1 Introduction

The initial public offering (IPO) is considered a very special period when a firm is transferred from private to public ownership. Prior studies suggest that IPO firms have strong incentives to engage in earnings management (Teoh et al., 1998). For example, managers tend to cook the book to avoid future litigation risks due to abnormal stock price declines after the listing (Teoh et al., 1998). Some scholars argue that firms are more likely to engage in real earnings management (REM) relative to accrual earnings management (AEM) because REM is harder to detect (Cohen et al., 2008; Graham et al., 2005; Roychowdhury, 2006) and easier to conduct throughout the year. As the balance sheet accumulates all the previous changes in accounting methods (Barton & Simko, 2002), firms that undertook extensive AEM in previous years potentially use REM in the current period (Barton & Simko, 2002; Gunny, 2010). Therefore, firms may be more willing to engage in REM in the IPO year.

Prior research shows that firms with more earnings management in the IPO year subsequently experience poorer stock performance (Cohen & Zarowin, 2010; Roychowdhury et al., 2012; Teoh et al., 1998). In a seminal paper, Teoh et al. (1998) prove that firms with more AEM behavior in the IPO year exhibit greater underperformance in the subsequent three years. Follow-up research shows that REM activities can also lead to underperformance and have worse consequences than AEM (Cohen & Zarowin, 2010; Roychowdhury et al., 2012). On the other hand, some studies pointed out that firms may use REM as a signaling mechanism for favorable future operating performance (Gunny, 2010), especially when they have weak information environments (Beyer et al., 2018). In the Chinese setting, although some studies have proved significant relationships between various forms of earnings management and the long-run market performance, there is little research on how REM and long-run post-IPO returns are related.

As one of the effective information intermediaries, the media usually serve as an external supervisor of managerial opportunism, alleviating information asymmetry and facilitating market efficiency (Bushee et al., 2009; Drake et al., 2014; Fang & Peress, 2009; Peress, 2014). Consistent with this view, Chen et al. (2021) find that the media improve the information environments of IPO companies, suppressing both AEM and REM practices. Additionally, Chahine et al. (2015) find that informative news could weaken the negative impact of pre-equity carve-outs (pre-ECO) earnings management on long-run ECO performance. Although the media could be helpful for investors, some news reports could be detrimental if they provide biased information or impose short-term performance pressure on managers (Dai et al., 2021). Accordingly, Media coverage may amplify the negative effect of earnings management on future returns. This study investigates the long-run role of the media in the capital market by examining the impact of media coverage on the consequences of IPO firms’ REM practices.

In this study, I use a sample of 1,006 Chinese IPOs issued over the 2010–2017 period. I focus on the Chinese setting for three reasons. First, according to the study of Boulton et al. (2011), the Chinese capital market is worth investigating, as it has the highest level of underpricing (120.72%), the highest score in average aggregate earnings managementFootnote 1 (35.87), and the highest score in average earnings opacityFootnote 2 (8.03) among 37 countries in their research. Likewise, Eng and Lin (2012) find that many Chinese companies, even those cross-listed in the US, have significant accounting deficiencies. In this sense, the association between earnings management and the firm’s future performance in China is more important than elsewhere. Second, unlike many stock markets around the world, most trades on the Chinese stock market are made by retail investors (Dai et al., 2016). Compared with institutional investors, retail investors typically encounter more difficulties in either acquiring or processing information related to corporate performance. Thus, the media could become one of their important information sources. Finally, most media are owned by the government in China (He et al., 2008). This unique characteristic presents an interesting question: Can media still effectively influence the capital market despite that media are mainly controlled by the government?

Inspired by Teoh et al. (1998) that investors may not have full knowledge of earnings immediately, I hypothesize that earnings management around the IPO temporarily boosts stock prices, which will revert gradually over time. Therefore, I expect a negative relationship between REM and the long-run market performance. Following Roychowdhury (2006), I measure REM through abnormal production costs, abnormal reduction of abnormal discretionary expenses, and abnormal reduction of operating cash flows. The results present that the market consequences of REM vary depending on the form of REM practices. Abnormal reduction of operating cash flows and abnormal reduction of abnormal discretionary expenses are negatively related to long-run returns. This finding implies that the impact of cuts in operating cash flows or discretionary expenses is reversed in the long run, the biased market belief tends to correct itself and prices decline. In contrast, I find that abnormal production costs are positively related to long-run returns, which is not in line with my expectation or some previous research. This finding implies that some REM practices could predict positive long-run stock returns. One potential explanation is that managers engage in overproduction to signal future operating performance. Investors are initially not aware of the implication of managerial behavior and undervalue the firm based on high abnormal production costs. When the firm’s performance becomes available, investors revise their belief upward, resulting in positive stock returns in the long term. This argument is examined and confirmed in the additional analyses using data on the IPO firms and their matching non-IPO firms.

Next, I examine whether media coverage could influence the relationship between REM behavior in the IPO year and subsequent stock returns. I find that the positive association between abnormal production costs and stock returns becomes weaker if an IPO firm is more visible through the media. Meanwhile, the negative relationship between the abnormal reduction of discretionary expenses and stock returns is alleviated if an IPO company has more media exposure. These findings suggest that media coverage could mitigate the influence of REM on stock returns, improving market efficiency. Thus, my empirical results support the monitoring role of media in the capital market.

Further, as retail investors may have worse information-processing abilities, they are more likely to temporarily misprice the IPO firms’ REM activities based on their biased beliefs than institutional investors. Thus, I expect and confirm that the relationship between REM and future performance is less pronounced when the IPO firm is with higher institutional ownership. I also investigate the different effects of media coverage on retail and institutional investors. Retail investors are more likely to be influenced by news articles than institutional investors because they have less access to corporate information and the media could become one of their important information sources. Consistent with this view, I expect that the media have fewer effects on the relationship between REM and future performance when the IPO firm is with higher institutional ownership. My findings support this conjecture. I also compare the role of audit quality and analyst following with the role of media coverage and the results confirm the effective monitoring role of media coverage in the IPO year.

This study makes several contributions to the literature. First, this study contributes to the earnings management literature by documenting how different forms of REM activities influence post-IPO performance. This study should be useful to market participants and academic researchers who are interested in the consequences of REM on public companies and capital markets. Second, this research adds new empirical evidence to the ongoing debate about the role of the media in the Chinese capital market from the perspective of REM. This study shows that media coverage has a real impact on the market by mitigating the positive (negative) influence of overproduction (reduction of discretionary expenditures) on long-run stock performance. Although these findings imply that the effects of media on the consequences of REM vary with the forms of REM practices, they are generally consistent with the literature about the monitoring role of the media in the capital market (Dai et al., 2015; Liu & McConnell, 2013), suggesting that media coverage plays an important role in improving the market’s ability to efficiently incorporate accounting information into stock prices.

The rest of the paper is organized as follows. Section 2 reviews existing literature and describes hypothesis development. Section 3 describes the sample selection and data. Section 4 shows the empirical results. Section 5 shows the results of additional analyses, and Sect. 6 concludes.

2 Relevant Literature and Hypothesis Development

2.1 REM Around the IPO

REM refers to managers using real activities to manipulate earnings. Roychowdhury (2006) defines REM as “departures from normal operational practices, motivated by managers’ desire to mislead at least some stakeholders into believing certain financial reporting objectives have been achieved in the normal course of operations.” REM includes opportunism actions such as reducing discretionary spending (e.g., R&D, advertising, and maintenance), delaying new project launches, overproduction, and accelerating sales.

Prior research suggests that IPO firms have strong incentives to engage in earnings management. Teoh et al. (1998) indicate that the lock-up restriction on managers' stock sales after IPO, avoiding future litigation risks due to abnormal post-IPO share price declines, executive compensation, and meeting earnings forecasts are strong drivers of upward earnings management at the end of the IPO year. Numerous studies have found evidence that IPO firms increase reported earnings through AEM (Aharony et al., 1993; Teoh et al., 1998) and/or REM practices (Darrough & Rangan, 2005; Fedyk et al., 2017) around the IPO to maintain high stock prices.

Some scholars argue that firms are more likely to engage in REM relative to AEM for several reasons. First, REM is less likely to be detected by auditors, regulators, and investors than AEM (Cohen et al., 2008; Graham et al., 2005; Roychowdhury, 2006). Second, REM occurs throughout the year, while AEM is only undertaken at the end of a quarter or fiscal year. Managers may not be able to meet their earnings goals if they only manage earnings with accrual accounts. Finally, as the balance sheet accumulates all the previous changes in accounting methods (Barton & Simko, 2002), firms that undertook extensive AEM in previous years tend to use REM in the current period (Barton & Simko, 2002; Gunny, 2010). Therefore, firms may prefer to engage in REM in the IPO year.

Overall, given the strong incentives of REM activities in the IPO year, it is important to investigate whether such activities influence a company's market performance post IPO and whether media coverage can serve as a monitoring mechanism to mitigate the influence of such activities.

2.2 Earnings Management and Post-IPO Market Performance

Earnings management activities could reduce firm value because actions taken to boost current earnings could negatively affect future cash flows (Roychowdhury, 2006). For instance, overproduction creates excess inventories, which increases inventory holding costs in the subsequent periods. Due to the influence of cognitive bias, investors may not fully understand managed earnings and temporarily overvalue the IPO firms based on their biased beliefs (Shen et al., 2014). As the effects of earnings management are reversed in the long term, the biased market belief tends to correct itself and prices decline (Shen et al., 2014). Consequently, earnings management that occurs in the IPO year potentially negatively influences stock returns in the post-IPO period (Chang et al., 2010; Fang & Peress, 2009; Teoh et al., 1998). Teoh et al., (1998) find that IPOs that conduct accrual earnings manipulation experience declines in long-run stock returns after the IPO. Shen et al. (2014) confirm that IPO firms with more managed accruals tend to be overvalued immediately after going public, leading to poorer stock performance in the long run.

REM activities have also been found to negatively influence subsequent stock returns and operating performance, and the consequences may be severer than those of AEM (Cohen & Zarowin, 2010; Roychowdhury et al., 2012). In the Chinese setting, Li (2010) finds that firms with higher abnormal production costs or lower cash flows from operations potentially have lower subsequent stock returns. Thus, I expect:

H1

REM in the IPO year is negatively related to stock returns in the three years thereafter.

In contrast, some studies document a positive relationship between REM and future performance (e.g. Beyer et al., 2018; Gunny, 2010). Managers can use REM to avoid losses or earnings declines in an effort to enhance the firm’s credibility and reputation with stakeholders (Bartov et al., 2002). The enhanced reputation allows the firm to perform better in the future. Consistent with this view, Gunny (2010) finds that REM is positively related to future performance when firms meet earnings benchmarks, suggesting that managers may use REM as a signaling mechanism for future operating performance. Beyer et al. (2018) show that firms with poor information environments are more likely to use REM to signal their favorable prospects. As IPO firms have weak information environments, their managers are more likely to engage in some REM activities to signal a positive future outlook, resulting in a positive relationship between some REM behavior and future operating performance. Meanwhile, as investors are not fully aware of the implication of these REM activities, they may initially undervalue the firm based on REM and correct their belief when information about favorable corporate performance becomes available. Thus, some REM activities may be positively related to long-run market returns. As a result, in the Chinese IPO setting, the relationship between REM and future returns is largely an empirical matter.

2.3 Market Performance and the Interaction Between Media Coverage and Earnings Management

Since earnings management is a common way for managers to maximize their own interests at the expense of the company's long-term value, it is important to examine whether the media have a positive or negative effect on earnings management, which in turn has an impact on the capital market. The "effective supervision hypothesis" supports that media could serve as an external supervisor of managerial opportunism, alleviating the degree of information asymmetry in the market, and thereby effectively inhibiting earnings management behavior. Accordingly, the more the media attention, the more supervision and governance effects it can have on listed companies and the weaker the association between earnings management and subsequent stock returns. Consistent with this view, Chahine et al. (2015) examine and confirm that a high level of informative news at the time of the equity carve-outs (ECO) will moderate the negative impact of pre-ECO earnings management on long-run ECO performance. Additionally, Drake et al. (2014) find that media attention mitigates cash flow mispricing and media coverage can help the market better understand accounting information.

On the other hand, the "market pressure hypothesis" indicates that media attention could impose short-term market pressure on managers, inducing them to manipulate earnings. When the firm is with more media exposure, managers may manage earnings to avoid announcing negative earnings news or to achieve short-term goals because the negative consequences of unfavorable earnings will be exacerbated by sensational news reports. Accordingly, the more the media attention, the stronger the association between earnings management and subsequent stock returns. Consistent with this view, several studies find a positive relationship between media reporting and REM (Dai et al., 2015; Meng, 2020), but it is unclear whether media coverage mitigates or amplifies the effect of REM in the IPO year on the long-run market performance. Overall, I expect:

H2

Media coverage in the IPO year mitigates the relationship between REM in the IPO year and stock returns in the three years thereafter.

3 Sample and Data

3.1 Sample Selection

This study will use A-share IPOs in Chinese market markets. Financial data and stock returns are obtained from the China Stock Market and Accounting Research (CSMAR) database, and data on media coverage are collected from the “Daily News” section of the RESSET database.Footnote 3 This section collects public news on specific stocks from well-known media sources, including Shanghai Securities News, Securities Times, China Economic Net, Sina Finance, and Tencent Finance. The content of these news articles includes financial results, corporate fraud, and product marketing information. The sample includes IPOs during the 2010–2017 period and excludes financial firms. I require that CSMAR coverage of the security includes the necessary data to analyze REM. Based on these restrictions, the sample has 1,006 IPO firms listed from 2010 to 2017. Table 1 presents the distribution of IPO firms by year and by industry. Panel A shows that the largest number of IPO firms was in 2017 (30.91 percent of the sample), while the number of IPOs in 2012 was small (4.17 percent of the sample). None of the IPOs in 2013 is included in this sample.Footnote 4 Except for these years, the distribution of IPOs is relatively even (6–21 percent of the sample per year). Panel B shows that the largest number of IPO firms was in the manufacturing industry (77.83 percent of the sample).

Table 1 Sample IPOs characteristics
Full size table

3.2 Measures of REM

Following prior literature (Cohen & Zarowin, 2010; Roychowdhury, 2006), this study measures REM using abnormal production costs, abnormal discretionary expenses, and abnormal operating cash flows and uses the model developed by Dechow et al. (1998) and implemented by Roychowdhury (2006). I first estimate the normal level of production costs using the following cross-sectional regression for each industry and year for all non-IPO firms.Footnote 5

$$\frac{{PROD_{i,t} }}{{AT_{i,t - 1} }} = \alpha_{0} + \beta_{1} \frac{1}{{AT_{i,t - 1} }} + \beta_{2} \frac{{SALE_{i,t} }}{{AT_{i,t - 1} }} + \beta_{3} \frac{{\Delta SALE_{i,t} }}{{AT_{i,t - 1} }} + \beta_{4} \frac{{\Delta SALE_{i,t - 1} }}{{AT_{i,t - 1} }} + \varepsilon_{i,t}$$
(1)

where i denotes firm, t denotes year, PROD is the sum of the cost of goods sold and the change in inventory from year t-1 to t, AT is the firm’s book assets; SALE is the firm’s sales revenue, and ε is the error term. I estimate the equation for each industry in each year and require each industry-year to have at least 6 observations. Abnormal production costs (ABPROD) are calculated as actual PROD minus the normal level of PROD estimated using the coefficients from regression (1). Higher values of ABPROD indicate more REM.

Further, the normal level of discretionary expenses can be expressed as follows:

$$\frac{{DISX_{i,t} }}{{AT_{i,t - 1} }} = \alpha_{0} + \beta_{1} \frac{1}{{AT_{i,t - 1} }} + \beta_{2} \frac{{SALE_{i,t - 1} }}{{AT_{i,t - 1} }} + \varepsilon_{i,t}$$
(2)

where DISX is the sum of R&D, advertising, and selling, general and administrative (SG&A) expenses. Following Cohen and Zarowin (2010), I set R&D and advertising expenses to zero if they are missing. All other variables are as defined in Eq. (1). I conduct the same industry-year regressions using non-IPO firms and define abnormal discretionary expenses (ABDISX) of IPO firms as actual DISX minus the normal level of DISX estimated using the coefficients from regression (2). Lower values of ABDISX indicate more REM. I multiply ABDISX by negative one and get the new variable (N_ABDISX) so that higher values of this measure indicate more REM.

Last, the normal level of operating cash flows can be expressed as follows:

$$\frac{{CFO_{i,t} }}{{AT_{i,t - 1} }} = \alpha_{0} + \beta_{1} \frac{1}{{AT_{i,t - 1} }} + \beta_{2} \frac{{SALE_{i,t} }}{{AT_{i,t - 1} }} + \beta_{3} \frac{{\Delta SALE_{i,t} }}{{AT_{i,t - 1} }} + \varepsilon_{i,t}$$
(3)

where CFO is the firm’s operating cash flows. All other variables are as defined in Eq. (1). I conduct the same industry-year regressions using non-IPO firms and define abnormal operating cash flows (ABCFO) of IPO firms as actual CFO minus the normal level of CFO estimated using the coefficients from regression (3). Lower values of ABCFO indicate more REM. I multiply ABCFO by negative one and get the new variable (N_ABCFO) so that higher values of this measure indicate more REM. Finally, I combine the three individual measures into one comprehensive REM measure (TREM) as ABPROD plus N_ABDISX and N_ABCFO.

3.3 Measures of Long-Run Market Performance

Following Ritter and Welch (2002), this research estimates the buy-and-hold abnormal returns (BHAR) using a market benchmark. Returns are calculated from the beginning of the fifth month after the IPO-year fiscal year-end so that all the earnings information contained in the first financial statement can be known for sure to investors.Footnote 6BHAR is given in Eq. (4).

$$BHAR\left[ {0,T} \right] = \frac{{\mathop \sum \nolimits_{i = 1}^{N} \left[ {\mathop \prod \nolimits_{t = 1}^{T} \left( {1 + r_{i,t} } \right) - \mathop \prod \nolimits_{t = 1}^{T} \left( {1 + r_{m,t} } \right)} \right]}}{N}$$
(4)

where \({r}_{i,t}\) and \({r}_{m,t}\) are monthly raw return for stock \(i\) and market return, and \(N\) is the number of stocks in the month \(t\). I estimate the long-run performance up to 36 months post-IPO.

3.4 Measures of Media Coverage

To measure media coverage, I first obtain the number of news articles with the firm as the subject in the IPO year from the RESSET database. If the news article is issued on a non-trading day, the issue date is adjusted to the next trading day. The logarithm of one plus the sum of news reports in the sample frame (ln(News)) is used in the analyses.

3.5 Control Variables

Following prior studies (e.g., Bergstresser & Philippon, 2006), this research uses the following variables to control for several firm-level characteristics that could affect REM behavior or long-run returns. The Appendix provides detailed definitions of these variables.

Dsale is asset-scaled sales growth in the IPO year. DNI is the asset-scaled net income growth in the IPO year. DCapExp is the asset-scaled mean capital expenditures in years 1, 2, and 3 less the asset-scaled mean expenditures in years − 1 and 0, where the year 0 is the IPO year. IR is an IPO firm’s initial return. MktRet is the contemporaneous three-year value-weight buy-and-hold return from the exchange where the IPO is listed, calculated from the listing month. ln(AGE) is the logarithm of one plus the firm's age. ln(MV) is the logarithm of market value at the first post-IPO fiscal year-end. ln(BM) is the logarithm of the book-to-market ratio (BV/MV). ln(Analyst) is the logarithm of one plus the sum of analyst reports covering the firm in the IPO year. Big10 is an indicator variable, which takes a value of one if the firm has a big-ten auditor and zero otherwise. Underwriter is an indicator variable, which takes a value of one if the underwriter of the IPO firm in the current year ranks in the top ten among all underwriters and zero otherwise. ln(Days) is the logarithm of the number of days between the IPO date and the first post-IPO fiscal year end.

Finally, following the modified Jones (1991) model of Dechow et al. (1995), I estimate abnormal discretionary accruals as follows:

$$\frac{{TA_{i,t} }}{{AT_{i,t - 1} }} = \alpha_{0} + \beta_{1} \frac{1}{{AT_{i,t - 1} }} + \beta_{2} \left( {\frac{{\Delta REV_{i,t}}}{{AT_{i,t - 1} }}}- {\frac{{\Delta AR_{i,t}}}{{AT_{i,t - 1} }}}\right) + \beta_{3} \frac{{PPE_{i,t} }}{{AT_{i,t - 1} }} + \varepsilon_{i,t}$$
(5)

where TA is total accruals, calculated as the difference between income before extraordinary items and operating cash flows; ΔREV is the change in sales from year t−1 to t; ΔAR is the change in accounts receivable from year t−1 to t; and PPE is gross property, plant, and equipment. All other variables are as defined in Eq. (1). I estimate the regression in each year for each industry and require each industry-year to have at least 6 observations. Discretionary accruals are calculated as the residuals from Eq. (5). Because discretionary accruals can be either income-increasing or income-decreasing, I use the absolute value of discretionary accruals (Abs(DTAC)) as the control variable. Higher values of Abs(DTAC) indicate more AEM.

3.6 Empirical Model

This study examines the long-run post-IPO market consequence of REM and analyzes the effect of media coverage on the relationship between REM and long-run post-IPO returns. In the following models, i denotes firm, t denotes year, and ε is the error term. I test the hypothesis H1 by the significance and direction of \({\beta }_{1}\) in Eq. (6). Similarly, the significance and direction of the significance and direction of \({\beta }_{3}\) in Eq. (7) examine the hypothesis H2.

$$\begin{aligned} BHAR_{i,t} = & \alpha_{0} + \beta_{1} REM_{i,t} + \beta_{2} {\text{Abs}}\left( {DTAC} \right)_{i,t} + \beta_{3} Dsale_{i,t} + \beta_{4} DNI_{i,t} + \beta_{5} CapExp_{i,t} \\ & + \beta_{6} IR_{i,t} + \beta_{7} MktRet_{i,t} + \beta_{8} \ln \left( {AGE} \right)_{i,t} + \beta_{9} \ln \left( {MV} \right)_{i,t} + \beta_{10} \ln \left( {BM} \right)_{i,t} \\ & + \beta_{11} \ln \left( {Analyst} \right)_{i,t} + \beta_{12} Big10_{i,t} + \beta_{13} UnderWriter_{i,t} \\ & + \beta_{14} \ln\left( {Days} \right)_{i,t} + Year + Industry + \varepsilon_{i,t} \\ \end{aligned}$$
(6)
$$\begin{aligned} BHAR_{i,t} = & \alpha_{0} + \beta_{1} {\text{ln}}\left( {News} \right)_{i,t} + \beta_{2} REM_{i,t} + \beta_{3} REM_{i,t} {\text{ * ln}}\left( {News} \right)_{i,t} \\ & + \beta_{4} {\text{Abs}}\left( {DTAC} \right)_{i,t} + \beta_{5} Dsale_{i,t} + \beta_{6} DNI_{i,t} + \beta_{7} DCapExp_{i,t} + \beta_{8} {\text{IR}}_{i,t} \\ & + \beta_{9} MktRet_{i,t} + \beta_{10} \ln \left( {AGE} \right)_{i,t} + \beta_{11} \ln \left( {MV} \right)_{i,t} + \beta_{12} \ln \left( {BM} \right)_{i,t} \\ & + \beta_{13} \ln \left( {Analyst} \right)_{i,t} + \beta_{14} Big10_{i,t} + \beta_{15} UnderWriter_{i,t} + \beta_{16} \ln\left( {Days} \right)_{i,t} \\ & + Year + Industry + \varepsilon_{i,t} \\ \end{aligned}$$
(7)

4 Results

4.1 Descriptive Statistics

Table 2 presents the summary statistics of the variables in the regression analysis. To mitigate the effects of outliers, all the continuous variables are winsorized at both the 1st and 99th percentiles. The mean values of the REM measures, ABPROD, N_ABDISX, N_ABCFO, and TREM are 0.04, − 0.02, − 0.03, and 0.00, respectively. The average 3-year BHAR is 0.12 and the average ln(News) is 3.72. For the control variables, the average Abs(DTAC) in the sample is 0.12. On average, these firms have a sales growth (Dsale) of 19%, a net income growth (DNI) of 2%, a capital expenditures growth (DCapExp) of 12%, initial returns (IR) of 40%, logged ages (ln(AGE)) of 8.21, a logged market value (ln(MV)) of 21.66, and a logged book-to-market ratio (ln(BM)) of − 0.52. The mean value of MktRet is 4%. The average ln(Analyst) is 2.88 and the average ln(Days)) is 4.74. About 57% of firms’ first post-IPO financial statements are audited by the Big 10, and about 32% of firms have an underwriter ranking in the top ten.

Table 2 Descriptive statistics
Full size table

4.2 REM and the Long-Run Market Performance

4.2.1 Univariable Analysis

In Figs. 1, 2, and 3, IPO firms are ranked in quartiles based on three REM measures, ABPROD, N_ABDISX, and N_ABCFO, respectively. Figures 1, 2, and 3 plot the time series of the average BHAR of four quartile portfolios. Figure 1 shows that IPOs in the fourth quartile overperform those in the first quartile over three years, indicating that firms engaging more in overproduction tend to have higher long-run abnormal returns. Both Figs. 2 and 3 present that the first quartile outperforms the fourth quartile over three years. These results suggest that more abnormal cuts in discretionary expenses or operating cash flows will result in lower long-run abnormal returns.

Fig. 1
figure 1

ABPROD and Average BHAR

Full size image
Fig. 2
figure 2

N_ABDISX and Average BHAR

Full size image
Fig. 3
figure 3

N_ABCFO and Average BHAR

Full size image

Table 3 reports the average BHAR for each quartile. BHAR is provided for each of the one, two, and three subsequent years, calculated from the beginning of the fifth month after the first post-IPO fiscal year-end. IPO firms are ranked in quartiles based on the magnitude of REM. In Columns 1, 2, and 3, ABPROD is positively related to BHAR for the first, second, and third years. In contrast, Columns 4, 5, and 6 show that N_ABDISX is negatively related to BHAR. Columns 7, 8, and 9 indicate an insignificant relationship between N_ABCFO and BHAR. In summary, these results demonstrate the effect of REM on long-run returns varies with the forms of REM activities.

Table 3 Time-series means of equal weighted portfolio abnormal stock returns
Full size table

4.2.2 Multivariable Analysis

I further conduct a regression analysis to examine the relationship between various REM activities and 3-year post-IPO abnormal returns. Table 4 reports the results. In Column 1, the coefficient of ABPROD is positive, suggesting that IPO firms with more abnormal production costs in the IPO year have higher post-IPO long-run returns. This finding is inconsistent with some previous research (e.g. Li, 2010), which supports that overproduction leads to long-run underperformance. In Columns 2, 3, and 4, the coefficients of N_ABDISX, N_ABCFO and TREM are negative, indicating that IPO firms reducing more discretionary expenses, operating cash flows or the total amount of REM have lower long-run returns. These findings are consistent with some prior studies (e.g. Li, 2010; Roychowdhury, 2006), which suggest that REM activities have negative influences on long-run performance.

Table 4 The associations between REM and 3-year BHAR
Full size table

Overall, these results are in line with those of univariable analysis but inconsistent with H1. One potential explanation for the different effects of various REM practices is that IPO firms with higher abnormal production costs (higher discretionary expenses or operating cash flows) potentially have better (worse) operating performance and the market initially misprices these REM activities, resulting in higher (lower) long-run stock returns. I will analyze this argument in Sect. 5.1.

4.3 The Effect of Media Coverage on the Relationship Between REM and Returns

4.3.1 Double Sorting on Both Media and REM Measures

I sort individual stocks into five portfolios based on the media coverage measure and select the stocks from the highest and lowest media coverage portfolios. Then I sort selected stocks into five portfolios based separately on the specific REM measures per media coverage portfolio and obtain equal-weighted BHAR for each portfolio. In Figs. 4, 5, and 6, I plot the BHAR of a long-short portfolio with a long position in the highest REM portfolio and an offsetting short position in the lowest REM portfolio.

Fig. 4
figure 4

Hedge portfolio BHAR and ABPROD

Full size image
Fig. 5
figure 5

Hedge portfolio BHAR and N_ABDISX

Full size image
Fig. 6
figure 6

Hedge portfolio BHAR and N_ABCFO

Full size image

Figure 4 shows the hedge portfolio BHAR is positive and becomes greater with less media coverage. These results indicate that ABPROD is positively related to BHAR and this relationship becomes weaker if the firm is more visible through the media. In contrast, Fig. 5 reports that the hedge portfolio BHAR is negative and becomes lower with less media coverage. These results suggest that media coverage mitigates the negative association between BHAR and N_ABDISX. In Fig. 6, the hedge portfolio BHAR is positive when the firm is with more media coverage, but it is negative when the firm is with less media coverage in most months within the 36-month post-IPO period. The line for the hedge portfolio BHAR and N_ABCFO is less volatile if the firms are with more media exposure, indicating that media coverage alleviates the relationship between BHAR and N_ABCFO. Overall, these findings support the prediction that media coverage mitigates the effect of REM on long-run stock returns.

4.3.2 Multivariable Analysis

Further, I conduct a regression analysis to test the effect of media coverage on the relationship between various REM activities and long-run returns. Table 5 presents the results. In Column 1, the coefficient of ABPROD * ln(News) is significantly negative, suggesting that the association between ABPROD and BHAR becomes less positive with more IPO-year news articles. Column 2 shows a positive coefficient on the interaction of N_ABDISX and ln(News). This result suggests that the association between N_ABDISX and BHAR becomes less negative with more news reports in the IPO year. Column 3 (4) reports that the coefficient of N_ABCFO * ln(News) (TREM * ln(News)) is positive but insignificant, indicating that media coverage does not influence the relationship between N_ABCFO (TREM) and BHAR. Overall, these results provide evidence that news coverage mitigates the association between specific REM practices and long-run returns, supporting H2.Footnote 7

Table 5 The effect of media coverage on the associations between REM measures and 3-year BHAR
Full size table

In summary, Fig. 7 shows the results of H1 and H2.

Fig. 7
figure 7

The results of Hypotheses

Full size image

4.4 Robustness Tests

I check the robustness of my findings using an alternative measure of media coverage. The RESSET database defines the information level, InfoLvl, as the influence of news articles on the market and ranks it on a scale of 1 to 5. InfoLvl equals 1 if the information is relatively unimportant, which has no direct impact on the market or has a slight impact on the market. InfoLvl equals 2 if the news is about daily information of the general level. InfoLvl equals 3 if the information is relatively important, which has a significant impact on the securities market, industry market, and regional market. InfoLvl equals 4 if the information is a headline in an important media outlet (except for non-substantive content, such as newspaper suspension notices, information guides, error correction statements, etc.), important economic data disclosures (CPI, PPI, PMI, etc.), or important information released by important national institutions (the China Securities Regulatory Commission, the China Banking Regulatory Commission, the National Development and Reform Commission, the Bureau of Statistics, etc.). InfoLvl equals 5 if the information is bombshell news, which has a significant impact on the market and the macro-environment, such as the circuit breaker mechanism. I use the logarithm of one plus the sum of the information level of each news report in the IPO year, ln(InfoLvl), as an alternative media coverage proxy to re-examine the effect of media coverage on the relationship between REM and long-run stock returns. The control variables are the same as in the previous analysis. The results in Panel A of Table 6 are similar to those reported in Table 5, verifying that my findings are robust.

Table 6 Robustness check: an alternative measure of media coverage
Full size table

Moreover, I conduct robustness checks using an alternative measure of long-run market performance. I use the HUSHEN 300 Index as the benchmark to adjust stock returns and the untabulated results are qualitatively similar to those reported in Tables 4 and 5, confirming the robustness of my findings.

Further, I use Srivastava (2019)’s model to estimate the abnormal production costs and re-examine the relationship between abnormal production costs and long-run post-IPO stock returns. The untabulated results show abnormal production costs are positively related to stock returns within manufacturing companies but negatively related to stock returns within non-manufacturing firms. These findings indicate that my results do not change qualitatively when using an alternative measure of abnormal production costs and that production cost manipulation can only be fully utilized by manufacturing firms.

5 Additional Analyses

5.1 A Potential Explanation for the Positive Relationship Between Abnormal Production Costs and Future performance

In the previous subsections, I find firms with higher abnormal production costs potentially have better long-run market performance. This finding is inconsistent with many previous studies, which document that more REM activities will result in worse market consequences (e.g. Li, 2010; Roychowdhury, 2006). However, my result is consistent with the finding of Gunny (2010). Gunny (2010) finds that REM is positively related to future performance when firms meet earnings benchmarks, which supports that managers may use REM as a signaling mechanism for future operating performance. Managers engage in REM to avoid losses or earnings declines in an effort to enhance the firm’s credibility and reputation with stakeholders (Bartov et al., 2002). The enhanced reputation allows the firm to perform better in the future. Thus, I assume that managers are confident in their ability to sell products in the future, so they carry out overproduction, leading to high ABPROD in the IPO year. Meanwhile, investors are not fully aware of the implications of ABPROD in the IPO year on future performance, so they initially undervalue the firm based on high ABPROD and correct their belief until the information about favorable corporate performance becomes available. Accordingly, ABPROD is positively related to both operating and market performance in the long run.

To examine this assumption, I test the relationship between ABPROD and future performance using the full sample and two subsamples with meeting and missing the earnings benchmark. I use the industry-adjusted change in ROA in the third year following the IPO (∆ROA) to capture the IPO firms’ subsequent operating performance. The benchmark is defined as the earnings in the IPO year being higher than the earnings in the year prior to the IPO. Table 7 shows the results. Column 1 indicates a positive relationship between ABPROD and ∆ROA, suggesting that IPO firms with higher abnormal production costs potentially perform better in subsequent years.Footnote 8 Columns 2 and 3 show significant positive coefficients of ABPROD, indicating that firms using abnormal production costs to avoid earnings declines potentially perform better in the future. Columns 4 and 5 report an insignificant relationship between ABPROD and ∆ROA. This means that the use of abnormal production is not able to improve a firm’s reputation or signal future performance when the firm is suffering earnings declines. These results support my expectation that managers can use overproduction as a signaling mechanism for favorable future performance.

Table 7 Abnormal production costs and future performance
Full size table

Further, Beyer et al. (2018) show that firms with poor information environments are more likely to use REM to signal their favorable prospects. As IPO firms have weaker information environments than established firms, their managers are more likely to engage in REM to signal a positive future outlook. To test this assumption, I match each IPO firm with a non-IPO firm based on a host of firm characteristics—including assets, age, return on assets, Big 10 auditor, analyst following, industry, and year. I follow Rosenbaum and Rubin (1985) in implementing the propensity score matching and finally select a non-IPO listed firm that has the closest propensity score as the matched non-IPO firm for each IPO firm. I examine whether ABPROD is positively related to future operating and market performance within these non-IPO firms. Columns 6 and 7 of Table 7 show negative and significant coefficients of ABPROD, confirming my assumption that firms with better information environments are less likely to use REM to signal future performance.

Overall, these results confirm the positive effect of abnormal production costs on future performance, which is opposite to the effect of abnormal reductions of discretionary expenses or operating cash flows on future performance shown in Sect. 4.2. These findings suggest that in a Chinese IPO setting, the impact of REM on future performance depends on the form of REM activities.

5.2 Different Effects of Media Coverage on Retail and Institutional Investors

Compared with institutional investors, retail investors may have worse information-processing abilities and be more likely to temporarily misprice firms’ REM activities based on their biased beliefs (Collins et al., 2003). Thus, I expect the relationship between REM and future performance is less pronounced when the IPO firm is with higher institutional ownership. Moreover, retail investors may be more likely to be influenced by news articles than institutional investors. Compared with institutional investors, retail investors have less access to corporate information. Therefore, the media could become one of their important information sources and influence their investment decisions. Bushee et al. (2020) document that retail investors are more susceptible to media-driven trade immediately following the IPO, resulting in worse investing outcomes. However, few studies have investigated the different effects of media coverage on retail and institutional investors from the view of the relationship between REM and long-run stock returns.

To test the above predictions, I split the original sample into two subsamples with low- and high-institutional ownership at the end of the IPO year. Panel A of Table 8 shows the relationship between REM and long-run stock returns in these two subsamples. The control variables in Table 8 are the same as in the main analysis. In Columns 1 to 4, the coefficient of each REM measure is significant. In columns 5 to 8, the coefficients of ABPROD and N_ABDISX are significant, while the coefficients of N_ABCFO and TREM are insignificant. The signs of these coefficients are the same as those in Table 4. These findings are consistent with my prediction that retail investors are more likely to misunderstand managed earnings and misprice the IPO firms based on their biased beliefs than institutional investors.

Table 8 Different effects of media coverage on retail and institutional investors
Full size table

Panel B of Table 8 reports the different effects of media coverage on retail and institutional investors. In Columns 1 to 4, the coefficients of ABPROD * ln(News) and N_ABDISX * ln(News) are significant and the signs of these coefficients are the same as those in Table 5. In contrast, Columns 5 to 8 show that none of the coefficients on the interaction of REM measure and media coverage measure is significant. These findings confirm my prediction that media coverage can help investors, especially retail investors, understand managerial activities.

5.3 The Role of Audit Quality and Analyst Following

Extensive studies document the effect of audit quality on earnings management activities (e.g. Balsam et al., 2003; Francis & Yu, 2009). High-quality auditors can detect managerial opportunism in financial reporting and take corrective actions to improve the earnings quality of their client firms. Therefore, I expect that high-quality auditors can also play a monitoring role in mitigating the influence of REM on long-run stock returns. Additionally, I test whether other information intermediaries play a similar role as media coverage in the capital market. Prior literature shows that analyst earnings forecasts are informative about stock prices (Lys & Sohn, 1990) and that analyst following improves the market’s ability to efficiently incorporate information about future earnings into stock prices (Ayers & Freeman, 2003). Thus, I expect that analyst following can mitigate the relationship between REM activities and long-run stock returns.

I employ Big10 and ln(Analyst) as the measure of audit quality and analyst following, respectively. The control variables are the same as in the previous analysis. Table 9 presents the results. The coefficient of the interaction term between each REM measure and Big10 (ln(Analyst)) is insignificant. These findings are inconsistent with my predictions and suggest that compared with media coverage, audit quality or analyst following has a relatively less important effect on the consequences of REM activities in the Chinese IPO setting.

Table 9 The role of audit quality and analyst following
Full size table

6 Conclusions

The paper examines the relationship between REM activities in the IPO year and the market performance in subsequent years. The result shows abnormal reduction of discretionary expenses or operating cash flows is negatively related to subsequent three-year stock returns. These results suggest that investors are unable to fully understand managed earnings and behave as if they are fixated on these high earnings, resulting in disappointing investment outcomes in the long term.

Surprisingly, I find that abnormal production costs are positively related to subsequent stock performance. This finding is inconsistent with some previous studies (e.g. Li, 2010; Roychowdhury, 2006), but provides evidence to support the studies of Gunny (2010) and Beyer et al. (2018). This result indicates that IPO firms potentially use some REM activities (e.g. overproduction) to signal their favorable prospects. As investors may not fully understand managerial behavior, they initially undervalue the firms based on high abnormal production costs. When more information on corporate performance becomes available, investors revise their beliefs upward, resulting in a positive relationship between abnormal production costs and long-run stock returns.

Moreover, this study tests the effects of media coverage on the relationship between REM activities in the IPO year and long-run returns. The empirical results show that both the positive relationship between overproduction and returns and the negative relationship between abnormal reduction of discretionary expenses and returns become weaker when the firm is with more media exposure. These results indicate that media coverage serves as a monitoring mechanism, which mitigates the influence of REM and improves the market’s ability to efficiently incorporate accounting information into the stock price. These findings are useful for investors who want to utilize earnings information and media coverage to differentiate issuers.

In the additional tests, I confirm that a negative relationship between abnormal production costs and long-run future performance exists in non-IPO firms, supporting my prediction that firms with better information environments are less likely to use REM activities to signal their future performance. I also find that the relationship between REM activities and long-run returns is more pronounced and that the media has a more important effect on this relationship when the IPO firm is with high institutional ownership. This result suggests that institutional investors have a better understanding of REM activities, and their investment decisions are less likely to be influenced by the media than retail investors. Finally, my study compares the role of audit quality and analyst following with the role of media coverage and confirms the effective monitoring role of media coverage in the IPO year.

Nevertheless, I recognize the limitations of my research, which should be addressed in future studies. In this study, the control variable, initial returns, is found insignificantly related to long-run future performance. One potential explanation is that initial returns consist of pre-market deliberate underpricing and aftermarket overpricing and only the overpricing component influences long-run post-IPO performance in the Chinese stock market (Gao, 2010). Considering the procedure to measure the overpricing component is complicated and many relevant studies use initial returns as the control variable, I keep using initial returns in my analysis. Future studies can continue to improve research designs by separating initial returns into underpricing and overpricing components.