Earnings management of initial public offering firms: evidence from regulation changes in China

2.1. Regulation of the IPO process

The economic reforms and liberalization that started in 1978 have transformed the Chinese economy from a centrally planned economy to a market-oriented one. A key aspect of economic reform was the Chinese Government's establishment of the SHSE in 1990, followed by the SZSE in 1991. The Chinese Government exerts strong power over the IPO process; it regulates the method of determining the offering price and stipulates a series of criteria for selecting firms that go public. When the Chinese stock market was established, the most common method of determining the offering price was the online fixed-price offering method. Under this method, the offering price was determined by after-tax earnings per share, multiplied by an assigned P/E ratio. The China Securities Regulatory Commission (CSRC) determines the P/E ratio according to the ratios of listed firms in the same locality and industry. The online fixed-price offering method was adopted during the period 1996–2004.1 The P/E ratio was fixed at 15 before 1999, but increased to 50 in January 1999 and then declined to 20 in 2002. Under this method, issuers have a strong incentive to increase accounting earnings to achieve a high offering price.

However, in December 2004, the CSRC abolished the online fixed-price offering method and adopted the book-building mechanism. Book building is a common practice in the IPO process of developed countries. Under this mechanism, the price of securities is determined by the issuing company along with the lead underwriter in consideration of feedback from investors and market intermediaries during a certain period. The mechanism is regarded as a transparent and flexible price determination method for IPOs that potentially reduces information asymmetry between issuers and investors. Importantly, this change (the abolition of the fixed-price offering procedure) will decrease IPO firms' incentive to manipulate accounting earnings for a high offering price.

2.2. Government control

A critical purpose of the establishment of the stock market is to help privatize SOEs. However, these firms are controlled by the state, even after privatization. IPOs provide governments with an opportunity to inject equity capital into SOEs without public expenditures. The growth of SOEs through additional equity capital will contribute to economic growth and thereby benefit governments. Especially, the success of IPOs is often viewed as a sign that local government officials are performing well (Li, 1998; Chen et al., 2008). These facts give governments an incentive to make SOEs that go public manipulate earnings to achieve a high offering price.

State-owned enterprises are controlled by various levels of government (e.g., the central government, local governments (province, city, country and town) and other SOEs). Previous studies suggest that different levels of government exert different levels of intervention on SOEs. Li et al. (2011) suggest local governments interact more closely and frequently with SOEs because local governments rely on SOEs for their revenues. Chen et al. (2008) argue that the decentralization reform in the 1990s induced local governments to compete against each other for equity capital in order to develop local economies. These facts suggest that local governments have a strong incentive to use IPOs to inject equity capital into SOEs.

2.3. Preferential access to bank debt

During the central-planned economy, all firms were SOEs and the government allocated budgets to each firm and collected all their profits. After the economic reform, the four Chinese state-owned commercial banks were instructed to lend money to SOEs until 1998 (Héricourt and Poncet, 2009; Poncet et al., 2010).2 Although the lending restrictions have since been removed, banks still tend to think of non-SOEs as more risky than SOEs (Park and Sehrt, 2001; Héricourt and Poncet, 2009). As a result, SOEs receive bank loans more easily than non-SOEs due to government support and guarantees (Tian, 2001; Wang, 2005). These facts suggest that SOEs receiving preferential access to bank debt suffer less from financial constraints. On the other hand, firms that do not have access to bank debt will face financial constraints and need to seek equity financing. Those firms will have a strong incentive to manipulate earnings for a high offering price in the IPO process.

2.4. Regional variations

From the 1980s, the Chinese Central Government established four ‘Special Economic Zones’ (SEZs), which have embodied preferential policies to attract foreign investments.3 SEZs include 14 open coastal cities and development zones. Under the Chinese banking system, non-SOEs introduced foreign capital to mitigate financial constraints (Héricourt and Poncet, 2009; Poncet et al., 2010). As a result, Chinese firms in provinces with a greater intensity of foreign direct investment (FDI) may face less financial constraints, whereas firms located in provinces with low FDI will suffer from such financial constraints. Indeed, Poncet et al. (2010) find that the geographical presence of foreign firms alleviates credit constraints of Chinese private firms. This fact suggests that firms located in SEZs have a weak incentive to raise much money in an IPO.

3.1. Hypotheses and variables

We attempt to re-examine factors previous studies have shown to affect earnings management of IPO firms by using recent Chinese data. Previous studies argue that earnings management by IPO firms is mitigated if reputable underwriters and auditors are involved with the IPO process (Titman and Trueman, 1986; Aharony et al., 1993; Becker et al., 1998; Klein, 2002). We adopt a dummy variable that takes a value of one if a company has a Big Four international accounting firm as its auditor as a measure of auditor quality (AUDITOR) (See Table 1 for a definition of variables). It is difficult to precisely measure the quality of the underwriters of Chinese IPOs, because government-controlled Chinese securities companies underwrite all sample firms. Following Francis et al. (2009), we picked the top five underwriters based on the frequency of underwriting for the largest 100 IPOs in offering size. We then made a dummy variable that takes a value of one if an IPO firm is underwritten by one of the top five underwriters (UNDERWRITER).

We adopt the change in managerial ownership between the IPO year and the preceding year (Ch_MANAGEROWN) as a proxy for the number of shares managers sell in the IPO process. Ch_MANAGEROWN is predicted to have a negative impact on earnings manipulation (Aharony et al., 1993; DuCharme et al., 2001; Darrough and Rangan, 2005). Peasnell et al. (2005) argue that independent boards effectively decrease discretionary earnings management. We adopt the percentage of independent directors over total board members (INDIR) in our regressions. The audit committee has specific responsibility for the production of financial statements (Peasnell et al., 2005). We define a dummy variable that takes a value of one for firms equipped with an audit committee and zero for others (A_COMMITTEE). Previous studies suggest that firms with long history have low abnormal accruals (Bergstresser and Philippon, 2006; Burgstahler et al., 2006; Fan, 2007). We adopt firm age (AGE) to test this idea. Aharony et al. (1993) and Bergstresser and Philippon (2006) also present that the level of earnings management is negatively related to firm size. We use the natural logarithm of assets as a measure of firm size (LNASSET). Aharony et al. (1993) find a positive relationship between leverage and earnings management for US IPOs. Young companies that go public tend to have rich growth opportunities and therefore should decrease leverage to reduce bankruptcy costs and agency costs of debt (Jensen and Meckling, 1976; Myers, 1977; Pagano et al., 1998). This fact gives IPO firms with high leverage an incentive to manipulate earnings for a high offering price. We compute leverage as total liabilities divided by assets (LEVERAGE) to test this idea.

As mentioned, unique Chinese characteristics such as regulation changes on the IPO process, government control, preferential access to bank debt and regional variations may impact earnings management of IPO firms. We make a dummy variable that takes a value of one for SOEs, and zero for non-SOEs (D_SOE) to examine the effect of government control. As mentioned, different levels of governments potentially have different effects on IPO firms' earnings management. We also make three dummy variables to test this idea: one for SOEs controlled by the Chinese Central Government, and zero for others (SOE_C); one for SOEs controlled by the local government (SOE_L), and zero for others; and one for SOEs controlled by other SOEs and zero for others (SOE_S). If an IPO firm has access to bank loans at the point of IPO, it will face less financial constraints and does not have a strong incentive of earnings management for a high offering price. To test this idea, we use bank debt divided by liabilities (BANKDEBT) as well as a dummy variable that takes a value of one for firms that issue bank debt and zero for others (BANKL). Poncet et al. (2010) find that the geographical presence of foreign firms alleviates credit constraints of private Chinese firms. This discussion suggests that firms located in provinces that attract foreign capital have less incentive to boost earnings. We make a dummy variable that takes a value of one for IPO firms located in the top five provinces or regions that attract foreign capital and zero for others (COASTALREGION).

Finally, the online fixed-price offering method was abolished in China and the book-building mechanism was adopted in December 2004. This regulation change will decrease IPO firms' incentive to manage earnings for a high offering price. We make a dummy variable that takes a value of one for firms that went public in 2005 and after and zero otherwise (BOOKBUILD). We also stress that the factors discussed previously should have weaker impacts on earnings management for the book-building period compared with the online fixed-price offering period if those factors really motivate IPO firms to manage earnings for a high offering price.

3.2. Variable for earnings management

To the best of our knowledge, there are only a limited number of articles that investigate earnings management of Chinese IPO companies. Aharony et al. (2000) show that Chinese SOEs that sold shares to foreign investors tend to manipulate earnings before the IPO. Aharony et al. (2010) find that 185 Chinese firms that went public in the period 1999–2001 opportunistically used related-party sales to manage earnings upwards. Kao et al. (2009) show that government regulations affect a firm's reported performance and earnings forecasts surrounding IPO. All of those studies use a component of return on assets (net income divided by assets; hereafter denoted by ROA) as a proxy for earnings management to investigate Chinese IPOs before 2000. In contrast, Aerts and Cheng (2011) investigate the effect of causal disclosures on earnings management of Chinese IPO firms during the period 2007–2008 by using the modified Jones Model. Moreover, recent non-Chinese articles commonly use the modified Jones Model (Teoh et al., 1998a,b; DuCharme et al., 2001, 2004; Klein, 2002; Roosenboom et al., 2003; Darrough and Rangan, 2005; Othman and Zeghal, 2006). Dechow et al. (1995) examine alternative accrual computation models and argue that the modified Jones Model is the most powerful measure of earnings management. We follow those previous studies and adopt the modified Jones Model to compute an earnings management measure for our sample companies.

Teoh et al. (1998a,b) divide total accruals into current and long-term components and argue that managers are able to exert more discretion over current accruals. We calculate the current accruals (CA) as:

(1)

The current accruals can be further divided into nondiscretionary accruals, which are attributable to business conditions, rules, and institutions (Roosenboom et al., 2003) and DCA. We adopt DCA as the key variable in this research. We use the cross-sectional modified Jones model to estimate the nondiscretionary current accruals. For each IPO company, we estimate the following equation by using data of all the same industry firms excepting the IPO firm (estimation sample) at the IPO year.

(2)

where ΔSales is the change in sales and TA is total assets. Subscripts j and t denote firm and year. The nondiscretionary current accruals of an IPO firm i are computed by:

(3)

where and are estimated coefficients for Equation (2). ΔAC is the change in accounts receivables. The DCA of the IPO firm i is computed by:

(4)

3.3. Sample selection

We analyse Chinese A-share IPOs on the SHSE and the SZSE during the period 1999–2009. We obtain corporate financial and ownership structure data at the IPO year from the China Centre for Economic Research (CCER) Database. We hand-collect those data for the preceding year from the prospectus of each firm. We start our analytical period with year 1999 because the CCER database includes corporate ownership data for that year and thereafter. We do not include B-share IPOs because there were only six B-share IPOs during the period.4 We also exclude financial companies because of their different accounting statement formats. As a result of those procedures, our sample consists of 880 companies, of which 456 firms are listed on the SHSE and 424 firms on the SZSE. Our sample is more comprehensive in terms of sample size and period than that of previous Chinese articles (Aharony et al., 2000, 2010; Kao et al., 2009; Aerts and Cheng, 2011).

Panel A of Table 2 shows the distribution of our sample by calendar year. During the sample years, IPO markets were more active in 2000, 2004 and 2007. Panel B presents the industry distribution of our sample firms. Manufacturing firms account for a substantial part of the sample companies.

Table 3 shows the descriptive statistics for the entire sample. Panel A of Table 3 shows DCA for the entire sample during the IPO process. Although the mean DCA is not significantly different from zero, the median value DCA is 0.016 and significantly different from zero at the one percent level, which suggests that Chinese IPOs tend to manipulate earnings upward in the IPO year.

In our sample, only 44 firms (5 per cent) have an internationally reputable accounting firm as their auditors; about 24 per cent of our sample firms adopted the top five underwriters during the IPO process (Panel C of Table 3). After the introduction of the book-building mechanism, 349 firms (about 40 per cent) went public. In our sample, 492 firms (55.91 per cent) are SOEs, 156 of which (about 31.71 per cent) are controlled by a local government. The figure of BANKL shows that only 45 per cent of our sample companies (400 firms) issued bank debt in the year before IPO. In our sample, 357 firms (about 41 per cent) are located in the top five provinces or regions that attract foreign capital.

4.1. Univariate analyses

Table 4 shows the univariate test results. We divide sample companies into several groups upon variables and compare the mean DCA across groups. For dummy variables, we simply divide sample companies into two groups (Panel A). In Panel B, we construct groups upon nondummy variables by the following procedures: For LEVERAGE (at the year before IPO) and INDIR (at the year of IPO), sample firms are equally divided into four groups (Group 4 is the highest group). For Ch_MANAGEROWN, for which many observations take a value of zero, we establish a group that consists of firms that take a value of zero (Group 3) and then divide the remaining firms equally into two groups (Group 1 is firms with the largest reduction of managerial ownership). For BANKDEBT, for which many observations take a value of zero, we establish a group that consists of firms that take a value of zero (Group 1) and then divide the remaining firms equally into two groups (Group 3 is firms with the highest ratio of bank debt divided by liabilities).

Panel A of Table 4 shows firms that went public during the book-building period manipulated earnings downward. Although the mean DCA is not significantly different between the two subperiods, this finding is consistent with the idea that the abolition of the fixed-price offering system decreases IPO firms' incentive to manage earnings for a high offering price. Firms that issue bank debt before the IPO do not manipulate earnings upward at the point of IPO (the mean DCA is −0.012). Importantly, the mean DCA of firms that issue bank debt before the IPO year is −0.042 under the fixed-price offering method, which is significantly lower than the DCA of firms that do not issue bank debt before IPO under the same period (the mean DCA is 0.064) at the 1 per cent level. In contrast, there is an opposite relationship between the two groups under the book-building system. This finding is consistent with the notion that preferential access to bank debt mitigates financial constraints and thereby decreases the incentive of earnings management for a high offering price. The univariate test does not find a significant relationship between underwriter or auditor reputation (UNDERWRITER; AUDITOR), existence of audit committee (A_COMMITEE), IPO firms' location (COASTALREGION), SOEs (D_SOE) and DCA.

Consistent with Aharony et al. (1993), Panel B of Table 4 shows that, for the entire sample, DCA has a clear monotonic positive relationship to LEVERAGE. The difference in DCA between the highest and lowest leverage groups (0.087) is both economically and statistically significant. We also find that the highest leverage group shows a significantly larger DCA than the lowest group under the book-building period as well as the fixed-price offering period. This finding sheds doubt on the view that IPO firms with high leverage engage in earnings management to achieve high offering prices. We find a clear negative relationship between DCA and BANKDEBT, especially for the fixed-price offering period. This result is consistent with the idea that preferential access to bank debt mitigates financial constraints and thereby decreases the incentive to manipulate earnings for a high offering price. There is a positive and significant relationship between DCA and INDIR, which contradicts the view that independent directors monitor earnings management by IPO firms. The univariate test does not find a clear relationship between DCA and Ch_MANAGEROWN.

4.2. Regression results

To further test our hypotheses, we conduct regression analyses that adopt DCA as a dependent variable. Following previous studies, we include AUDITOR, UNDERWRITER, Ch_MANAGEROWN, INDIR, A_COMMITTEE, AGE, LNASSET and LEVERAGE as independent variables. We also include proxies of unique Chinese characteristics in our analysis: BOOKBUILD, D_SOE, BANKL (or BANKDEBT) and COASTALREGION. In each regression, we delete observations for which the dependent variable takes a value greater (lower) than its 0.99 (0.01) percentile to address abnormal values. When necessary independent variables are not available, the observation is also deleted from the analysis. Table 5 shows no serious correlations among independent variables.

Table 6 shows the regression results for the whole sample period. Models 1–4 engender a negative and significant coefficient on BOOKBUILD. Given that the mean DCA is 0.003, the estimated coefficient is economically significant. This finding supports our basic idea that the introduction of the book-building procedure (i.e., the abolition of the fixed-price offering system) significantly decreased Chinese IPO firms' incentive to manipulate earnings upward after controlling for various factors. We also find there is a positive and significant coefficient on LEVERAGE in all the models. This result is consistent with Aharony et al.'s (1993) US finding. All models suggest that BANKL and BANKDEBT have a negative and significant impact on discretionary current accruals. This result supports the idea that IPO firms that issue bank debt do not need to manipulate earnings upward because those firms do not face severe financial constraints. Models 1 and 2 carry an insignificant coefficient on D_SOE. On the other hand, Models 3 and 4 show a positive and significant relationship between local government control (SOE_L) and earnings management.

Regarding the other variables, AUDITOR, UNDERWRITER, Ch_MANAGEROWN, A_COMMITTEE and AGE have insignificant coefficients for the entire sample period. INDIR has a positive and significant coefficient, suggesting that firms with more independent boards manage earnings more during the IPO process. A possible interpretation is that independent directors in China are not really independent and do not effectively monitor management (Wang, 2007). In China, large IPO firms tend to manipulate earnings upward more than small firms do.

Although we find a significant effect of LEVERAGE, BANKL, BANKDEBT, SOE_L on DCA, the analysis potentially suffers from endogeneity problems. Some unobservable variables might affect both DCA and these variables. It is also likely that firms with specific characteristics tend to manipulate earnings for different objectives. The former analysis presents clear evidence that Chinese IPOs have weakened incentives to manage earnings after the abolition of the fixed-price offering system. If the former finding is really attributable to IPO firms' desire for a high offering price, we should find a weaker relationship between those variables and DCA for the book-building period. Similarly, variables on which the former analysis engenders an insignificant coefficient might affect DCA during the fixed-price offering period. To address these issues, we divide our sample into two subperiods and separately conduct regression analyses: the period of the fixed-price offering method (1999–2004) and the book-building period (2005–2009).

Panel A of Table 7 shows the regression results under the fixed-price offering method. As with the former analysis, all models show that firms that have high leverage manipulate accounting earnings more upward. However, Panel B of Table 7 (results for the book-building period) also finds a positive and significant coefficient on LEVERAGE. The coefficient becomes large in absolute value during the latter period when IPO firms should have weak incentives for earnings management for a high offering price. Panel C of Table 7 shows the regression results that include the interaction terms of several independent variables and BOOKBUILD. This analysis confirms that the LEVERAGE effect does not weaken for the book-building period. These results suggest that the positive relationship between leverage and discretionary accruals is attributable to IPO firms' nonoffering price objectives or hidden variable effects.

As with the former analysis, Panel A of Table 7 finds a negative and significant coefficient on BANKL and BANKDEBT. However, these variables lose significance in the regression for the book-building period (Panel B of Table 7); once offering price loses its direct link to accounting earnings, bank debt does not affect earnings management. The results in Panel C suggest that the impact of bank loans on discretionary current accruals significantly decreased after the abolition of the fixed-price offering system. This result provides robust evidence that bank debt availability affects Chinese IPO firms' incentives for earnings management for a high offering price.

Panels A and B show that in the fixed-price offering period the coefficient of SOE_L is positive and significant at the 10 per cent level, whereas it becomes insignificant for the book-building period. Similarly, SOE_C has a positive and marginally significant coefficient for the fixed-price offering period but does not during the book-building period. In addition, Models 1 and 2 of Panel C find a significant difference in the SOE_C coefficient between the two subperiods. The result suggests that both central and local governments make IPO firms manipulate earnings upward to attract more equity capital under the fixed-price offering method. In Models 3 and 4 of Panel C, we include a new dummy variable that takes a value of one for SOEs controlled by the central or local governments (SOE_G). These models engender a positive and significant coefficient on SOE_G, and its interaction term with BOOKBUILD (SOE_G*BOOKBUILD) has a negative and significant coefficient. Several early studies suggest that local governments have higher incentives to make IPO firms manipulate earnings than the central government does. In contrast to these studies, our results do not find a clear difference in the incentive of earnings management between the central and local governments.

Regarding other variables, Ch_MANAGEROWN, which has an insignificant coefficient in the entire period analysis, has a negative and significant coefficient for the fixed-price offering period (Panel A). However, this relationship disappears after the abolition of the fixed-price offering method (Panel B). Panel C suggests that Ch_MANAGEROWN has a significantly weaker impact for the book-building period compared with the fixed-price offering period. Consistent with Darrough and Rangan's (2005) finding, these results present robust evidence that IPO firms manipulate earnings more upward to achieve high offering prices when the manager sells more shares. All models of Panel A find that AUDITOR has a negative coefficient, and two estimations provide a significant one. On the other hand, Panel B engenders a positive and insignificant coefficient on AUDITOR. Although Panel C finds no significant difference in the AUDITOR coefficient between the subperiods, the results provide weak evidence that reputable auditors mitigate IPO firms' earnings management. Differently from previous studies, large firms with a long history present high abnormal accruals during the fixed-price offering period. A potential explanation is that firm size serves as a proxy for political influence because governments are likely to pay attention to large firms (Watts and Zimmerman, 1990). Information asymmetry does not provide a good explanation of Chinese IPOs' earnings management. We find no evidence that reputable underwriters mitigate earnings management in China.

4.3. Additional analyses

We have so far found a significant correlation between leverage and earnings management of IPO firms. Although the relation is potentially attributable to different motivations or endogeneity biases, it would be interesting to separately relate short-term and long-term debt to earnings management to further examine the leverage effect. Long-term leverage is related to agency costs of debt and future financing ability, whereas short-term leverage represents the risk of short-term liquidity shortage or the financial distress risk. We conduct a regression analysis that uses short-term debt over total assets (SLEVER) and long-term debt over total assets (LLEVER). The results in Table 8 illustrate a significant correlation between SLEVER and DCA for both fixed-price offering and book-building periods (Models 1 and 2). There is no significant difference in the SLEVER coefficient between the subperiods (Model 3). In contrast, we find no significant coefficients on LLEVER. A possible interpretation is that firms with high short-term financial distress risk manipulating earnings to conceal their situation, but not to boost offering price. Another possibility is that some hidden variables are related both to short-term debt and earnings management.

As mentioned, the P/E ratio for offering price determination declined from 50 to 20 in 2002. Under the reduced P/E ratio, Chinese IPO firms are likely to have a strong incentive to manipulate earnings to obtain high offering prices. To test the effect of this regulation change, we separately conduct a regression analysis for the three subperiods (Models 1–3 of Table 9): the fixed-price offering period, with a P/E ratio of 50 (1999–2001); the fixed-price offering period, with a P/E ratio of 20 (2002–2004); the book-building period (2005–2009). Model 4 is for the entire sample period and adds a dummy variable that takes a value of one in the period 2002–2004 (P/E_20) as well as its interaction terms with Ch_MANAGEROWN, LEVERAGE, SOE_G and BANKDEBT.

Consistent with former findings, Ch_MANAGEROWN and BANKDEBT have a negative and significant coefficient for both the two fixed-price offering periods (Models 1 and 2), which becomes insignificant during the book-building period (Model 3). Model 4 suggests that the Ch_MANAGEROWN coefficient is not significantly different between the two fixed-price offering periods but decreases significantly after the introduction of the book-building system. Interestingly, the BANKDEBT effect significantly decreased after reduction of the P/E ratio in 2002. A potential interpretation is that even IPO firms with preferential access to bank debt aggressively manipulate earnings during the reduced P/E ratio period. Indeed, Model 4 engenders a positive and significant coefficient on P/E_20, suggesting that IPO firms have stronger incentives for earnings management as a result of the decreased fixed P/E ratio. Model 4 also suggests that earnings management is not significantly different in size between the first fixed-price offering period and the book-building period. The reduced P/E ratio gives Chinese IPO firms a strong incentive for earnings management. The BANKDEBT coefficient also significantly decreases after the introduction of the book-building system. Table 9 shows that SOE_G has a significant coefficient only for the first subperiod. However, Model 4 finds no significant difference in the SOE_G coefficient between the two fixed-price offering periods. Consistent with the former finding, the effect of SOE_G significantly weakens after the abolition of the fixed-price offering method. Model 2 engenders an insignificant coefficient on LEVERAGE during the reduced P/E ratio period. However, we do not find a significant difference in LEVERAGE coefficients between the subperiods. Overall, the results are consistent with our main arguments.

As mentioned earlier, Aharony et al. (2000) and Kao et al. (2009) adopt a different proxy for earnings management, which is defined as a component of ROA. Following them, we compute another earnings management measure as net income minus income from core operations, divided by 1-year lagged assets (noncore ROA). In our sample, the mean (median) noncore ROA increased from −15.10 per cent (−12.21 per cent) for the year preceding IPO to −1.02 per cent (−7.70 per cent) for the IPO year. However, the noncore ROA does not show a significant reduction in the 3 years following IPO. This finding is inconsistent with the idea that firms manipulate earnings upward by borrowing future earnings. The finding is also inconsistent with Aharony et al.'s (2000) finding.5 We interpret that the ROA-based earnings management measure does not work well for our sample period.