Cross-sectional determinants of post-IPO stock performance: evidence from China
This paper was previously titled ‘Investor Sentiment, Governance Mechanisms and Post-IPO Performance in China’. We thank Robert Faff (the editor) and an anonymous referee for their helpful suggestions. We also thank Charles Shi, Anna Vong, Desmond Yuen and conference participants at the European Financial Management Association 2007 Annual Meeting at Vienna and the Seventh (2007) Accounting and Finance Conference at Xiamen University for valuable comments on our manuscript. All remaining errors are ours.
Abstract
This paper examines the cross-sectional determinants of post-IPO long-term stock returns in China. We document that the aftermarket P/E ratio has the most robust negative association with post-IPO stock returns. The negative relation indicates that the market corrects the aftermarket overvaluation of IPO firms in the long run. Underwriter reputation has a positive effect on post-IPO stock returns, while board size has a negative impact, consistent with the views that reputable underwriters mitigate the information asymmetry in IPO pricing and over-sized boards reduce the effectiveness of corporate governance. However, we find little evidence indicating that the equity ownership structure is significantly associated with post-IPO stock returns.
1. Introduction
We examine the determinants of post-IPO long-term stock returns in China. While it has been well documented that newly listed firms underperform their benchmarks in the USA (Ritter, 1991; Loughran and Ritter, 1995, 1997), evidence of post-IPO underperformance in China is mixed. Studying IPOs between 1990 and 1993, Mok and Hui (1998) find that underpriced IPOs listed on the Shanghai Stock Exchange earn a post-IPO market-adjusted return insignificantly different from zero, while overpriced IPOs earn a significantly positive post-IPO market-adjusted return. By contrast, Chan et al. (2004) examine both stock and operating performance of IPOs between 1993 and 1998, and find that A-share IPOs in China only slightly underperform the size and/or B/M benchmarks, while B-share IPOs outperform the benchmarks.1 In addition, post-IPO returns are shown to reflect changes in operating performance. However, both studies aim to document the average long-term performance of IPOs in China and compare the difference in performance between A-share and B-share IPOs. Surprisingly, there have been few studies looking into the cross-section of post-IPO stock returns.2
The major objective of this study was to examine the cross-sectional determinants of post-IPO stock returns in China, especially from the perspective of mispricing in the aftermarket. Several factors suggest that shares of newly listed firms in China are likely to be mispriced. Unlike the US bookbuilding process in which underwriters interact frequently with potential investors to solicit information and price shares according to investors’ response, application, price setting and allocation of IPOs in China are tightly regulated by administrative policies rather than economic rules. The offer price in China is fixed and is set as a multiple of earnings per share, and shares are predominantly allocated by lot drawing. In addition, as the government imposes a limit on the aggregate value of new shares that can be issued every year, the supply of new shares hardly meets the demand from investors in China. The absence of a feedback mechanism, compounded by the artificially limited supply of new shares, might help induce an overvaluation of newly listed shares in the immediate aftermarket. We thus use aftermarket valuation, measured by P/E ratio, as a proxy for misvaluation driven by market sentiment. To the extent that the misvaluation is corrected by the market in the long run, we expect a negative relation between aftermarket valuation (P/E ratio) and long-term stock performance in China.
The creation of multiple classes of shares with different trading restrictions is another special feature of China’s stock market. In China, the state and its related parties (legal persons) still closely control most listed companies and their ownership has been non-tradable and non-transferable until recently. Given the fact that their objectives are not necessary value maximization, previous studies document that ownership structure explains cross-sectional variation in post-IPO performance. However, they focus mainly on operating performance but not stock performance, and they document inconclusive results regarding the effects of state ownership on post-IPO operating performance.3
We also consider two potentially important factors, namely, underwriter reputation and the size of board of directors, which have been ignored by prior studies on post-IPO stock performance in China. Using US data, Carter et al. (1998) show that reputable underwriters can mitigate the long-term underperformance of IPOs by reducing the information asymmetry in IPO pricing. Yermack (1996) and Eisenberg et al. (1998) suggest that excessively large boards are dysfunctional and value destructive. We thus expect that both underwriter reputation and the size of board are related to the post-IPO stock performance in China.
Our paper helps address a void in the literature on IPOs in China by examining the effects of the aforesaid factors on the post-IPO stock performance, using a more extensive sample and more state-of-the-art empirical techniques to check the robustness of results. We measure post-IPO stock returns over a 3 year window after the IPO.4 Apart from the abovementioned determinants of post-IPO stock returns, we also control for share turnover and risk factors. In the light of the criticisms by Fama (1998) and Mitchell and Stafford (2000) regarding the bad-model problems in previous long-term event studies, we examine post-IPO stock performance using three different approaches: the buy-and-hold return approach, the Fama–MacBeth regression approach and the calendar-time regression approach.5 All three approaches yield qualitatively similar results as follows:
-
First, aftermarket valuation matters. Firms with a higher aftermarket P/E ratio earn a lower post-IPO long-term return, consistent with Purnanandam and Swaminathan (2004) that overvalued IPO firms underperform their benchmarks in the long run.6
-
Second, underwriter reputation and board size matter. Underwriter reputation is positively related to post-IPO stock performance and board size is negatively related to post-IPO stock returns. Both findings are consistent with previous studies that reputable underwriters can reduce the information asymmetry problems in IPO pricing, thereby reducing long-term underperformance (Carter et al., 1998) and that over-sized boards tend to reduce the effectiveness of corporate monitoring (Yermack, 1996; Eisenberg et al., 1998).
-
Third, risk factors matter. IPOs of smaller firms and firms with higher B/M ratios earn higher post-offering returns, consistent with Wang (2004) that both size and B/M ratios explain stock returns in China. Share turnover also predicts negatively future stock returns but in a less consistent manner.
-
Fourth, ownership structure is not a significant determinant of post-IPO stock performance. In contrast to Sun and Tong (2003) and Wei et al. (2003) who document a negative relation between state ownership and post-IPO operating performance, we do not find a significant effect of state ownership or legal-person ownership on post-IPO stock performance. This suggests that the effects of ownership structure on operating performance might have been incorporated into the aftermarket stock prices.
Overall, our empirical results demonstrate that post-IPO stock performance in China is affected by a set of proxies for firm valuation, corporate governance, underwriter reputation and risk factors. Among all the factors considered, aftermarket P/E ratio offers the most consistent results across different estimation methods and estimation periods. The negative relation between the P/E ratio and post-IPO stock returns suggests that IPOs in China are mispriced in the aftermarket, and the misvaluation is corrected in the long run. On the other hand, the findings that underwriter reputation and board size also have significant impacts on post-IPO performance suggest that investors in Chinese stock markets do factor into their valuation the impact of corporate governance and underwriting efforts.
The remainder of the paper is organized as follows: Section 2 briefly describes China’s state-owned enterprise (SOE) reforms and the evolution of its primary market. In Section 3, we review the related literature and develop our hypotheses. Section 4 describes the data and methodology. Section 5 reports summary statistics of key explanatory variables. Section 6 presents subgroup analysis on post-IPO returns in a calendar-time regression setting and examines the determinants of post-IPO stock performance. Section 7 concludes the paper.
2. China’ s SOE reform and IPO market
The main objective of SOE reform in China is to improve the operation of unprofitable, inefficient state-owned enterprises (SOEs) by ultimately separating the ownership from control. With the exception of some industries deemed of strategic importance to national interests, the Chinese government aims to fully float its ownership to the stock market. As an intermediate stage of the reform process, the Chinese government sells a small portion of ownership to public investors and retains a significant percentage of ownership in the SOEs to help ensure that certain social objectives, such as maintaining the employment level, can be achieved. In early years, the application for going public was constrained by a large number of factors. For example, firms were required to have two consecutive years of positive operating income (Aharony et al., 2000). Additionally, the government sets a target for the aggregate value of new shares issued and allocates the quota to provinces and municipalities (Chan et al., 2004). Therefore, applications had to be approved and recommended by the provincial governments before going to the China Securities Regulatory Commission (CSRC hereafter) for initial review. When the CSRC considered the applications, besides listing requirements and profitability of the applicants, it sought advice from other departments on whether the applicants belonged to national industrial policies and ‘protected’ industries such as petrochemicals, energy and raw materials (Aharony et al., 2000).
The pricing of IPOs in China is also regulated. Before 1999, the pricing method was stipulated as the ‘multiplier’ method. The offer price was set to be the product of net earnings per share and a chosen multiplier, with the multiplier validated by the CSRC (Tian and Megginson, 2006).7 Post-1999, in determining the offer price, the issuing firm needs to consider the industrial situation, the company development prospects and the P/E ratio in the secondary market. However, as the validated multiplier was usually set to a level significantly lower than the current market PE, most IPOs were still deeply underpriced relative to ‘intrinsic values’. In addition, the offer price was usually chosen 1–2 months before the official listing date, a period much longer than the elapsed time for developed countries.8
The IPO share allocation system has also undergone several changes over time. Before 1999, IPO shares were mainly allocated by a fixed price method in which the offer price was set fixed prior to the IPO. Ma and Faff (2007) document that several fixed price procedures existed in 1990s. In the early- to mid-1990s, IPO shares were allocated based on a lot drawing system in which subscription warrants (limited or unlimited) are sold to potential investors. Later, new fixed price procedures were introduced to allow investors to bid through the electronic trading system, to bid with deposits (as opposed to purchase warrants), and to offer shares to investors of secondary market. In 1999, with the rise of institutional investors, the CSRC allowed large companies to adopt an offering method combining lot drawings (for public investors) and placements to legal entities, a method similar to that of Hong Kong.9
3. Review of related literature and development of hypotheses
3.1. Firm valuation and post-IPO performance
As discussed in Section 2, compared with the US bookbuilding method, IPOs in China lack a feedback system to allow underwriters to obtain useful information from investors for setting the offer price and offer size. Particularly during early years, the offer prices of newly listed shares were artificially fixed at low levels, and the allocation of shares was based on lot drawing. Therefore, investors had low incentive to acquire new information. As such, the lack of information resulted in large uncertainty regarding true firm value and IPOs were severely underpriced, with underpricing being largely affected by market conditions and uncertainty during the listing period (Ma and Faff, 2007).10
The problem was compounded by the limit set by the government on the aggregate value of new shares that could be issued every year. As newly listed firms sell only a small portion of ownership to the public, the supply of new shares could not meet the demand from investors in China who have very few alternative investment choices due to stringent capital controls. When too much money was chasing too few shares, it was not uncommon for stock prices of newly listed companies to jump by more than +100 per cent on the first day of trading.
The above arguments lead us to expect that IPOs in China could be largely overpriced in the immediate aftermarket due to the joint effect of a lack of information production in the pricing process and the government policy to limit the supply of shares. If the market cannot fully correct mispricing immediately, we expect to observe a negative relationship between aftermarket valuation and long-term stock returns. Indeed, even in more developed markets, mispricing exists of IPO shares in the aftermarket. In contrast to previous evidence on IPO underpricing, Purnanandam and Swaminathan (2004) find that most IPOs in the USA are indeed overpriced when they use alternative benchmarks based on price multiples.11 Examining a sample of French IPOs, Derrien (2005) shows that overpricing and positive first-day returns are not necessarily inconsistent, in the presence of overoptimistic noise traders. Collectively, the two studies suggest that mispricing in IPOs persists and that it has a significant impact on post-IPO stock returns.
In the light of Purnanandam and Swaminathan (2004) and Derrien (2005), we hypothesize that:
H1: firms with a higher aftermarket P/E ratio at IPO, earn lower future stock returns.
We use P/E ratio to proxy for firm valuation because it is the ratio that issuers and regulators are concerned about in setting the offer price. However, as IPOs in China were priced based on the multiplier method mentioned above, the P/E ratio implied by the offer price may reflect neither the fundamental value nor overvaluation. Therefore, unlike Purnanandam and Swaminathan (2004) who use the offer P/E ratio, we use the aftermarket P/E ratio which is equal to the first-day closing price, divided by the fully diluted earnings per share. In addition, it is possible that an IPO is overvalued because of exuberant market sentiment attracted by all firms in a particular industry. Therefore, we also include the median P/E ratio in the industry, to control for the industry valuation.
3.2. Other determinants of post-IPO performance
Apart from the aftermarket P/E ratio, we also consider the following factors that have been ignored by prior studies on post-IPO performance in China stock markets.
The first factor is underwriter reputation. While previous studies find that IPOs generally underperform their benchmarks after offering of shares, Carter et al. (1998) find that IPOs managed by more reputable underwriters outperform those led by less reputable underwriters. Chemmanur and Fulghieri (1994) argue that by managing IPOs which perform better over the long term, investment banks gain reputation. Therefore, we hypothesize that:
H2: post-IPO stock performance is positively associated with underwriter reputation.
The second additional factor is board size. A number of studies suggest that excessively large boards may be dysfunctional. Yermack (1996) and Eisenberg et al. (1998) show that having eight or fewer members on the board would be optimal for firm value and performance. In addition, firms with larger board size also tend to be slower in replacing CEOs in the face of declines in performance. We thus hypothesize that:
H3: post-IPO stock return is negatively related to the size of board.
The third additional factor is ownership structure. The situation in China is complicated by the fact that non-tradable ownership is further subdivided into state ownership, legal-person ownership and employees’ ownership. While it is argued that legal persons have interests more in line with outside investors, their close relation with the government may create conflicts of interest with other shareholders. Empirical results are also mixed in the literature. While Sun and Tong (2003) find that state ownership is negatively related to post-IPO operating performance, they also find that ownership held by legal persons, who have a close relationship with the government, predicts positively post-IPO operating performance. Wang (2005) finds that state ownership does not have significant explanatory power on the change in operating performance around IPO, but legal person ownership has a curvilinear effect on the change in performance. Given the mixed results in the literature, we do not propose a directional prediction on the effect of ownership structure on post-IPO performance. We thus hypothesize (in null form) that:
H4: post-IPO stock return is not associated with ownership structure.
4. Data and construction of variables
We analyse all listed companies that conducted initial public offerings after 1993.12 We exclude IPOs of B shares because Chan et al. (2004) show that A-share and B-share IPOs behave very differently in terms of underpricing and post-IPO performance. Following common practice, we also exclude companies in financial industries. Basic information on IPOs, financial information, stock returns, ownership information, board information and underwriter information are obtained from the Chinese Stock Market and Accounting Research (CSMAR) Database. Following Chan et al. (2004), we also exclude firms that have an elapsed time between offering date and listing date which exceeds 360 days. The final sample consists of 1194 IPOs.
4.1. Post-IPO stock performance
We evaluate the IPO stock performance using three alternative approaches: (i) pooled regressions using buy-and-hold returns; (ii) Fama–MacBeth regressions using post-IPO monthly returns; and (iii) the α from the Fama–French three-factor model.
To compute the buy-and-hold returns, we assume that the IPO stock is purchased at the end of the listing month and held for a period of 12 or 36 months. For each IPO firm, we also find a matched seasoned firm based on market capitalization (size) and book-to-market ratio of equity (B/M).13 To be included as a candidate for a matched firm, the seasoned firm has to have existed in the CSMAR database for at least 24 months at the time of an IPO listing. Absolute percentage differences in size and B/M, respectively, are calculated between the IPO firms and all potential candidates. The one with the smallest sum of the two differences is the matched firm. In our pooled regression analysis, we also include the buy-and-hold return of the matched firm as a control in the regression models that explain post-IPO buy-and-hold return.
Fama–MacBeth regressions are also estimated for post-IPO monthly stock returns, from the first month to the 36th month after listing. Therefore, there are 36 OLS regressions in total. In each regression j, the dependent variable is the stock return in month j after listing. To control for general market movements, we include the market index return (sourced from CSMAR) in each regression. Average coefficients are then calculated from the OLS regressions.
Fama (1998) and Mitchell and Stafford (2000) argue that the traditional buy-and-hold return analysis in long-term event studies is seriously flawed because the systematic errors associated with the bad model problems are compounded over a long horizon. In addition, positive correlation among long-term returns of different stocks leads to non-trivial inflation of test statistics. Accordingly, they advocate the use of the calendar-time regression method in long-term event studies. Specifically, for each month, firms that went public during the previous 36 months are included in the portfolio formation. A calendar-time portfolio is formed by those stocks (equal weighting) and monthly portfolio returns are calculated. Portfolio returns are then regressed on the Fama–French factors to obtain the estimated portfolio abnormal performance measure (α). We also take subsamples of IPO firms according to different firm characteristics to examine the effects of those characteristics on return performance.
We construct the Fama–French factors based on Wang and Chin (2004). In each July, firms are ranked according to firm size and B/M ratio, where firm size is the market capitalization in June, and B/M ratio is the book value of equity in the fiscal year ended in last calendar year divided by the market capitalization in last December. Firms are then divided into two equal groups (S and B) according to firm size. In an independent sorting, firms are also divided into three groups (L, M and H) according to their B/M ratio. Firms in the top (bottom) 30 percentile are assigned to H group (L group). Thus, we have six portfolios, S/L, S/M, S/H, B/L, B/M and B/H and monthly value-weighted returns are calculated for each of them. The HML factor is defined as the average returns of S/H and B/H, minus the average returns of S/L and B/L. The SMB factor is defined as the average returns of S/L, S/M and S/H, minus the average returns of B/L, B/M and B/H.
As the Chinese government bond market was developed much later than its stock markets, we use the 3 month household deposit rate as a proxy for risk-free interest rate. Wang (2004) shows that the common return factors in China work in a way similar to those in broader financial markets, confirming the legitimacy of applying the three-factor model to China’s stock market. In unreported results, we find that the average monthly returns of the HML and SMB portfolios are 0.45 per cent and 1.07 per cent, respectively.
4.2. Key explanatory variables
- 1
Aftermarket P/E ratio (PE), defined as the closing price on the first trading day divided by after-tax profit per share. To mitigate the impact of outliers, we set the P/E ratio to a missing value if it is greater than 1000.14 In our preliminary data check, all sample firms have positive earnings in the year prior to their IPOs, consistent with Aharony et al. (2000) who state that only firms with two consecutive years of positive operating income are eligible to go public. Therefore, negative net profit is not a concern in the calculation of P/E ratio.
- 2
Industry-median P/E ratio (IND_PE), defined as the industry-median P/E ratio at the end of the listing month of an IPO, where the P/E ratio is defined as market capitalization at the end of the listing month divided by net income in the most recent fiscal year. We exclude firms that have a listing history shorter than 2 years.
- 3
Underwriter reputation measured by the number (UW_CNT) or the value (UW_VAL) of IPOs (co-)managed by the underwriter. We use a measure of underwriter reputation based on: (i) the number of IPOs (co-)managed by an underwriter throughout the database (Kirkulak and Davis, 2005) and (ii) the value of IPOs (co-)managed by an underwriter as a percentage of total value of offerings throughout the database (Megginson and Weiss, 1991).15 We consider both lead managers and co-managers in our measurement of underwriter reputation. If an IPO is co-managed by N underwriters, then the count (or the value) for an underwriter in this deal will be multiplied by 1/N. The underwriter reputation for an IPO is the average reputation (based on count or market share) of all lead- and co-managers.
- 4
Board size (BOARD), defined as the number of directors on the board at the time of IPO.
- 5
Post-IPO state ownership (STATE) and legal-person ownership (LP).
5. Summary statistics and correlation analysis
5.1. Summary statistics
Table 1 presents the summary statistics of the key explanatory variables for the full sample, from 1993 to 2004. The average aftermarket P/E ratio (PE) is 71.9 and the median ratio is 53.4, higher than the industry mean and median values. The IPO underwriting business of China is rather competitive. Even the largest player captures only 4.4 per cent of the market, in terms of both the value of shares underwritten and the number of IPOs underwritten. The median board size (BOARD) is nine. The sum of state ownership and legal-person ownership averages at about 65 per cent, consistent with other studies (see Aharony et al., 2000; Sun and Tong, 2003) which show that the state, legal entities and employees usually altogether hold more than 60 per cent of the ownership of the privatized companies after listing. The mean book-to-market ratio (BM) is 0.25 and the median is 0.23, similar to the values documented by Wang (2004). Finally, on average, 57.5 per cent of tradable shares change hands on the first trading day, a turnover comparable with that for the US market.18
| Mean | Median | Min | Max | N | |
|---|---|---|---|---|---|
| Aftermarket P/E ratio (PE) | 71.9 | 53.4 | 8.3 | 693.1 | 1155 |
| Industry-median P/E ratio (IND_PE) | 64.8 | 52.8 | 9.8 | 563.5 | 1110 |
| Underwriter reputation measured by the number of IPOs (UW_CNT) | 17.7 | 15.0 | 1.0 | 52.8 | 1184 |
| Underwriter reputation measured by the value of IPOs (UW_VAL) | 1.4% | 1.2% | 0.0% | 4.4% | 1184 |
| Board size (BOARD) | 9.7 | 9.0 | 4.0 | 20.0 | 1170 |
| Post-IPO state ownership (STATE) | 43.6% | 51.6% | 0.0% | 88.9% | 1194 |
| Post-IPO legal-person ownership (LP) | 21.1% | 11.2% | 0.0% | 90.7% | 1194 |
| Firm size in RMB millions (SIZE) | 2034.7 | 884.0 | 47.4 | 365,884.3 | 1193 |
| Book-to-market ratio (BM) | 0.25 | 0.23 | 0.02 | 2.18 | 1103 |
| First-day tradable share turnover (TURNOVER) | 57.5% | 59.2% | 0.9% | 198.1% | 1192 |
| Modified Altman’s Z-score (AZ) | 1.32 | 1.26 | −0.31 | 5.53 | 1194 |
- Summary statistics for the key variables are reported. Basic information on IPOs, financial information, stock returns, ownership information, board information and underwriter information all come from the Chinese Stock Market and Accounting Research (CSMAR) Database. Variables include (a) aftermarket P/E ratio (PE), defined as the P/E ratio given by CSMAR, multiplied by offer price and divided by the first-day closing price; (b) Industry-median P/E ratio (IND_PE), defined as the industry-median P/E ratio at the end of the listing month of an IPO; (c) underwriter reputation, measured by the number (UW_CNT) or the value (UW_VAL) of IPOs (co-)managed by the underwriter over the sample period, divided by the total number or value of IPOs in the sample period; (d) board size (BOARD), defined as the number of directors in the board at the time of IPO; (e) post-IPO state ownership (STATE), ownership held by the central or local government; (f) post-IPO legal-person ownership (LP), ownership held by other state-owned enterprises and non-bank financial institutions; and (g) firm size (SIZE) in RMB millions, defined as market capitalization implied by the offer price; (h) book-to-market ratio (BM), defined as pre-IPO book value of equity plus the proceeds from the offering, divided by the market capitalization at the end of the listing month; (i) first-day tradable share turnover (TURNOVER), defined as the number of shares traded on the first trading day divided by the total number of tradable shares at the end of listing month; (j) modified Altman’s Z-score (AZ).
5.2. Pairwise correlations among key explanatory variables
Table 2 reports pairwise correlations among key explanatory variables. The aftermarket P/E ratio is statistically uncorrelated with underwriter reputation and board size, suggesting aftermarket investors are relatively unconcerned about the roles of underwriters and the board of directors in monitoring the IPO firms. The P/E ratio is positively related to legal-person ownership. This is consistent with Sun and Tong (2003) that legal-person ownership has a positive effect.
| ln(PE) | ln(UW_CNT) | UW_VAL | ln(BOARD) | STATE | |
|---|---|---|---|---|---|
| ln(UW_CNT) | −0.02 (0.40) | ||||
| UW_VAL | −0.01 (0.77) | 0.84*** (0.00) | |||
| ln(BOARD) | −0.04 (0.16) | −0.01 (0.82) | 0.01 (0.69) | ||
| STATE | −0.04 (0.20) | 0.20*** (0.00) | 0.19*** (0.00) | 0.04 (0.17) | |
| LP | 0.06** (0.03) | −0.10*** (0.00) | −0.10*** (0.00) | −0.04 (0.16) | −0.87*** (0.00) |
- Data come from the Chinese Stock Market and Accounting Research (CSMAR) Database. Pearson correlations are reported. The aftermarket P/E ratio (PE) is defined as the P/E ratio given by CSMAR, multiplied by offer price and divided by the first-day closing price. Underwriter reputation is measured by the number (UW_CNT) or the value (UW_VAL) of IPOs (co-)managed by the underwriter over the sample period, divided by the total number or value of IPOs in the sample period. Board size (BOARD) is the number of directors in the board at the time of IPO. Post-IPO state ownership (STATE) is the level of ownership held by the central or local government after IPO. Post-IPO legal-person ownership (LP) is the level of ownership held by other state-owned enterprises and non-bank financial institutions. Values are significant at *10 per cent, **5 per cent and ***1 per cent levels.
Both measures of underwriter reputation (UW_CNT and UW_VAL) have similar correlations with other variables. Underwriter reputation is associated with higher state ownership, while the legal-person ownership and underwriter reputation are negatively related. Board size is uncorrelated with underwriter reputation and ownership structure variables.
In all cases, state ownership and legal-person ownership have contrary relationships with the other variables. While it is consistent with previous studies that state ownership and legal-person ownership have different impacts on IPO valuation and post-IPO operation performance, it could also suggest that firms controlled by the state have characteristics very different from firms controlled by legal persons.
6. Post-IPO stock performance
6.1. Summary statistics on post-IPO performance
Table 3 reports summary statistics on post-IPO stock performance.19 We calculate 36 month buy-and-hold returns starting at the end of the listing month. Then the return to a company matched by (i) size, (ii) B/M and (iii) size and B/M is subtracted from the associated raw IPO return. In general, the stocks of IPO firms only slightly underperform their benchmarks, ranging from −2.7 per cent to −7.8 per cent for 36 months post-IPO. The finding is consistent with Chan et al. (2004) that A-share IPOs only slightly underperform their benchmarks.
| Mean | Median | N | |
|---|---|---|---|
| 36 month stock return, size adjusted | −6.2% | −3.9%*** | 1130 |
| 36 month stock return, B/M adjusted | −2.7% | −4.2% | 1023 |
| 36 month stock return, size and B/M adjusted | −7.8%* | −3.0% | 1020 |
- Stock returns and financial data come from the Chinese Stock Market and Accounting Research (CSMAR) Database. Post-IPO benchmark-adjusted returns are reported. For each IPO, buy-and-hold return is measured over a 36 month period, starting at the end of the listing month. The buy-and-hold return to the IPO is adjusted for three different benchmark returns: (i) the return to a size-matched firm, (ii) the return to a B/M-matched firm and (iii) the return to a size- and B/M-matched firm, over the same period. Values are significant at *10 per cent, **5 per cent and ***1 per cent levels.
6.2. Post-IPO performance adjusted using the Fama–French three-factor model
While Table 3 shows that IPOs slightly underperform their benchmarks for 36 months after their offerings, these figures might be misleading. Specifically, as Fama (1998) and Mitchell and Stafford (2000) argue, the test statistics tend to be biased due to the correlations among overlapping return series. Therefore, they suggest using a calendar-time portfolio approach to solve the problem. Specifically, for each calendar month, a portfolio is formed by a set of stocks which were listed within a 36 month period prior to the portfolio formation. Equally weighted return is then calculated for the portfolio in each month. Monthly returns are regressed on the three factors, HML, SMB and the market return.
Table 4 reports the coefficients of calendar-time regressions. As the first IPO enters our sample in January 1993 and exits in January 1996, we exclude the calendar-time portfolios from January 1993 to December 1995, to ensure our results are not affected by those months with few observations. Therefore, each calendar-time regression is performed on 108 observations, from January 1996 to December 2004.
| Full sample | Group 1 | Group 2 | Group 3 | Group 4 | Group 5 | (5) − (1) | |
|---|---|---|---|---|---|---|---|
| Panel A: By aftermarket P/E ratio (1 = lowest, 5 = highest) | |||||||
| Excess market return | 0.995*** (52.5) | 0.990*** (41.7) | 1.006*** (37.4) | 1.027*** (35.9) | 0.966*** (41.8) | 0.991*** (46.8) | 0.001 (0.0) |
| SMB | 0.399*** (10.4) | 0.159*** (3.3) | 0.305*** (5.6) | 0.523*** (9.1) | 0.437*** (9.4) | 0.534*** (12.5) | 0.375*** (6.8) |
| HML | −0.074** (−2.3) | 0.197*** (4.8) | −0.057 (−1.2) | −0.146*** (−3.0) | −0.261*** (−6.6) | −0.214*** (−5.9) | −0.411*** (−8.9) |
| Constant | 0.002 (1.0) | 0.004** (2.1) | 0.003 (1.4) | 0.003 (1.1) | 0.002 (0.8) | −0.003* (−1.7) | −0.007*** (−3.2) |
| Observations | 108 | 108 | 108 | 108 | 108 | 108 | 108 |
| R 2 | 0.97 | 0.95 | 0.94 | 0.94 | 0.95 | 0.96 | 0.50 |
| Group 1 | Group 2 | Group 3 | Group 4 | Group 5 | (5) − (1) | |
|---|---|---|---|---|---|---|
| Panel B: By underwriter reputation (UW_VAL) (1 = lowest, 5 = highest) | ||||||
| Excess market return | 1.035*** (31.9) | 1.028*** (40.1) | 0.975*** (43.9) | 0.944*** (46.6) | 0.992*** (49.7) | −0.043 (−1.5) |
| SMB | 0.586*** (8.9) | 0.491*** (9.5) | 0.354*** (7.9) | 0.307*** (7.5) | 0.243*** (6.0) | −0.343*** (−6.1) |
| HML | 0.060 (1.1) | −0.004 (−0.1) | −0.037 (−1.0) | −0.239*** (−6.9) | −0.174*** (−5.1) | −0.234*** (−4.9) |
| Constant | 0.000 (0.0) | 0.001 (0.5) | 0.001 (0.6) | 0.001 (0.5) | 0.004** (2.6) | 0.004* (1.9) |
| Observations | 108 | 108 | 108 | 108 | 108 | 108 |
| R 2 | 0.93 | 0.95 | 0.96 | 0.96 | 0.96 | 0.47 |
| Group 1 (Size ≤ 7) 283 firms | Group 2 (8 ≤ Size ≤ 9) 422 firms | Group 3 (10 ≤ Size ≤ 11) 259 firms | Group 4 (Size ≥ 12) 206 firms | (4) − (1) | |
|---|---|---|---|---|---|
| Panel C: By board size (1 = smallest, 5 = largest) | |||||
| Excess market return | 0.999*** (39.1) | 1.015*** (44.7) | 0.992*** (49.0) | 0.948*** (44.4) | −0.052* (−1.9) |
| SMB | 0.407*** (7.9) | 0.426*** (9.3) | 0.320*** (7.8) | 0.414*** (9.6) | 0.007 (0.1) |
| HML | −0.094** (−2.2) | −0.078** (−2.0) | −0.136*** (−3.9) | −0.032 (−0.9) | 0.062 (1.3) |
| Constant | 0.004** (2.1) | 0.002 (1.0) | 0.001 (0.9) | −0.002 (−1.1) | −0.006*** (−2.8) |
| Observations | 108 | 108 | 108 | 108 | 108 |
| R 2 | 0.95 | 0.96 | 0.96 | 0.96 | 0.04 |
| Group 1 | Group 2 | Group 3 | Group 4 | Group 5 | (5) − (1) | |
|---|---|---|---|---|---|---|
| Panel D: By state ownership (1 = lowest, 5 = highest) | ||||||
| Excess market return | 1.059*** (30.1) | 0.988*** (45.6) | 0.996*** (42.9) | 0.979*** (46.5) | 0.952*** (40.8) | −0.107*** (−3.0) |
| SMB | 0.454*** (6.4) | 0.438*** (10.0) | 0.533*** (11.4) | 0.333*** (7.8) | 0.246*** (5.2) | −0.208*** (−2.9) |
| HML | −0.076 (−1.3) | −0.093** (−2.5) | −0.051 (−1.3) | −0.099*** (−2.7) | −0.044 (−1.1) | 0.031 (0.5) |
| Constant | 0.005* (1.7) | 0.002 (0.9) | −0.000 (−0.2) | 0.000 (0.2) | 0.001 (0.4) | −0.004 (−1.5) |
| Observations | 108 | 108 | 108 | 108 | 108 | 108 |
| R 2 | 0.91 | 0.96 | 0.96 | 0.96 | 0.95 | 0.17 |
| Group 1 | Group 2 | Group 3 | Group 4 | Group 5 | (5) − (1) | |
|---|---|---|---|---|---|---|
| Panel E: By legal-person ownership (1 = lowest, 5 = highest) | ||||||
| Excess market return | 0.997*** (46.3) | 0.915*** (37.8) | 0.989*** (42.8) | 0.998*** (40.0) | 1.040*** (36.8) | 0.042* (1.8) |
| SMB | 0.318*** (7.3) | 0.354*** (7.3) | 0.417*** (9.0) | 0.513*** (10.2) | 0.440*** (7.7) | 0.122** (2.6) |
| HML | 0.042 (1.1) | −0.118*** (−2.9) | −0.089** (−2.2) | −0.008 (−0.2) | −0.149*** (−3.1) | −0.191*** (−4.8) |
| Constant | 0.001 (0.7) | −0.001 (−0.4) | 0.001 (0.3) | 0.001 (0.6) | 0.004* (1.8) | 0.003 (1.5) |
| Observations | 108 | 108 | 108 | 108 | 108 | 108 |
| R 2 | 0.96 | 0.94 | 0.95 | 0.95 | 0.94 | 0.21 |
- Stock returns come from the Chinese Stock Market and Accounting Research (CSMAR) Database. For each month, firms that went public during the previous 36 months are included for portfolio formation. A calendar-time portfolio is formed by those stocks with an equal weight assigned to each stock and monthly portfolio excess return minus the risk-free interest rate is calculated. Portfolio excess returns are then regressed on the Fama–French three factors to get the abnormal performance measure, α. The Fama–French factors are constructed based on Wang and Chin (2004). Excess market return is the value-weighted return for all A shares listed on the two stock exchanges minus the risk-free interest rate, SMB is the return on a hedging portfolio formed by taking a long position in small stocks and a short position in large stocks. HML is the return on a hedging portfolio formed by taking a long position in high B/M stocks and a short position in low B/M stocks. As the government bond market was developed much later than the stock market, the 12 month household deposit rate is used as a proxy for risk-free interest rate. The whole sample is also divided into quintiles according to different firm characteristics, the aftermarket P/E ratio (Panel A), underwriter reputation (Panel B), board size (Panel C), state ownership (Panel D) and legal-person ownership (Panel E), in order to examine the effects of those characteristics on post-IPO stock performance. Portfolio 1 contains firms with the lowest values and Portfolio 5 contains firms with the highest values for the sorting variable. Monthly excess returns of individual portfolios are regressed on the three factors. The last column of each panel reports the three-factor model for the return on a hedging portfolio formed by buying portfolio 5 and selling portfolio 1, based on the given sorting variable. Each calendar-time regression is performed on 108 observations from January 1996 to December 2004. Values are significant at *10 per cent, **5 per cent and ***1 per cent levels.
The first column of Panel A reports the full sample regression. The full sample regression shows that there is no significant underperformance of IPO firms, after controlling for the Fama–French factors. The next five columns report the calendar-time regressions for five groups of firms ranked by aftermarket P/E ratio (PE). For each year, IPO firms are sorted into quintiles based on PE. Group 1 (Group 5) contains firms with the lowest (highest) PE. Consistent with H1, firms with higher aftermarket P/E ratios earn lower future stock returns. Across the groups, Group 1 IPO firms have a positive α, while Group 5 firms have a negative α, and both of the αs are statistically significant (5 per cent and 10 per cent, respectively). In addition, the hedging portfolio formed by taking a long position in Group 5 stocks and a short position in Group 1 stocks earns a negative monthly return of 0.7 per cent after controlling for the Fama–French factors. The inverse relation between the aftermarket P/E ratio and long-term performance suggests that the overvaluation in the immediate aftermarket is corrected in the long run.
Panel B reports regressions of subgroups formed by underwriter reputation (UW_VAL). Consistent with H2, IPOs led by the most reputable underwriters earn significantly positive abnormal returns. On the other hand, less prestigious underwriters do not lead to significant abnormal post-IPO performance. This suggests that even in China, where the stock market is still developing and investors are characterized as unsophisticated, underwriters are selective in conducting their businesses to preserve their reputation. The estimated coefficients on SMB and HML in the last column (i.e. the high versus low reputation differential) also suggest that more reputable underwriters tend to manage IPOs by larger firms and firms with better opportunities (low B/M).
Panel C reports regressions of subgroups sorted by board size. As the board size is a discrete variable, firms are divided into four groups to make sure each group has a similar number of observations. Consistent with H3, the result shows that firms with the smallest board size earn a positive and significant abnormal return, while firms with the largest board size earn insignificant abnormal return. The difference between the two groups is significant at the 1 per cent level. This suggests that a larger board tends to reduce shareholders’ value, as shown by previous studies. While board size positively correlated with firm size, the insignificant estimated coefficient on SMB in the last column suggests that size has a negligible role in explaining the difference in returns between the two groups of firms.
Panels D and E report results from regressions of subgroups formed by state ownership and legal-person ownership, respectively. Notably, the two types of ownership have opposite effects on post-IPO stock performance. Firms with the highest state ownership tend to underperform those with the lowest state ownership, while firms with the highest legal-person ownership tend to earn a higher return than those with the lowest legal-person ownership. The findings are consistent with Sun and Tong (2003) who document that state ownership has a negative impact, while legal-person ownership has a positive effect on post-IPO operating performance. However, as the αs of the two hedging portfolios are both statistically insignificant, we cannot reject H4 that post-IPO stock return is not associated with ownership structure.
6.3. Regression analysis for post-IPO stock performance
One shortcoming of calendar-time regression approach is that multiple sorting of the sample is impractical, especially for emerging stock markets where the number of listed firms is small. Therefore, it is difficult to separate one effect from another and get a clean inference from the results. Accordingly, we perform a pooled regression analysis for post-IPO stock performance. All the variables have been defined in the previous sections. In addition, we add the return to a size- and B/M-matched firm as a further control for risk factors. For all the regressions, the t-statistics are adjusted for time clustering. Specifically, the residuals are assumed to be correlated for IPOs listed in the same month but uncorrelated for IPOs listed in different months.
Table 5 reports the results from the regressions of buy-and-hold returns of different periods (12 and 36 month). Panel A reports the results for 12 month post-IPO stock return. Column (1) includes only firm risk characteristics and matched-firm returns. It shows that smaller firms and firms with higher B/M ratios earn a higher post-IPO return, consistent with the general findings regarding common risk factors of stock returns in China. On the other hand, firms with higher Altman’s Z-score earn higher returns. This is inconsistent with our expectation that firms with lower bankruptcy risk (higher Z-score) should earn lower returns, but consistent with Dichev (1998) and Lamont et al. (2003) that firms with higher bankruptcy risks and more financially constrained firms earn lower returns. Column (2) includes four additional variables for underwriter reputation, board size and ownership structure. Underwriter reputation and board size also have significant impacts on post-IPO stock returns and their signs are consistent with our findings in Table 4. On the other hand, the effects of state ownership and legal-person ownership remain statistically insignificant, regardless of model specifications and measurement periods. Column (3) includes two more proxies for firm valuation and industry valuation. Consistent with the finding from the calendar-time regression analysis, the aftermarket P/E ratio is negatively related to post-IPO stock performance. In addition, industry median P/E ratio is negatively related to post-IPO stock returns. This suggests that the negative relationship between aftermarket P/E ratio and post-IPO stock returns is also driven by the industry-wide market sentiment.
| (1) | (2) | (3) | |
|---|---|---|---|
| Panel A: Dependent variable ln(1 + 12 month post-IPO return) | |||
| Aftermarket P/E ratio | −0.11*** (−4.5) | ||
| Industry-median P/E ratio | −0.27*** (−6.5) | ||
| Underwriter reputation | 3.90*** (3.1) | 5.50*** (4.6) | |
| ln(board size) | −0.12*** (−3.0) | −0.13*** (−3.6) | |
| State ownership | −0.02 (−0.2) | 0.11 (1.2) | |
| Legal-person ownership | 0.02 (0.2) | 0.17* (1.7) | |
| ln(firm size) | −0.06*** (−3.2) | −0.07*** (−3.2) | −0.07*** (−3.7) |
| Book-to-market ratio | 0.40** (2.4) | 0.41** (2.4) | 0.04 (0.4) |
| First-day tradable share turnover | 0.18* (1.7) | 0.15 (1.4) | −0.01 (−0.2) |
| Altman’s Z-score | 0.07** (2.5) | 0.06** (2.2) | 0.04 (1.3) |
| ln(1 + matched-firm return) | 0.48*** (11.6) | 0.46*** (12.5) | 0.37*** (9.1) |
| Constant | 0.24 (1.4) | 0.51*** (2.7) | 2.11*** (8.3) |
| Observations | 1100 | 1077 | 1014 |
| R 2 | 0.27 | 0.29 | 0.38 |
| (1) | (2) | (3) | |
|---|---|---|---|
| Panel B: Dependent variable ln(1 + 36 month post-IPO return) | |||
| Aftermarket P/E ratio | −0.26*** (−6.5) | ||
| Industry-median P/E ratio | −0.22*** (−3.8) | ||
| Underwriter reputation | 0.74 (0.4) | 2.90* (1.8) | |
| ln(board size) | −0.11* (−1.8) | −0.05 (−0.9) | |
| State ownership | −0.12 (−0.7) | −0.11 (−0.7) | |
| Legal-person ownership | −0.03 (−0.2) | 0.03 (0.2) | |
| ln(firm size) | −0.14*** (−4.5) | −0.13*** (−3.8) | −0.15*** (−4.4) |
| Book-to-market ratio | 1.08*** (3.7) | 1.03*** (3.5) | 0.58*** (3.2) |
| First-day tradable share turnover | −0.02 (−0.1) | −0.02 (−0.2) | −0.13 (−1.1) |
| Altman’s Z-score | 0.24*** (4.9) | 0.24*** (5.1) | 0.24*** (4.6) |
| ln(1 + matched-firm return) | 0.52*** (15.5) | 0.52*** (15.6) | 0.42*** (11.7) |
| Constant | 0.65** (2.6) | 0.86*** (3.0) | 2.93*** (8.0) |
| Observations | 1048 | 1026 | 963 |
| R 2 | 0.45 | 0.46 | 0.51 |
- Basic information on IPOs, financial information, stock returns, ownership information, board information and underwriter information all come from the Chinese Stock Market and Accounting Research (CSMAR) Database. The dependent variable is the buy-and-hold return to IPO firm, over a 12 or a 36 month period. Explanatory variables include (a) aftermarket P/E ratio, defined as the P/E ratio given by CSMAR, multiplied by offer price and divided by the first-day closing price; (b) industry-median P/E ratio, defined as the industry-median P/E ratio at the end of the listing month of an IPO; (c) underwriter reputation, measured by the value of IPOs (co-)managed by the underwriter over the sample period, divided by the value of IPOs in the sample period; (d) board size (natural log), defined as the number of directors in the board at the time of IPO; (e) post-IPO state ownership, ownership held by the central or local government; (f) post-IPO legal-person ownership, ownership held by other state-owned enterprises and non-bank financial institutions; and (g) firm size in RMB millions (natural log), defined as market capitalization implied by the offer price; (h) book-to-market ratio, defined as pre-IPO book value of equity plus the proceeds from the offering, divided by the market capitalization at the end of the listing month; (i) first-day tradable share turnover, defined as the number of shares traded on the first trading day divided by the total number of tradable shares at the end of listing month; (j) modified Altman’s Z-score (AZ); and (k) the buy-and-hold return to a size- and B/M-matched firm. Industry dummy variables are included but not reported. t-statistics, adjusted for time clustering, are reported in parentheses. Values are significant at *10 per cent, **5 per cent and ***1 per cent levels.
Panel B of Table 5 reports the results from the regressions of 36 month post-IPO stock returns. The results are qualitatively the same as those of Panel A, except that in Panel B, underwriter reputation and board size have less significant impacts on 36 month post-IPO stock performance.
In sum, Table 5 shows that in pooled regression analysis, the aftermarket P/E consistently predicts post-IPO stock return, and ownership structure is essentially not associated with post-IPO stock return. On the other hand, the impacts of underwriter reputation and board size are relatively short lived. In other words, Table 5 consistently supports H1 and H4 across different windows of analysis, but it fails to consistently support H2 and H3.
Table 6 reports the Fama–MacBeth regressions for post-IPO monthly stock returns. Compared with the calendar-time regression approach, the Fama–MacBeth approach allows for multi-variable controls. Thirty-six regressions are performed for monthly returns from the first month to the 36th month post-IPO. To control for the general market movements, we include the market index return from CSMAR in each regression. Table 6 reports the average coefficients, with the t-statistics, from the 36 regressions.
| (1) | (2) | (3) | |
|---|---|---|---|
| 1st to 36th month | 1st to 12th month | 13th to 36th month | |
| Aftermarket P/E ratio | −0.004*** (−3.7) | −0.003 (−1.5) | −0.004*** (−3.4) |
| Industry-median P/E ratio | −0.002 (−1.1) | −0.003 (−1.1) | −0.001 (−0.6) |
| Underwriter reputation | 0.191*** (3.0) | 0.394*** (4.6) | 0.090 (1.2) |
| ln(board size) | −0.003 (−1.5) | −0.009*** (−4.7) | 0.002 (0.1) |
| State ownership | 0.006 (1.3) | 0.004 (0.9) | 0.007 (1.1) |
| Legal-person ownership | 0.010* (1.8) | 0.013 (1.5) | 0.008 (1.2) |
| ln(firm size) | −0.004*** (−5.7) | −0.004** (−2.8) | −0.004*** (−4.9) |
| Book-to-market ratio | 0.016*** (3.7) | 0.012 (1.4) | 0.018*** (3.6) |
| First-day tradable share turnover | −0.007 (−1.4) | −0.009 (−1.0) | −0.005 (−1.0) |
| Altman’s Z-score | 0.005*** (4.3) | 0.005** (2.7) | 0.005*** (3.3) |
| Market return | 0.997*** (64.5) | 0.979*** (58.3) | 1.007*** (46.6) |
| Constant | 0.049*** (5.9) | 0.063*** (3.9) | 0.042*** (4.4) |
| Number of regressions | 36 | 12 | 24 |
- The dependent variable is monthly stock return. Cross-sectional OLS regressions are run from the first month up to 36th month post-IPO, and time-series average of coefficients. Explanatory variables include (a) aftermarket P/E ratio, defined as the P/E ratio given by CSMAR, multiplied by offer price and divided by the first-day closing price; (b) Industry-median P/E ratio, defined as the industry-median P/E ratio at the end of the listing month of an IPO; (c) underwriter reputation, measured by the value of IPOs (co-)managed by the underwriter over the sample period, divided by the value of IPOs in the sample period; (d) board size (natural log), defined as the number of directors in the board at the time of IPO; (e) post-IPO state ownership, ownership held by the central or local government; (f) post-IPO legal-person ownership, ownership held by other state-owned enterprises and non-bank financial institutions; and (g) firm size in RMB millions (natural log), defined as market capitalization implied by the offer price; (h) book-to-market ratio, defined as pre-IPO book value of equity plus the proceeds from the offering, divided by the market capitalization at the end of the listing month; (i) first-day tradable share turnover, defined as the number of shares traded on the first trading day divided by the total number of tradable shares at the end of listing month; (j) modified Altman’s Z-score (AZ); and (k) contemporaneous value-weighted market return, which is calculated based on all A-shares in the market, to control for the general market movement. t-statistics are reported in parentheses. Values are significant at *10 per cent, **5 per cent and ***1 per cent levels.
In Column (1), most coefficients are consistent with the results in Column (3) of Panel B, Table 5, except that the industry-median P/E ratio is statistically insignificant in Fama–MacBeth regressions. Column (2) reports the results from regressions from the first month to the 12th month post-IPO. Again, the results are generally consistent with Column (3) of Panel A, Table 5. Column (3) reports the results for regressions from the 13th month to the 36th month post-IPO. As opposed to Column (2), underwriter reputation and board size have insignificant effects on stock returns beyond 1 year after the IPO, confirming our findings in Table 5.
7.Conclusion
This study offers a comprehensive exploration of the determinants of post-IPO stock returns in China. While there are a few studies which have examined post-IPO stock performance in a general context, studies documenting its determinants are sparse. Our study fills this gap by identifying potential factors derived from the IPO literature and previous studies on China’s stock market and testing which consistently explain post-IPO stock returns.
We find that firm valuation, proxied by the aftermarket P/E ratio, is the most robust determinant of post-IPO stock performance in China, consistent with Purnanandam and Swaminathan (2004) who document that firms with high price multiples at IPO are generally overpriced. We control for firm risk characteristics in our regression models and find that the negative relationship is robust.
We also show that reputable underwriters and board size have significant effects on post-IPO stock returns and that their effects are consistent with the results from studies for more developed markets. In addition, firm size and the B/M ratio also have significant predictive power for post-IPO stock returns. This suggests that while it is true that China’s stock market is still developing, many theories and findings which have shown to apply to the more developed markets are also relevant to the Chinese setting.
