National Congress years, which could be an artifact of the intensity of the National Congress period, the localized aspect of political promotions, or the long‐window nature of the political promotion process. 14

Together, the time‐series pattern of negative stock return skewness (relative to nontreatment firm‐years) is consistent with the systematic suppression and subsequent release of negative information around these two political events and limitations in the ability of market participants to unravel silence as bad news (e.g., Giglio and Shue [2013]).15 More importantly, this observed stock price behavior is being observed in two settings where (1) local politicians have an acute incentive to exert influence over firms under their direct and indirect control to minimize their personal cost of reporting bad news during these sensitive periods and (2) executives of local firms have a self‐interested incentive to suppress negative information to protect valuable connections and curry favor from affected politicians. The next three subsections examine these patterns conditional upon factors correlated with the expected political costs of revealing negative information around these events.

4.2. INFLUENCE OF POLITICAL CONNECTIONS ON THE INCENTIVE TO SUPPRESS NEGATIVE FINANCIAL INFORMATION

Next, we examine whether firms with political connections experience incrementally fewer stock price crashes around our two political events. We measure the firms’ political connectivity using one executive‐level measure (i.e., CEO background) and two indirect measures of the strength of the firm's connections (i.e., the extent to which the firm receives preferential treatment from the government, in the form of subsidies and access to state bank financing). We estimate variations of the following pooled, cross‐sectional model to examine how the firms’ political connections influenced incentives around our two events:

$$\begin{array}{ccc}\hfill {\mathit{Ncskew}}_{i,t}& =& \alpha +\mathit{Year}+\mathit{Province}+{\beta }_1{\mathit{Political}}_{i,t}\times {\mathit{Connection}}_{i,t}\hfill \\ & & +{\beta }_3\mathit{Post}-{\mathit{Political}}_{i,t}{+\beta }_4\mathit{Post}-{\mathit{Political}}_{i,t}\times {\mathit{Connection}}_{i,t}\hfill \\ & & +{\beta }_5{\mathit{Connection}}_{i,t}{+\beta }_6{\mathit{LogSize}}_{i,t}+{\beta }_7{\mathit{Growth}}_{i,t}+{\beta }_{8}Sigm{a}_{i,t}\hfill \\ & & +{\beta }_9{\mathit{Turnover}}_{i,t}+{\beta }_{10}{\mathit{Turnover}}_{i,t-1}+{\beta }_{11}Bet{a}_{i,t}+{\beta }_{12}{\mathit{Return}}_{i,t}\hfill \\ & & +{\beta }_{13}{\mathit{Return}}_{i,t-1}{+\beta }_{14}{\mathit{SOE}}_{i,t}{+\varepsilon }_{\mathit{i},\mathit{t}}.\hfill \end{array}$$ (3)

In the first set of estimations, Connection_i,t is an indicator variable equal to one if the firm's CEO has past or concurrent work experience in government or the Party, zero otherwise. In the second set of estimations, Connection_i,t is an indicator variable equal to one if the firm received a government subsidy in year t, zero otherwise. In the third set of estimations, Connection_i,t is an indicator variable equal to one if the firm's balance sheet reflects long‐term debt from a state bank in year t, zero otherwise. Data on CEO backgrounds are gathered through the firm's annual report.16 Data on the receipt of government subsidies and the use of long‐term state bank loans are gathered through CSMAR.17

Our estimations, presented in table 5, find that the political background of the firm's chief executive has no impact on Ncskew_i,t around either event. The lack of an observed relation could be because CEO backgrounds do not adequately capture variation in the strength of political connections in China. For example, individuals appointed to run state companies are selected by the government; as a result, these CEOs are likely connected in spite of their previous work experience. Similarly, the strength of a successful entrepreneur's political connection is more likely to be related to the success of his company and less about his/her previous work experience.

Table 5.

Political Incentives to Suppress Negative Information Conditional upon the Political Connections of the Listed Firm

Political Event:	National Congress			Provincial‐Level Political Promotion
Measure of	Executive	Government	State Bank	Executive	Government	State Bank
Connection:	Connection	Subsidy	Loans	Connection	Subsidy	Loans
Politicalt	−0.124***	−0.234***	−0.113**	−0.069**	−0.088*	−0.074**
	(−3.18)	(−4.34)	(−2.26)	(−2.36)	(−1.96)	(−2.47)
Politicalt × Connection	0.019	−0.089*	−0.162**	−0.011	−0.049*	0.067
	(0.44)	(−1.73)	(−2.48)	(−0.32)	(−1.78)	(1.41)
Post‐Politicalt	0.147***	0.075*	0.153***	0.129***	0.132**	0.123***
	(4.03)	(1.95)	(3.81)	(4.11)	(2.42)	(3.57)
Post‐Politicalt × Connection	0.042	−0.048	−0.017	−0.038	0.058	−0.033
	(1.10)	(−1.18)	(−0.53)	(−1.60)	(1.33)	(−0.65)
Connection	0.017	0.051**	0.023	0.044*	0.004	0.009
	(0.95)	(2.49)	(0.97)	(2.02)	(0.14)	(0.28)
Control Variables	Included	Included	Included	Included	Included	Included
Year Fixed Effects	Included	Included	Included	Included	Included	Included
Provincial Fixed Effect	Included	Included	Included	Included	Included	Included
Cumulative (Connection = 0)	−0.271	−0.309	−0.266	−0.198	−0.220	−0.197
Cumulative (Connection = 1)	−0.294	−0.350	−0.411	−0.171	−0.327	−0.097
Difference (1 minus 0)	−0.023	−0.041	−0.145	0.027	−0.107	0.100
(F‐statistic)	(0.22)	(0.46)	(4.37)**	(0.46)	(2.36)	(5.09)**
Adjusted R  2	0.087	0.201	0.087	0.095	0.153	0.095
Number of Observations	11,657	7,326	11,657	11,657	7,326	11,657

This table presents select coefficients from pooled, cross‐sectional estimations of the following model:

$$\begin{array}{ccc}\hfill {\mathit{Ncskew}}_{i,t}& =& \alpha +\mathit{Year}+\mathit{Province}+{\beta }_1{\mathit{Political}}_{i,t}+{\beta }_2{\mathit{Political}}_{i,t}\times {\mathit{Connection}}_{i,t}+{\beta }_3\mathit{Post}-{\mathit{Political}}_{i,t}+{\beta }_4\mathit{Post}-{\mathit{Political}}_{i,t}\times {\mathit{Connection}}_{i,t}+{\beta }_5{\mathit{Connection}}_{i,t-1}\hfill \\ & & +{\beta }_6{\mathit{LogSize}}_{i,t}+{\beta }_7{\mathit{Growth}}_{i,t}+{\beta }_{8}Sigm{a}_{i,t}+{\beta }_9{\mathit{Turnover}}_{i,t}+{\beta }_{10}{\mathit{Turnover}}_{i,t-1}+{\beta }_{11}Bet{a}_{i,t}+{\beta }_{12}{\mathit{Return}}_{i,t}+{\beta }_{13}{\mathit{Return}}_{i,t-1}+{\beta }_{14}{\mathit{SOE}}_{i,t}+{\varepsilon }_i.\hfill \end{array}$$

The dependent variable Ncskew is the firm's third moment of excess daily stock returns scaled by its cubed standard deviation times minus one. In these estimations, Political is an indicator variable equal to one for the year of a National Congress meeting (first set of estimations) or the year preceding and corresponding to a provincial‐level political promotion event (second set of estimations), zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following the specific political event, zero otherwise. The indicator variable Connection captures one alternative measure of the firm's political connectivity. In the first estimation, Connection is an indicator variable equal to one if the CEO of the company has prior or concurrent work experience in government or the Party, zero otherwise. In the second estimation, Connection is an indicator variable equal to one if the firm has received a government subsidy, zero otherwise. In the third estimation, Connection is an indicator variable equal to one if the firm has loans from a state bank, zero otherwise. All other variables are defined in table 1 and appendix A. t‐statistics derived using clustered standard errors by province are presented in parentheses. Models include annual and provincial fixed effects (Year and Province, respectively; coefficients not reported). The superscripts ***,**, and * indicate that the estimated coefficient is significantly different from zero at the 1%, 5%, and 10% level (two‐tailed test), respectively. Differences in the cumulative pattern of suppression and reversal effects, measured as the sum of Political coefficients (β₁ + β₂) minus the sum of Post‐Political coefficients (β₃ + β₄), conditional upon the realization of Connection are tested using an F‐test.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

To overcome these measurement weaknesses, we also use the firm's ability to garner economic benefits from the state—in the form of a government subsidy or access to long‐term state bank loans—to measure the strength of the company's connections to local politicians. Prior research shows that the ability to gain access to preferential state bank loans or receive government subsidies is correlated with the strength of the firm's political connections (Khwaja and Mian [2005], Faccio [2006]); as such, our two measures of government benefits should capture the implicit strength of the firm's political relationships. Estimations using these two empirical proxies reveal that politically connected firms experience a greater reduction in Ncskew_i,t in advance of our political events, especially National Congress meetings; however, we are unable to document an incremental increase in crash frequency after the event for politically connected firms. The preevent variation, and tabulated cumulative suppression effects, provide evidence consistent with firms garnering economic benefits from the state responding to incrementally stronger incentives to temporarily suppress negative information in advance of these meetings; the postevent behavior suggests that either that these tests are also misspecified or that connected firms are being affected by an omitted force preventing or obscuring the post‐National Congress stock price crash reversal. Finally, we find only limited evidence supporting incrementally greater suppression by connected firms around political promotions.

4.3. INCENTIVES ARISING FROM THE POLITICAL CHARACTERISTICS OF KEY PROVINCIAL POLITICIANS

Next, we examine whether the personal political characteristics of the provincial politicians influence incentives to suppress negative information. From a research design perspective, documenting that shifts in negative skewness around our political events vary by the political attributes of the provincial politician will help tighten the link between personal political costs, information suppression incentives, and observed stock price behavior. We focus on two observable attributes used in prior research (Nathan and Tsai [1995], Shih [2008], Shih, Adolph, and Liu [2012]): the politician's affiliation with the Shanghai political faction and his/her princeling status. Each attribute is expected to impact the costs and benefits of releasing or suppressing negative information during these political periods and to capture differences in the politician's ability to directly or indirectly influence other economic agents in the province (e.g., local politicians and party members, management of listed firms, news media, and other information dissemination outlets).

The Shanghai faction was the most influential political faction during our sample period; as such, politicians affiliated with this faction wielded greater power during our sample period. This power translates into an ability to affect the disclosure decisions of firms under their direct and indirect control and to pressure media outlets in the province to censor their reporting and investigative activities in advance of promotion events or National Congress meetings. Similarly, politicians categorized as princelings—the children or grandchildren of former revolutionary leaders—are extremely powerful through their family and social connections and their control over key sectors of the economy. Additionally, because princelings and faction‐affiliated politicians are capable of conferring considerable benefits to supporters, local firms have an incentive to suppress negative information to maintain strong relations with these politicians.

We estimate variations of the following cross‐sectional model to examine how these personal attributes influenced incentives around the two political events:

$$\begin{array}{ccc}\hfill {\mathit{Ncskew}}_{i,t}& =& \alpha +\mathit{Year}+\mathit{Province}{+\beta }_1{\mathit{Political}}_{i,t}{+\beta }_2{\mathit{Political}}_{i,t}\times {\mathit{Attribute}}_{i,t}\hfill \\ & & +{\beta }_3\mathit{Post}-{\mathit{Political}}_{i,t}{+\beta }_4\mathit{Post}-{\mathit{Political}}_{i,t}\times {\mathit{Attribute}}_{i,t}\hfill \\ & & +{\beta }_5{\mathit{Attribute}}_{i,t}+{\beta }_6{\mathit{LogSize}}_{i,t}{+\beta }_7{\mathit{Growth}}_{i,t}{+\beta }_{8}Sigm{a}_{i,t}\hfill \\ & & +{\beta }_9{\mathit{Turnover}}_{i,t}{+\beta }_{10}{\mathit{Turnover}}_{i,t-1}+{\beta }_{11}{\mathit{Beta}}_{i,t}+{\beta }_{12}{\mathit{Return}}_{i,t}\hfill \\ & & +{\beta }_{13}{\mathit{Return}}_{i,t-1}+{\beta }_{14}{\mathit{SOE}}_{i,t}{+\varepsilon }_{\mathit{i},\mathit{t}}.\hfill \end{array}$$ (4)

In the first set of estimations, Attribute_i,t is an indicator variable equal to one if in year t the promoted politician or, in provinces without a promotion, either the provincial governor or party chairman, is a member of the Shanghai political faction, zero otherwise. In the second set of estimations, Attribute is an indicator variable equal to one if in year t the promoted politician or, in provinces without a promotion, either the provincial governor or party chairman, is a princeling, zero otherwise. In these estimations Attribute_i,t will capture, incremental to the provincial fixed effect, the provincial leaders’ political characteristics because, for each province, Attribute_i,t is set equal to one only in those periods when the province's governor or party chairman belongs to the Shanghai faction or is a princeling. Data on these political attributes are gathered through China Information Bank and http://zh.wikipedia.org, supplemented with Google searches.

Table 6 presents select coefficients from these estimations; two key results emerge. First, both the transitory suppression, and subsequent release, of negative information is incrementally stronger around promotion events involving politicians affiliated with the Shanghai faction. This amplification of promotion behavior is consistent with these incumbent, faction‐affiliated politicians bearing larger costs if adverse news is revealed and possessing greater power to temporarily suppress the information before their promotion. For the National Congress meetings, we find firms in provinces controlled by Shanghai‐faction‐affiliated politicians to exhibit more bad news suppression prior to the event, but there are no significantly more negative information releases after the event. Second, princelings and nonprincelings engage in similar levels of suppression in advance of our two political events. However, firms affiliated with princelings experience larger stock price crashes following these events (as noted by the positive coefficient on the term Post‐Political_i,t × Attribute_i,t in both estimations).18 One possible reason for the less consistent results for the National Congress meetings with the Shanghai‐faction‐affiliated politicians and the two political events for the princelings is the lack of power in the tests. The politicians’ factional affiliations may have a stronger effect on the release of bad news around their promotions than National Congress meetings. Also, not all princelings in our sample exert the same level of power as the Shanghai‐faction‐affiliated politicians, which could contribute to the weaker results for the two events involving princelings. Further, the cumulative suppression effects around political promotions for firms operating in provinces controlled by princelings or members of the Shanghai faction are significantly stronger than provinces not controlled by these powerful politicians. This variation is consistent with the characteristics of local politicians, and resultant incentives, shaping the information environment of listed firms around these political events.

Table 6.

Political Incentives to Suppress Negative Information Conditional upon Political Attributes of the Local Politicians

Political	National		Provincial‐Level
Event:	Congress		Political Promotion
Politician	Shanghai		Shanghai
Attribute:	Faction	Princeling	Faction	Princeling
Politicalt	−0.191***	−0.240***	−0.041**	−0.076***
	(−3.66)	(−3.71)	(−2.09)	(−4.39)
Politicalt × Attribute	−0.053	0.050	−0.101***	−0.033
	(−1.28)	(1.13)	(−3.25)	(−0.92)
Post‐Politicalt	0.055*	0.039*	0.055***	0.111***
	(1.98)	(1.77)	(2.97)	(7.04)
Post‐Politicalt × Attribute	0.008	0.075**	0.145***	0.057*
	(0.19)	(2.27)	(5.80)	(1.82)
Attribute	0.008	−0.057*	−0.039	−0.035
	(0.33)	(−1.98)	(−1.60)	(−1.49)
Control Variables	Included	Included	Included	Included
Year Fixed Effects	Included	Included	Included	Included
Provincial Fixed Effect	Included	Included	Included	Included
Cumulative (Attribute = 0)	−0.246	−0.279	−0.096	−0.187
Cumulative (Attribute = 1)	−0.307	−0.304	−0.342	−0.277
Difference (1 minus 0)	−0.061	−0.025	−0.246	−0.090
(F‐statistic)	(1.23)	(1.85)	(49.85)***	(4.72)**
Adjusted R  2	0.174	0.174	0.131	0.127

This table presents select coefficients from pooled, cross‐sectional estimations of the following model:

$$\begin{array}{ccc}\hfill {\mathit{Ncskew}}_{i,t}& =& \alpha +\mathit{Year}+\mathit{Province}+{\beta }_1{\mathit{Political}}_{i,t}+{\beta }_2{\mathit{Political}}_{i,t}\times {\mathit{Attribute}}_{i,t}+{\beta }_{3}Post-{\mathit{Political}}_{i,t}\hfill \\ & & +{\beta }_{4}Post-{\mathit{Political}}_{i,t}\times {\mathit{Attribute}}_{i,t}+{\beta }_5{\mathit{Attribute}}_{i,t-1}+{\beta }_6{\mathit{LogSize}}_{i,t}+{\beta }_7{\mathit{Growth}}_{i,t}\hfill \\ & & +{\beta }_{8}Sigm{a}_{i,t}+{\beta }_9{\mathit{Turnover}}_{i,t}+{\beta }_{10}{\mathit{Turnover}}_{i,t-1}+{\beta }_{11}Bet{a}_{i,t}+{\beta }_{12}{\mathit{Return}}_{i,t}\hfill \\ & & +{\beta }_{13}{\mathit{Return}}_{i,t-1}+{\beta }_{14}{\mathit{SOE}}_{i,t}+{\varepsilon }_i.\hfill \end{array}$$

The dependent variable Ncskew is the firm's third moment of excess daily stock returns scaled by its cubed standard deviation times minus one. In these estimations, Political is an indicator variable equal to one for the year of a National Congress meeting (first set of estimations) or the year preceding and corresponding to a provincial‐level political promotion event (second set of estimations), zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following the specific political event, zero otherwise. Attribute captures one of two political aspects of the promoted provincial politician: (1) affiliation with the Shanghai political faction and (2) princeling status. In the first estimation, Attribute is an indicator variable equal to one if the newly promoted politician, the provincial party chairman, or provincial governor is a member of the Shanghai political faction. In the second estimation, Attribute is an indicator variable equal to one if the newly promoted politician, the provincial party chairman, or provincial governor is a princeling. All other variables are defined in table 1 and appendix A. t‐statistics derived using clustered standard errors by province are presented in parentheses. Models include annual and provincial fixed effects (Year and Province, respectively; coefficients not reported). The superscripts ***, **, and * indicate that the estimated coefficient is significantly different from zero at the 1%, 5%, and 10% level (two‐tailed test), respectively. Differences in the cumulative pattern of suppression and reversal effects, measured as the sum of Political coefficients (β₁ + β₂) minus the sum of Post‐Political coefficients (β₃ + β₄), conditional upon the realization of Attribute, are tested using an F‐test. N = 12,723.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

4.4. STATE‐CONTROLLED FIRMS VERSUS NON‐STATE‐OWNED FIRMS

In this section, we focus on whether state ownership influences the degree of negative information suppression around political events. Our arguments suggest that both state‐ and non‐state‐owned companies will be affected by these political events. However, these information effects could vary by ownership type if negative news about state‐controlled firms creates incrementally larger costs for the politicians. Additionally, if the mechanism(s) through which suppression occurs differs by firm type, one might observe differences across ownership. For example, if direct control increases the ability of a politician to influence the information dissemination activities of listed firms, suppression should be stronger among state‐controlled companies.19 Yet, if the decentralization of SOE's has been effective, both state‐controlled and non‐state‐owned firms could exhibit very similar behavior around these events. Similarly, if suppression occurs through the local politicians’ supervision of newspapers and other media channels, politicians will be able to affect the information environment of both state and nonstate firms equally.

To test for these differential effects, we estimate pooled cross‐sectional estimations of equation (2) conditional upon firm type. Table 7 presents this evidence. The first and second pairs of columns present select coefficients from estimations focusing on the impact of National Congress meetings and provincial political promotions, respectively, on negative skewness. For each pair of estimations presented, the first column presents coefficients from estimations using our sample of state‐controlled firms; the second column presents coefficients from estimations using our sample of non‐state‐owned firms. Differences between state and nonstate firms’ sensitivity to the political events are tested using a chi‐square statistic. We estimate a separate model for each subsample to ensure that any state/non‐state effects are not the result of different parameter loadings on the determinants of Ncskew_i,t across firm type.

Table 7.

Influence of Political Events on the Incentive to Suppress Negative Financial Information: State‐Controlled Versus Non‐State‐Owned Firms

Political Event:	National Congress		Provincial‐Level Political Promotion
Type of	State‐	Non‐State‐	State‐	Non‐State‐
Firm:	Owned Firms	Owned Firms	Owned Firms	Owned Firms
Politicalt	−0.246***	−0.123***	−0.053*	−0.065**
	(−6.73)	(−2.84)	(−1.80)	(−2.01)
Difference in Politicalt coefficients:	Diff. = −0.123	Diff. = 0.012
	Chi‐square = 4.30**	Chi‐square = 0.10
Post‐Politicalt	0.051*	0.081*	0.124***	0.118***
	(1.88)	(1.99)	(3.54)	(3.13)
Difference in Post‐Politicalt coefficients:	Diff. = 0.030	Diff. = 0.006
	Chi‐square = 0.56	Chi‐square = 0.04
LogSizet	−0.048***	−0.022	−0.059***	−0.035*
	(−5.03)	(−1.361)	(−4.85)	(−1.89)
Growtht	−0.020	−0.023	−0.009	−0.021
	(−1.14)	(−1.368)	(−0.534)	(−1.22)
Sigmat	−8.391***	−12.405***	−19.004***	−19.892***
	(−4.82)	(−2.93)	(−8.20)	(−4.40)
Turnovert	0.932***	0.964***	0.303*	0.602**
	(6.10)	(4.66)	(1.83)	(2.61)
Turnovert− 1	−0.368**	−0.595***	−0.030	−0.543***
	(−2.12)	(−5.90)	(−0.10)	(−3.64)
Betat	−0.275***	−0.257***	−0.316***	−0.249***
	(−6.85)	(−4.81)	(−8.44)	(−4.82)
Returnt	−0.040**	−0.044**	0.027	−0.002
	(−2.05)	(−2.48)	(1.48)	(−0.10)
Returnt− 1	0.092***	0.040***	0.092***	0.034***
	(7.78)	(3.49)	(6.35)	(2.90)
Year Fixed Effects	Included	Included	Included	Included
Provincial Fixed Effects	Included	Included	Included	Included
Adjusted R  2	0.187	0.143	0.140	0.096
Number of Observations	8,608	4,115	8,608	4,115

This table presents coefficients from various pooled, cross‐sectional estimations of the following model:

$$\begin{array}{ccc}\hfill {\mathit{Ncskew}}_{i,t}& =& \alpha +Year+Province+{\beta }_{1}Politica{l}_{i,t}+{\beta }_{2}Post-{\mathit{Political}}_{i,t}+{\beta }_{3}Logsiz{e}_{i,t}+{\beta }_{4}Growt{h}_{i,t}\hfill \\ & & +{\beta }_{5}Sigm{a}_{i,t}+{\beta }_{6}Turnove{r}_{i,t}+{\beta }_{7}Turnove{r}_{i,t-1}+{\beta }_{8}Bet{a}_{i,t}+{\beta }_9{\mathit{Return}}_{i,t}+{\beta }_{10}{\mathit{Return}}_{i,t-1}+{\varepsilon }_{i,t}.\hfill \end{array}$$

The dependent variable Ncskew is the firm's third moment of excess daily stock returns scaled by its cubed standard deviation times minus one. In the first set of estimations, Political is an indicator variable equal to one for the years that a National Congress of the CCP was held, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following each National Congress event, zero otherwise. In the second set of estimations, Political is an indicator variable equal to one for the year preceding and corresponding to a provincial‐level political promotion event, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following the provincial‐level promotion event, zero otherwise. All other variables are defined in table 1 and appendix A. t‐statistics derived using clustered standard errors by province are presented in parentheses. Models include annual and provincial fixed effects (Year and Province, respectively; coefficients not reported). The superscripts ***, **, and * indicate that the estimated coefficient is significantly different from zero at the 1%. 5%, and 10% level (two‐tailed test), respectively. The significance of differences in estimated coefficients on Political and Post‐Political across the state‐firm and non‐state‐firm specifications is tested using chi‐squared tests.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

Focusing on National Congress meetings, both state‐controlled and privately owned listed firms experience a significant reduction in negative stock price crashes during National Congress years relative to non‐congress years. However, comparing across estimations, we find that the promotion period effect is significantly more pronounced among state‐controlled firms, consistent with the politician bearing greater costs for the release of negative news about state firms around these highly visible national events. For both sets of firms, this decline is followed by the predicted increase in negative stock return skewness in the year immediately following the National Congress. Focusing on impending provincial‐level promotion events, the incentives appear to affect both types of firms equally. Specifically, the intertemporal pattern of significantly lower Ncskew_i,t before the promotion, followed by significantly higher Ncskew_i,t after the event, is statistically and economically similar across the two types of firms. Taken together, the results suggest that state ownership heightens incentives to suppress negative news prior to the National Congress meetings, but has no incremental effect on either the subsequent releases of the news after the meeting or bad news suppression and subsequent releases around political promotion. This is consistent with the conjecture that local politicians can, through indirect intervention, exert equally strong pressure on nonstate firms to control the release of bad news, especially around their promotions, and/or nonstate firms have an equally strong incentive as the state firms to engage in strategic news releases because the former need to build political connections with the politicians.

4.5. INFLUENCE OF FOREIGN LISTING CHOICE ON THE INCENTIVE TO SUPPRESS NEGATIVE NEWS

Prior research shows that firms cross‐listing onto exchanges characterized by strong investor protections experience an improvement in corporate governance and financial reporting practices. These bonding‐related effects yield a lower cost of capital and heightened valuations for the cross‐listed firms (Hail and Leuz [2009], Diodge, Karolyi, and Stulz [2004]). The most direct route for Chinese firms to access foreign markets, especially during our sample period, is through an H‐share listing on the Hong Kong Stock Exchange. If Hong Kong institutions create credible reporting incentives, H‐share listings should exert a positive influence on the disclosure practices of listed Chinese firms. However, a Hong Kong listing may fail to produce bonding‐related benefits for Chinese firms. For example, Hung, Wong, and Zhang [2012] find that Chinese state‐owned firms’ overseas listing decisions are primarily determined by politicians’ agendas, with the performance of these overseas listed firms deteriorating following the foreign listing. Absent bonding benefits, a Hong Kong listing would have no effect on reporting incentives around political events. Finally, Chinese firms listed on foreign exchanges are the most prominent and visible Chinese companies. The release of bad news by these firms during one of our political event windows is expected to impose a large, visible cost on the local politician and central government; as such, these Hong Kong listed firms could face strong countervailing incentives to temporarily suppress the release of bad news during politically sensitive event periods.

Table 8 presents coefficients from estimations using two separate samples of firms. The first set of estimations mirror the analysis in table 4 on a small sample of non‐central government‐owned Chinese firms with H‐shares listed in Hong Kong. These estimations document the same pattern of suppression found among mainland firms around these two political events. Second, using our pooled sample of mainland (A‐share) listed Chinese firms, we estimate an empirical model that includes an indicator variable for the presence of a Hong Kong cross‐listing. These estimations reveal that the suppression of bad news around our two political events is not attenuated by the presence of a Hong Kong listing. Instead, contrary to our expectations, the suppression effect is actually incrementally stronger for visible, Hong Kong listed firms domiciled in a province experiencing a political promotion. The heightened level of suppression among these Hong Kong listed firms is consistent with the overseas listings decision of Chinese firms being politically motivated; any bad news of these firms during politically sensitive events will bring greater embarrassment to their affiliated politicians. 20 This evidence casts further doubt on the ability of a Hong Kong listing to fully resolve governance weaknesses among China's listed entities.

Table 8.

Political Incentives to Suppress Negative Information Conditional upon the Presence of a Foreign Stock Exchange Listing

Political Event:	National Congress		Provincial‐Level Political Promotion
	Hong Kong	Hong Kong +	Hong Kong	Hong Kong +
Sample:	Sample	Mainland Sample	Sample	Mainland Sample
Politicalt	−0.855***	−0.228***	−0.285***	−0.065**
	(−2.94)	(−7.03)	(−3.41)	(−2.71)
Politicalt × Hong Kongi,t	−	0.025	−	−0.186*
	−	(0.25)	−	(−1.99)
Post‐Politicalt	−0.373	0.053***	0.287*	0.125***
	(−1.27)	(2.69)	(1.96)	(4.00)
Post‐Politicalt × Hong Kongi,t	−	0.220**	−	0.157*
	−	(2.44)	−	(1.82)
Hong Kongi,t	−	−0.063	−	−0.024
	−	(−0.97)	−	(−0.34)
Control Variables	Included	Included	Included	Included
Year Fixed Effects	Included	Included	Included	Included
Provincial Fixed Effects	Excluded	Included	Excluded	Included
Cumulative (Hong Kong = 0)		−0.281		−0.190
Cumulative (Hong Kong = 1)		−0.476		−0.533
Difference (1 minus 0)		−0.195		−0.343
(F‐statistic)		(2.84)*		(10.97)***
Adjusted R  2	0.225	0.174	0.097	0.125
Number of Observations	273	12,723	273	12,723

This table presents select coefficients from pooled, cross‐sectional estimations of the following model:

$$\begin{array}{ccc}\hfill {\mathit{Ncskew}}_{i,t}& =& \alpha +Year+Province+{\beta }_{1}Politica{l}_{i,t}+{\beta }_2{\mathit{Political}}_{i,t}\times {\mathit{Hong}\mathit{Kong}}_{i,t}+{\beta }_3\mathit{Post}-{\mathit{Political}}_{i,t}+{\beta }_4\mathit{Post}-{\mathit{Political}}_{i,t}\times {\mathit{Hong}\mathit{Kong}}_{i,t}+{\beta }_5{\mathit{Hong}\mathit{Kong}}_{i,t-1}\hfill \\ & & +{\beta }_{6}LogSiz{e}_{i,t}+{\beta }_7{\mathit{Growth}}_{i,t}+{\beta }_{8}Sigm{a}_{i,t}+{\beta }_9{\mathit{Turnover}}_{i,t}+{\beta }_{10}{\mathit{Turnover}}_{i,t-1}+{\beta }_{11}Bet{a}_{i,t}+{\beta }_{12}{\mathit{Return}}_{i,t}+{\beta }_{13}{\mathit{Return}}_{i,t-1}+{\beta }_{14}SO{E}_{i,t}+{\varepsilon }_i.\hfill \end{array}$$

The dependent variable Ncskew is the firm's third moment of excess daily stock returns scaled by its cubed standard deviation times minus one. Political is an indicator variable equal to one if the firm‐year relates to a specific political event; Post‐Political is an indicator variable equal to one in the year immediately following the event. In the first set of estimations, Political is an indicator variable equal to one for the years that a National Congress of the CCP was held, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following each National Congress event, zero otherwise. In the second set of estimations, Political is an indicator variable equal to one for the year preceding and corresponding to a provincial‐level political promotion event, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following the provincial‐level promotion event, zero otherwise. For each event, the model is estimated using two samples: (1) a reduced form of the empirical model is estimated using a small sample of Hong Kong‐listed Chinese firms (first column) and (2) the full model is estimated using a pooled sample of Hong Kong‐listed Chinese firms and our sample of mainland‐listed Chinese firms (second column). In the second estimation, Hong Kong is an indicator variable equal to one if the firm‐year relates to a Hong Kong‐listed Chinese firm, zero otherwise. All other variables are defined in table 1 and appendix A. t‐statistics derived using clustered standard errors by province are presented in parentheses. Models include annual and provincial fixed effects (Year and Province, respectively; coefficients not reported). The superscripts ***, **, and * indicate that the estimated coefficient is significantly different from zero at the 1%, 5%, and 10% level (two‐tailed test), respectively. Differences in the cumulative pattern of suppression and reversal effects, measured as the sum of Political coefficients (β₁ + β₂) minus the sum of Post‐Political coefficients (β₃ + β₄), conditional upon the realization of Hong Kong are tested using an F‐test.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

4.6. INFLUENCE OF CAPITAL MARKET INCENTIVES ON INCENTIVES TO SUPPRESS NEGATIVE INFORMATION

Strong capital market forces have the ability to serve as a constraint on political incentives. To the extent that the capital market and reputational benefits of transparency rely upon a credible commitment to providing information in a timely manner, we would expect firms most exposed to market forces to be less sensitive to transitory political incentives. Stronger capital market settings are also associated with more conduits by which negative information can be gathered, used, and disseminated, thus increasing the difficulty for firms and local politicians to suppress negative information. However, politicians in decentralized, economically well‐developed provinces could be more sensitive to adverse market news about firms under their supervision because the public expects these firms to perform better than those in less‐developed provinces. Then a market orientation may produce a countervailing incentive, which both negates the aforementioned reputation and commitment pressures and intensifies the politicians’ efforts to control the firm's information environment.

To test these premises, we classify firm‐year observations on the basis of two proxies for capital market incentives—the relative importance of equity markets at the provincial level and whether the firm has tradable B‐shares. The first proxy, Equity market development, is measured as the total market capitalization of all listed firms in the province in year t, scaled by provincial gross domestic product. The indicator variable Market Development equals one if the province ranks in the top half of the distribution in year t, zero otherwise. The second proxy, B‐shares, is an indicator variable equal to one if the firm also issued a class of tradable B‐shares. Given that B‐shares are targeted for foreign investors, the existence of B‐share securities increases the firm's exposure to sophisticated, transparency‐demanding investors compared to firms only trading in the A‐share market (i.e., only domestic investors). The existence of B‐share prices also provides additional information signals to A‐share market participants. Contrary to these predictions, we find that neither Market Development nor B‐shares has an impact on Ncskew patterns around our two political events, both before or after conditioning upon the ownership of the company (see online appendix, table E). The lack of variation across these settings suggests that either our empirical proxies failed to capture meaningful variation in market‐based incentives or that market forces are not sufficiently developed in China to create the necessary offsetting incentives during these politically charged periods. It could also be that the increase in market forces that promote transparency in decentralized, well‐developed provinces are counterbalanced by the heighten incentives for politicians to suppress bad news when the public has a higher expectation of its firms to report good performance.

5. Further Analyses and Empirical Extensions

The preceding empirical analyses document the impact that political costs have on the incentives to temporarily suppress negative information about Chinese listed firms. The following sections extend these analyses. Specifically, we examine trends in the likelihood of large, negative monthly price movements in the months surrounding the two political events (section 5.1), the impact of these heightened political incentives on an alternative measure of information flow, stock return synchronicity (section 5.2), and whether political events are associated with a change in the flow of published newspaper articles about Chinese listed firms (section 5.3).

5.1. IMPACT OF POLITICAL EVENTS ON THE SUPPRESSION OF NEGATIVE INFORMATION: SHORT HORIZON TESTS

The consistency of the preceding results with Chen, Hong, and Stein's [2001] theoretical predictions of how stock prices should behave when negative information is temporarily suppressed is striking. Given that the two political events we are examining create ex ante incentives to suppress negative information, observing the “suppress and release” pattern serves as powerful, indirect evidence that such behavior is occurring. That said, from a research design perspective, the use of the negative skewness methodology outlined and used in prior research requires a tradeoff between theory and data—a shift in negative skewness is the theoretical prediction, but skewness is required to be measured over long windows (e.g., annually). These one‐year intervals generate coarse test variables (i.e., Political_i,t and Post‐Political_i,t), potentially limiting our ability to attribute changes in stock price behavior to the hypothesized incentives created by the political event.

Although the preceding negative skewness supports our information suppression hypotheses, we perform a secondary analysis that exploits granular monthly stock return. If these political events are creating heightened incentives to suppress negative information, then the information effects should be stronger in the months closest to the events. Using the following pooled cross‐sectional logistic model, we examine whether the likelihood of a large negative price change is lower (higher) in the months preceding (following) a given political event:

$$\begin{array}{c}\mathrm{Prob}(Cras{h}_{i,t}=1)=\mathrm{Logit}(\alpha +\mathit{Year}+\mathit{Province}+{\beta }_1\mathit{Pre}\hfill \\ -\mathit{Political}{[-6,-4]}_{i,t}+{\beta }_2\mathit{Pre}-\mathit{Political}{[-3,0]}_{i,t}\hfill \\ +{\beta }_3\mathit{Post}-\mathit{Political}{[1,3]}_{i,t}+{\beta }_{4}Post\hfill \\ -\mathit{Political}{[4,6]}_{i,t}+{\beta }_5{\mathit{Logsize}}_{i,t}+{\beta }_6{\mathit{Growth}}_{i,t}\hfill \\ +{\beta }_{7}Sigm{a}_{i,t}+{\beta }_8{\mathit{Turnover}}_{i,t}+{\beta }_9{\mathit{Turnover}}_{i,t-1}\hfill \\ +{\beta }_{10}Bet{a}_{i,t}+{\beta }_{11}{\mathit{Return}}_{i,t-1}+{\beta }_{12}{\mathit{SOE}}_{i,t-1}\hfill \\ +{\varepsilon }_{i,t}).\hfill \end{array}$$ (5)

In this model, the dependent variable Crash_i,t is an indicatory variable equal to one if the firm's market‐adjusted stock return is less than negative 20% in month t, zero otherwise. We test changes in the likelihood of large negative price movements by creating indicator variables meant to capture the firm‐month's proximity to the political event. As such, Pre‐Political_i,t is an indicator variable equal to one if the firm‐month precedes a specific political event within the indicated time range; zero otherwise. Analogously, Post‐Political_i,t is an indicator variable equal to one if the firm‐month follows the political event over the indicated time frame, zero otherwise. The remaining independent variables correspond to the relevant determinants of negative stock return skewness included in equation (2). This analysis is performed using 162,582 available firm‐month observations during our sample period.

Table 9 presents select coefficients (and t‐statistics using standard errors clustered by province) from various estimations of equation (5). In the first set of estimations, Pre‐Political_i,t and Post‐Political_i,t relate to the months preceding and following the month that the National Congress of the CCP was held. In the second set of estimations, Pre‐Political_i,t and Post‐Political_i,t relate to the months surrounding a provincial‐level political promotion event. These estimations reveal that the likelihood of large negative price movements is significantly lower in the three‐month period preceding and the month including the political event. Moreover, we observe that the reduced likelihood of negative news around the National Congress persists for a few months after the actual meeting. Because key policy decisions and high‐level political appointments are announced and scrutinized in the months immediately following a National Congress, the continued suppression of negative news during this politically sensitive time period is intuitive and reflects prevailing incentives in this context. In contrast, the reduction in negative news events in advance of a political promotion is concentrated in the months directly preceding the event (with the effect strongest in the three months immediately preceding the event), with the likelihood of large negative return shocks rebounding to baseline levels after the event. Once the promotion is publicly announced and implemented, the incentive to suppress in that province abates.

Table 9.

Monthly Analysis of the Likelihood of Stock Price Crash Around Political Events

	National		Provincial‐Level
Political Event:	Congress		Political Promotion
	(1)	(2)	(1)	(2)
Pre‐Political [‒6,‒4]	0.170*	0.171*	−0.068	−0.069
	(1.66)	(1.66)	(−0.72)	(−0.73)
Pre‐Political [−3,−0]	−0.468***	−0.469***	−0.438***	−0.439***
	(−3.78)	(−3.78)	(−3.98)	(−3.99)
Post‐Political [1,3]	−0.461***	−0.463***	−0.015	−0.016
	(−2.82)	(−2.84)	(−0.14)	(−0.15)
Post‐Political [4,6]	0.358***	0.356***	−0.140	−0.142
	(3.15)	(3.14)	(−1.27)	(−1.29)
LogSizet	0.043	0.049*	0.050*	0.056*
	(1.46)	(1.65)	(1.73)	(1.92)
Growtht	−0.330***	−0.328***	−0.323***	−0.322***
	(−5.73)	(−5.73)	(−5.65)	(−5.65)
Sigmat	31.755***	31.671***	31.120***	31.037***
	(9.84)	(9.81)	(9.68)	(9.65)
Turnovert	0.820***	0.820***	0.896***	0.897***
	(12.17)	(12.17)	(13.75)	(13.76)
Turnovert− 1	−0.060	−0.062	−0.038	−0.040
	(−0.94)	(−0.96)	(−0.62)	(−0.64)
Betat	−1.878***	−1.864***	−1.901***	−1.887***
	(−17.37)	(−17.10)	(−17.68)	(−17.39)
Returnt− 1	2.458***	2.455***	2.393***	2.390***
	(12.94)	(12.93)	(12.78)	(12.77)
SOE		−0.064		−0.065
		(−1.22)		(−1.22)
Year Fixed Effects	Included	Included	Included	Included
Provincial Fixed Effects	Included	Included	Included	Included
Pseudo R  2	0.147	0.147	0.146	0.146

This table presents coefficients from various pooled, cross‐sectional estimations of the following model:

$$\begin{array}{ccc}\hfill \mathrm{Prob}({\mathit{Crash}}_{i,t}=1)& =& \mathrm{Logit}(\alpha +Year+\mathit{Province}+{\beta }_1\mathit{Pre}-\mathit{Political}{[-6,-4]}_{i,t}\hfill \\ & & +{\beta }_{2}Pre-\mathit{Political}{[-3,0]}_{i,t}+{\beta }_3\mathit{Post}-\mathit{Political}{[1,3]}_{i,t}+{\beta }_4\mathit{Post}-\mathit{Political}{[4,6]}_{i,t}\hfill \\ & & +{\beta }_5{\mathit{Logsize}}_{i,t}+{\beta }_6{\mathit{Growth}}_{i,t}+{\beta }_{7}Sigm{a}_{i,t}+{\beta }_8{\mathit{Turnover}}_{i,t}+{\beta }_9{\mathit{Turnover}}_{i,t-1}\hfill \\ & & +{\beta }_{10}Bet{a}_{i,t}+{\beta }_{11}{\mathit{Return}}_{i,t-1}+{\beta }_{12}{\mathit{SOE}}_{i,t}+{\alpha }_{i,t}).\hfill \end{array}$$

The dependent variable Crash_i,t is an indicator variable equal to one if the firm's monthly market‐adjusted stock return in month t is less than ‒20%, zero otherwise. Pre‐Political is an indicator variable equal to one if the firm‐month precedes a specific political event over the indicated range; zero otherwise. Post‐Political is an indicator variable equal to one if the firm‐month follows the political event over the indicated range, zero otherwise. In the first set of estimations, Pre‐Political and Post‐Political relate to the months preceding and following the month that the National Congress of the CCP was held. In the second set of estimations, Pre‐Political and Post‐Political relate to the months surrounding a provincial‐level political promotion event. All other variables are defined in table 1 and appendix A. t‐statistics derived using clustered standard errors by province are presented in parentheses. Models include annual and provincial fixed effects (Year and Province, respectively; coefficients not reported). The superscripts ***, **, and * indicate that the estimated coefficient is significantly different from zero at the 1%, 5%, and 10% level (two‐tailed test), respectively. N = 138,472.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

5.2. POLITICAL EVENTS AND STOCK PRICE SYNCHRONICITY

Morck, Yeung, and Yu [2000] and Gul, Kim, and Qiu [2010] document that stock price movements are highly synchronous in China; this strong comovement is attributed to a limited flow of firm‐specific information into prices in this marketplace.21 Building on this interpretation of stock return synchronicity, we examine whether the overall flow of firm‐specific information into prices is affected by our two political events. Specifically, we estimate variations of the following cross‐sectional models:

$$\begin{array}{ccc}\hfill Synchronicit{y}_{i,t}& =& \alpha +Year+\mathit{Province}+{\beta }_1{\mathit{Political}}_{i,t}+{\beta }_2\mathit{Post}\hfill \\ & & -{\mathit{Political}}_{i,t}+{\beta }_3{\mathit{LogSize}}_{i,t}+{\beta }_4{\mathit{Growth}}_{i,t}\hfill \\ & & +{\beta }_{5}Std\_RO{A}_{i,t}+{\beta }_{6}Sigm{a}_{i,t}+{\beta }_7{\mathit{Turnover}}_{i,t}\hfill \\ & & +{\beta }_{8}Regulate{d}_{i,t-1}+{\beta }_{9}Industry\_Nu{m}_{i,t}\hfill \\ & & +{\beta }_{10}Industry\_Siz{e}_{i,t}+{\beta }_{11}SO{E}_{i,t}+{\varepsilon }_{i,t}.\hfill \end{array}$$ (6)

In these estimations, Synchronicity_i,t is measured as the log (R ²/(1+R ²)), where R ² is from an estimation of the market model of firm weekly returns against current and lagged domestic market and U.S. market returns in year t. Control variables are drawn from prior research and defined in appendix A.22 To the extent that political events are associated with a broad, temporary reduction in the amount of firm‐specific information influencing security prices, we expect a positive association between Political_i,t and Synchronicity_i,t and a negative relation between Post‐Political_i,t and Synchronicity_i,t.

Table 10 presents coefficients from various estimations of equation (6). These estimations reveal that Chinese listed firms experience a significant increase in stock return synchronicity around National Congress meetings and provincial‐level political promotion events, consistent with a temporary reduction in the total amount of firm‐specific information being released during this event window. Moreover, similar to the subsequent release of suppressed negative information, we observe an increase in firm‐specific information after the event. The combined evidence in table 4 through table 10 provides consistent, price‐based evidence on the impact that political forces have on the transparency of listed firms in China.

Table 10.

Impact of Political Incentives on Stock Return Synchronicity

	National		Provincial‐Level
Political Event:	Congress		Political Promotion
	(1)	(2)	(1)	(2)
Political	0.533***	0.522***	0.059*	0.059*
	(26.15)	(25.60)	(1.93)	(1.91)
Post‐Political	−0.081***	−0.113***	−0.038*	−0.037*
	(−3.59)	(−5.33)	(−1.86)	(−1.79)
LogSize	−0.036***	−0.040***	0.005	−0.001
	(−3.82)	(−4.26)	(0.51)	(−0.12)
Growth	−0.036**	−0.032*	−0.036**	−0.033*
	(−2.06)	(−1.89)	(−2.08)	(−1.93)
Std_ROA	−2.491***	−2.370***	−2.287***	−2.177***
	(−11.39)	(−10.77)	(−11.10)	(−10.43)
Sigma	−84.423***	−84.043***	−87.665***	−87.084***
	(−66.64)	(−65.45)	(−54.76)	(−53.39)
Turnover	1.982***	2.033***	2.564***	2.591***
	(16.87)	(16.79)	(20.62)	(20.16)
Regulated	0.067**	0.058**	0.069**	0.061*
	(2.61)	(2.21)	(2.34)	(1.99)
Industry_Num	−0.010	−0.008	0.013	0.016
	(−0.59)	(−0.46)	(0.75)	(0.92)
Industry_Size	0.015	0.014	−0.013	−0.015
	(0.78)	(0.76)	(−0.69)	(−0.84)
SOE	−	0.094***	−	0.094***
	−	(4.31)	−	(4.19)
Year Fixed Effects	Included	Included	Included	Included
Provincial Fixed Effects	Included	Included	Included	Included
Adjusted R  2	0.665	0.668	0.639	0.643

This table presents coefficients from various pooled, cross‐sectional estimations of the following model:

$$\begin{array}{ccc}\hfill {\mathit{Synchronicity}}_{i,t}& =& \alpha +Year+{\beta }_{1}Politica{l}_{i,t}+{\beta }_{2}Post-{\mathit{Political}}_{i,t}+{\beta }_{3}LogSiz{e}_{i,t}+{\beta }_{4}Growt{h}_{i,t}\hfill \\ & & +{\beta }_{5}Std\_RO{A}_{i,t}+{\beta }_{6}Sigm{a}_{i,t}+{\beta }_{7}Turnove{r}_{i,t}+{\beta }_{8}Regulate{d}_{i,t-1}\hfill \\ & & +{\beta }_{9}Industry\_Nu{m}_{i,t}+{\beta }_{10}Industry\_Siz{e}_{i,t}+{\beta }_{11}SO{E}_{i,t}+{\varepsilon }_{i,t},\hfill \end{array}$$

where Synchronicity is log(R ²/(1‒R ²)), where R ² is from estimation of market model of firm weekly return against current and one week lagged domestic market return and U.S. market return. Political is an indicator variable equal to one if the firm‐year relates to a specific political event; Post‐Political is an indicator variable equal to one in the year immediately following the event. In the first set of estimations, Political is an indicator variable equal to one for the years that a National Congress of the CCP was held, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following each National Congress event, zero otherwise. In the second set of estimations, Political is an indicator variable equal to one for the year preceding and corresponding to a provincial‐level political promotion event, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following the provincial‐level promotion event, zero otherwise. All other variables are defined in table 1 and appendix A. t‐statistics derived using clustered standard errors by province are presented in parentheses. Models include annual fixed effects (coefficients not reported). The superscripts ***, **, and * indicate that the estimated coefficient is significantly different from zero at the 1%, 5%, and 10% level (two‐tailed test), respectively. N = 7,786.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

5.3. POLITICAL EVENTS AND MEDIA COVERAGE: SHIFTS IN THE FLOW OF FINANCIAL NEWSPAPER ARTICLES

Our analyses using stock‐priced‐based measures of information flow, including both stock price crash metrics and stock return synchronicity, document patterns consistent with a shift in the information environment of listed firms around our political events. These patterns vary by political attributes correlated with incentives to suppress negative information, suggesting that the observed return behavior is driven by a shift in the information production and dissemination environment of these firms. However, as noted in the Introduction of the paper, the preceding analyses do not directly document the channels and methods by which the information environment, and hence stock price behavior, changes.

To shed preliminary evidence on this issue, we examine the flow of articles written about these firms and published in official (i.e., party papers) and professional (i.e., financial) newspapers in China around our two political events. The documentation of a shift in the flow of financial news articles would provide compelling evidence that the observed changes in stock price behavior are somehow linked to changes in these firms' information environment. The examination of newspaper articles is an ideal first attempt to address this issue, as these published articles represent the net output of both information production and dissemination activities, with the resultant flow of articles reflecting the incentives and decisions of the managers (i.e., disclosure choices), politicians (direct censorship of the firm and media), and the media (i.e., self‐censoring and commercialization).

To perform this test, we gather a comprehensive sample of 1,955,383 news articles written about our listed firms in Chinese newspapers over the period 1998 to 2010.23 The sample is split into two types of newspaper articles—those published in official newspapers, and those published in professional newspapers. Official newspapers, as the propaganda mouthpiece of the CCP, are expected to be strongly influenced by prevailing political incentives. Professional newspapers, in contrast, are expected to respond to conflicting incentives arising from both market readership demands and political incentives arising from licensing and political ownership arrangements.

For each firm‐year observation, we define Articles_i,t as the total number of newspaper articles about the firm published in either official or financial newspapers. Using these data, we estimate the following pooled cross‐sectional model for each political event published in each type of newspaper:

$$\begin{array}{c}\mathrm{Log}(1+{\mathit{Articles}}_{i,t}\left)\mathrm{or}\mathrm{Log}\right(1+{\mathit{Articles}}_{i,t}/Siz{e}_{i,t})\hfill \\ =\alpha +Year+\mathit{Province}{+\beta }_1{\mathit{Political}}_{i,t}{+\beta }_{2}Post-{\mathit{Political}}_{i,t}\hfill \\ {+\beta }_3{\mathit{Logsize}}_{i,t}{+\beta }_{4}Growt{h}_{i,t}{+\beta }_{5}Leverag{e}_{i,t-1}\hfill \\ {+\beta }_6{\mathit{Return}}_{i,t}{+\beta }_{7}SO{E}_{i,t}{+\varepsilon }_{i,t}.\hfill \end{array}$$ (7)

In these estimations, we utilize two different measures of newspaper activity: a logarithmic transformation of the number of articles written about firm (log(1+Articles_i,t)) and a logarithmic transformation of the number of articles scaled by the market value of equity (in billions of RMB).

Table 11 presents coefficients from these estimations. Consistent with our information suppression arguments, we observe that the average number of newspaper articles written about China's listed firms is significantly lower during our political event windows. The effect exists across both types of newspapers and using both measures of newspaper coverage. Interestingly, the effect appears to be incrementally stronger around National Congress meetings (vis‐à‐vis political promotions), which is also consistent with the observed relative magnitude of stock price behavior around these two events. Finally, we observe that the rate of published news articles immediately returns to baseline levels following the provincial political promotions, while the year after National Congress meetings actually experiences a significant increase in newspaper articles.24

Table 11.

Frequency of Firm‐Specific News Articles Around Political Events

	National				Provincial‐Level
Political Event	Congress				Political Promotion
	Official		Professional		Official		Professional
Type of Newspaper:	Newspaper		Newspaper		Newspaper		Newspaper
Dependent Variable:	Articlesi,t	Articlesi,t/Sizei,t	Articlesi,t	Articlesi,t/Sizei,t	Articlesi,t	Articlesi,t/Sizei,t	Articlesi,t	Articlesi,t/Sizei,t
Political	‒0.442***	‒0.325***	‒0.463***	‒0.503***	‒0.091***	‒0.079***	‒0.040**	‒0.036**
	(‒11.52)	(‒13.27)	(‒14.64)	(‒17.31)	(‒2.96)	(‒3.68)	(‒2.20)	(‒2.12)
Post‐Political	0.415***	0.332***	0.167***	0.144***	0.031	0.029	‒0.013	‒0.017
	(13.74)	(11.09)	(6.57)	(5.12)	(0.99)	(1.15)	(‒0.82)	(‒1.07)
LogSizet	0.603***	‒0.119***	0.626***	‒0.276***	0.612***	‒0.112***	0.630***	‒0.273***
	(22.91)	(‒6.41)	(29.48)	(‒13.03)	(22.62)	(‒5.88)	(29.25)	(‒12.84)
Growtht	0.012	0.020	‒0.052*	‒0.036	0.005	0.014	‒0.054*	‒0.038
	(0.24)	(0.47)	(‒1.80)	(‒1.14)	(0.09)	(0.33)	(‒1.88)	(‒1.21)
Leveraget	‒0.005	‒0.016	‒0.028	‒0.017	‒0.019	‒0.028*	‒0.034*	‒0.022
	(‒0.24)	(‒1.16)	(‒1.59)	(‒1.16)	(‒0.97)	(‒1.98)	(‒1.93)	(‒1.51)
Returnt	0.000	‒0.000	‒0.001***	‒0.001***	0.000	‒0.000	‒0.001***	‒0.001***
	(0.56)	(‒0.95)	(‒6.36)	(‒8.12)	(0.09)	(‒1.39)	(‒6.59)	(‒8.86)
SOE	‒0.044**	‒0.028**	0.024	0.013	‒0.035*	‒0.021	0.029	0.017
	(‒2.27)	(‒2.08)	(1.28)	(0.69)	(‒1.76)	(‒1.53)	(1.47)	(0.88)
Year Fixed Effects	Included	Included	Included	Included	Included	Included	Included	Included
Provincial Fixed Effects	Included	Included	Included	Included	Included	Included	Included	Included
Adjusted R  2	0.475	0.282	0.423	0.387	0.468	0.275	0.422	0.386

This table presents coefficients from various pooled, cross‐sectional estimations of the following model:

$$\begin{array}{ccc}\hfill \mathrm{Log}(1+Article{s}_{i,t})\mathrm{or}\mathrm{Log}(1+Article{s}_{i,t}/Siz{e}_{i,t})& =& \alpha +Year+Province+{\beta }_{1}Politica{l}_{i,t}+{\beta }_{2}Post-\mathit{Political}{}_{i,t}+{\beta }_{3}Logsiz{e}_{i,t}\hfill \\ & & +{\beta }_{4}Growt{h}_{i,t}+{\beta }_{5}Leverag{e}_{i,t-1}+{\beta }_6{\mathit{Return}}_{i,t}+{\beta }_{7}SO{E}_{i,t}+{\varepsilon }_{i,t}.\hfill \end{array}$$

The dependent variable, Articles_i,t, is one of two measures capturing the number of news articles published about firm i in year t in either official or professional newspapers. Log(1+Articles_i,t) is a logarithmic transformation of the number of articles published. Log(1+Articles_i,t/Size_i,t) is a logarithmic transformation of the number of articles published scaled by the market value of the firm (in RMB billions). Political is an indicator variable equal to one if the firm‐year relates to a specific political event; Post‐Political is an indicator variable equal to one in the year immediately following the event. In the first panel, Political is an indicator variable equal to one for the years that a National Congress of the CCP was held, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following each National Congress event, zero otherwise. In the second panel, Political is an indicator variable equal to one for the year preceding and corresponding to a provincial‐level political promotion event, zero otherwise; Post‐Political is an indicator variable equal to one for the year directly following the provincial‐level promotion event, zero otherwise. For each political event, the first (second) set of estimations reflects articles published in official (professional/financial) newspapers. All other variables are defined in table 1 and appendix A. t‐statistics derived using clustered standard errors by province are presented in parentheses. Models include annual and provincial fixed effects (Year and Province, respectively; coefficients not reported). The superscripts ***, **, and * indicate that the estimated coefficient is significantly different from zero at the 1%, 5%, and 10% level (two‐tailed test), respectively. N = 10,856.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

6. Robustness Tests and Alternative Interpretations

6.1. ROBUSTNESS TESTS: ALTERNATIVE MEASURES OF STOCK PRICE CRASHES

The preceding evidence relies primarily upon our dependent variable Ncskew_i,t, which reflects the negative skewness of each firm's daily residual return distribution. However, alternative measures of a firm's stock price crash tendencies exist. First, Chen, Hong, and Stein [2001] utilize a second crash statistic, Duvol_i,t that captures the “down‐to‐up volatility” of the individual stock. Duvol_i,t is measured as the log of the ratio of the standard deviation of residual returns on “down” days to the log of standard deviation of residual returns on “up” days, where down and up days are defined relative to the firm's mean daily residual return realization over the measurement period. Firms with higher level of Duvol_i,t are interpreted as being more crash prone. Replicating our principle analyses (i.e., table 4) using Duvol_i,t in lieu of Ncskew_i,t yields similar results (see online appendix, table 4C).25

Second, in table 9, we utilize a logistic regression specification to document changes in the relative likelihood of large negative stock price movements around our political events. This analysis, relying on granular monthly return data, quarterly political event windows, and firm‐month observations, yields suppression inferences consistent with our annual analysis using Ncskew.

Third, we measure the relative frequency that the firm experiences material negative weekly stock returns, and examine whether this frequency changes around our political events. Specifically, we define Fraction_i,t as the percentage of calendar weeks that the firm experiences a large, negative stock price drop during the fiscal year. A large negative stock price drop is defined as a negative weekly excess return more than 1.96 standard deviations below the sample mean or a price decline greater than minus 20%. In these estimations, firms with a higher level of Fraction_i,t are interpreted as being more crash prone. Pooled estimations using both estimates of Fraction_i,t as the dependent variable confirm the basic inferences obtained from table 4 (see online appendix, table 4D).

Fourth, we create an alternative measure of skewness based upon the 95th median and 5th percentile return realizations of the firm in year t. Specifically, we measure the ratio of the difference between the firm's 95th percentile and median return realization and the difference between the firm's median and 5th percentile realization. Pooled estimations using this nonparametric version of skewness as the dependent variable confirm the basic inferences obtained in table 4 (see online appendix).26

Together, these alternative methodologies confirm the primary intertemporal pattern of fewer stock prices crashes before our political events and more stock price crashes after the events. The robustness of these patterns suggests that our Ncskew results are not an artifact of measurement error or omitted determinants of negative skewness, but instead reflect the transitory suppression of negative information.

6.2. TEST OF ALTERNATIVE INTERPRETATION: SHIFT IN GOOD NEWS DISSEMINATION AROUND POLITICAL EVENTS

The political events we examine in this paper are expected to induce an asymmetric response on the part of politicians and affiliated firms. The presence of the political event raises the cost of reporting bad news; however, the event could also alter the incentive to release good news. In general, politicians and affiliated firms are expected to have been disclosing good news in a timely manner; such timely disclosure is inherently in line with their incentives. If politicians always report good news in a timely fashion, one would expect no incremental effect on good news disclosures during event periods, absent the production of new information by the politician. In contrast, politicians and affiliated firms can actively choose to withhold negative information as it arises (i.e., reduce the timeliness of reporting bad news over the event period). This is the asymmetric effect we predict, and our research setting and design are geared for identifying this negative information effect in security prices.

However, if political incentives affect the release of good news, the affected politicians and their affiliated firms could temporarily defer the release of good news in order to shift the timing of its arrival closer to the political event. Specifically, the frequency with which firms experience large positive stock returns (i.e., stock price jump) will therefore decrease in the year prior to the political events and increase during the political events. Such behavior is reasonable if the expected benefits of releasing good news are higher in the political event period than the preceding period. Moreover, given the planned and highly anticipated nature of the two political events in our study, affected agents will certainly have an opportunity to strategically time the release of good news through such a deferral mechanism. To test this good news deferral conjecture, we measure the percentage of weeks in a given year that the firm experiences a large price increase. A positive weekly excess return more than 1.96 standard deviations above the sample mean or a price increase greater than 20% is considered a large stock price increase. The empirical results do not support this good news deferral conjecture (see online appendix, tables A, B, and C). We are neither able to document a significantly lower frequency of price jumps in the one‐year period preceding the start of our respective political events, nor do we find any heightened incidence of stock price jumps during the two political event windows. 27 There is also no evidence that Ncskew_i,t is significantly higher in this preevent period, as would occur if good news were withheld. In addition, the failure to document an increase in stock price jump frequency during the political event windows casts doubt on the explanation that an increase in good news production and/or dissemination is responsible for the empirical stock price crash results documented in this paper.

6.3. TEST OF ALTERNATIVE INTERPRETATION: SHIFT IN REAL ECONOMIC PERFORMANCE

As an information flow metric, Ncskew_i,t reflects the arrival of new information into the price formation process. Such an arrival can reflect the timely dissemination of news about contemporaneous economic activity or the delayed arrival of news about prior economic events temporarily suppressed due to prevailing incentives. Under this interpretation, the significant decline in Ncskew_i,t around political events, and subsequent increase in Ncskew_i,t following the event, could reflect transitory strong economic performance during those political periods, followed by a reversion back to historical levels after the event. Such a spike in economic performance would be consistent with local politicians handing out positive NPV investment projects, new contracts, or other economic benefits to favored firms in order to improve the perceived performance of their locality in advance of the political event.

We search for this effect in our data by examining the level of firm performance (return on assets) and investment rates around these political events. The trends in ROA and investment rates do not support a simple, transitory politically created benefit argument for our Ncskew_i,t results (see online appendix, table D). For our sample, we observe that ROA and investment rates are not temporarily elevated in advance of the political events; instead, ROA remains constant and investment rates frequently decline in the year before the political events. These preevent trends document that affected firms were not experiencing heightened economic performance in advance of the event, consistent with the good news analyses discussed in the preceding section. Additionally, we document a temporary, postevent decline in ROA after National Congress meetings that is consistent with accounting information (which typically lags prices) reflecting bad news that was likely suppressed in advance of the event and/or the unraveling of propping activities around the political event.

6.4. CORRELATION IN CHANGES IN NEGATIVE SKEWNESS AROUND POLITICAL EVENTS

To the extent that the observed patterns in Ncskew are driven by the temporary suppression of negative information, the increase in stock price crash behavior observed after the event should be inversely related to (i.e., negatively correlated with) the decreased frequency of stock price crashes observed before the event. To test for this association, we compare the correlation in changes in Ncskew around these political events (i.e., correlation between ∆Ncskew_{i,Post‐Political} and ∆Ncskew_i,Political) against the baseline level of correlation between ∆Ncskew_i,t and lagged ∆Ncskew_i,t‐1 realizations outside these event windows. We find that increases in negative skewness following the political event are significantly negatively correlated with changes in negative skewness before the event; this autocorrelation in ∆Ncskew realizations around National Congress meetings and provincial promotions is ‒0.460 and ‒0.443, respectively. More importantly, these negative correlations are both significantly larger in magnitude than the baseline level of negative correlation observed outside these political windows (nonpolitical correlation = ‒0.376).28 This evidence is consistent with an observed jump in negative skewness after the political event being related to the decline in negative skewness preceding the event, thus reinforcing the information‐based interpretation of our main results.

6.5. ROBUSTNESS OF PRIMARY EMPIRICAL SPECIFICATION

To alleviate concerns about misspecifications of our primary empirical model (equation (2)), we have reestimated our results using several alternative models. We have reestimated table 4 without any control variables except for annual and provincial fixed effects; using only lagged versions of our control variables; using lagged versions of our control variables plus alternative measures of contemporaneous firm performance (market‐adjusted returns and return on assets); using logged returns to estimate Ncskew_i,t; after eliminating our two risk measures (Beta_i,t and Sigma_i,t) from the model; replacing sales growth with the firm's market‐to‐book ratio; and replacing provincial fixed effects with firm fixed effects. We continue to find that our political events retain their documented significant associations with Ncskew_i,t in all of these specifications (see online appendix, tables 4F, 4G, and 4H).

7. Conclusion

This paper examines the impact that political forces have on the flow of negative information into stock prices. Using a sample of listed Chinese firms, we test the proposition that politicians and politically connected managers have an incentive to temporarily suppress the flow of negative information around two political events—meetings of the National Congresses of the CCP and promotions of high‐level provincial politicians—that asymmetrically raise the cost of releasing such information. Our evidence supports this proposition. First, we document a significant and theoretically predicted shift in security price behavior around these political events, consistent with politicians and affiliated firms temporarily suppressing negative news. Second, we document that this pattern is incrementally stronger in settings where the political incentive to restrict such information is greater. Finally, we supplement this evidence with a direct examination of one aspect of the firm's information environment—the publication of newspaper articles—and find that the supply of firm‐specific information temporarily declines during these event periods.

Together, these results highlight the important role that political factors play in shaping the information environment of listed firms in highly political environments. We again add the caveat that our research design cannot identify the exact mechanism(s) or channel(s) through which this effect occurs. In spite of these limitations, the failure to document these theoretically predicted stock return patterns would have cast considerable doubt on our information suppression hypotheses. We view the paper as serving an important first step in documenting how the short‐term incentives of politicians and politically connected managers influence the information environment of emerging market firms. We encourage future research to exploit similar political settings to contextually examining the channels and mechanisms used by politicians and affiliated firms to temporarily suppress the flow of negative information into stock prices. Moreover, the results in the paper raise numerous questions about price formation in these settings. How does the market actually price politically driven information risk? Is this risk premium applied at the market level, or does the premium vary over time and/or based upon the political characteristics of the firm? How does the market interpret silence during these politicized periods, and what is the expected impact on realized returns over these windows? Do investors actively search for negative information during these political periods, and, if they do, what mechanisms are preventing the impounding of this information into prices before the political event?

Appendix A.

Variable Definitions

Variable	Description	Sources
Ncskew	This is the negative coefficient of skewness of daily excess return. It is defined as the negative of third moments of daily excess return scaled by cubed standard deviation of daily excess return in the year.	CSMAR
Crash	An indicator variable equal to one if the firm's monthly market adjusted stock return is less than ‒20%, zero otherwise.	CSMAR
Sychronicity	This is the synchronicity of the daily stock return with daily market return. It is defined as the log(R  2/(1‒R  2)), where R  2 is from the market model of firms daily return against current and one week lagged domestic weekly market return and U.S. weekly market return.	CSMAR
National Congress	This is the year of the National Congress of the Chinese Communist Party (CCP), which was held once every five years. The congress was held in 1997, 2002, and 2007 during our sample period.	Media Search
Provincial‐Level Politician Promotion	This variable captures the turnover of provincial party chairman or provincial governors. The turnover is defined as a promotion when he moves to a more senior position than his original one, which includes: (1) promotion within the same province (e.g., promotion of governor to party secretary), (2) promotion to another province (e.g., governor or party secretary position of a larger province), and (3) promotion to the minister level of the central government. Firms that operate in the same province as the promoted politician are affected by politician turnover; the promotion period includes the year before and year of the promotion.	Chinese Personnel Database; “China VIP” from China Information Bank; Web‐based searches
Political	An indicator variable equal to one if the firm‐year relates to a specific political event (i.e., National Congress or Provincial‐Level Political Promotion), zero otherwise.	Self‐constructed
Post‐Political	An indicator variable equal to one in the year immediately following the political event, zero otherwise.	Self‐constructed
LogSize	Log of total market value of equity at the end of the fiscal year.	CSMAR
Growth	Sales growth, defined as the log(salet/salet ‒1).	CSMAR
Sigma	Standard deviation of daily excess return in the year.	CSMAR
Turnover	Average of weekly turnover of the firm's shares, which is defined as the total value traded of tradable shares scaled by total value of tradable shares at the end of the week.	CSMAR
Turnovert‒ 1	One year lagged turnover.	CSMAR
Beta	The firm's beta, estimated using daily returns during the fiscal year.	CSMAR
Return	Annual market‐adjusted stock return of the firm.	CSMAR
Returnt‒ 1	One year lagged RET.	CSMAR
SOE	An indicator variable equal to one if the firm is state‐controlled. A firm is characterized as state‐controlled if the local government owns more than 20% of the outstanding shares of the company.	Annual Report
Std_ROA	The standard deviation of ROA in the past five years including current year with availability of at least three years data.	CSMAR
Regulated	An indicator variable equal to one for companies operating in regulated industry, zero otherwise.	CSMAR
Industry_NUM	The log of total number of listed companies in the two‐digit SIC industry at the fiscal year end.	CSMAR
Industry_SIZE	The log of total assets of all listed companies in the two‐digit SIC industry at the fiscal year end.	CSMAR
Executive Connection	An indicator variable equal to one if the firm's CEO has past or concurrent work experience in the government or political appointment, such as the People's Representative or Member of Chinese People's Political Consultative Conference.	Annual Report
Government Subsidy	An indicator variable equal to one if the firm received a government subsidy in the current year, zero otherwise.	CSMAR
State Bank Loan	An indicator variable equal to one if the firm's balance sheet reflects a long‐term loan from a state bank in the current year, zero otherwise.	CSMAR
Shanghai Faction	An indicator variable equal to one if the promoted politician, or, in provinces without a promotion, either the provincial governor or party chairman, is a member of the Shanghai political faction, zero otherwise. The politician is a member of the Shanghai political faction if he has an overlap in his work experience with Jiang Zemin before his appointment as the president of China.	China Information Bank; Chinese Wikipedia; Web‐based search
Princeling	An indicator variable equal to one if the promoted politician, or, in provinces without a promotion, either the provincial governor or party chairman, is characterized as a princeling, zero otherwise. A politician is characterized as a princeling if he is the son, daughter, or grandchild of a founding member of the CCP.	China Information Bank; Chinese Wikipedia; Web‐based search
Hong Kong	An indicator variable equal to one if the firm's equity is listed on the Hong Kong Stock Exchange (i.e., H‐Share), zero otherwise.	CSMAR
Articles	The firm's annual number of newspaper articles published in leading Chinese newspapers (either official or financial newspapers).	Wisenews; China Financial Newspaper Searching System

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.

Appendix B.

Empirical Timeline

This appendix presents a timeline representation of our primary empirical research design. Panel A (B) presents the timeline for National Congress meetings of the CCP (provincial‐level political promotions). In both settings, the indicator variable Political equals one if the firm‐year observations corresponds with the affected timing of the respective political event, zero otherwise. The indicator variable Post‐Political equals one if the firm‐year corresponds to the year following the political event. Firm‐years unaffected by these event and postevent periods serve as the baseline by which event and postevent stock price behavior is benchmarked.

This article is being made freely available through PubMed Central as part of the COVID-19 public health emergency response. It can be used for unrestricted research re-use and analysis in any form or by any means with acknowledgement of the original source, for the duration of the public health emergency.