Underwriter Reputation and Post-IPO Price Performance: New Evidence from IPO Fraud of Chinese Listed Firms

Underwriter Reputation and Post-IPO Price Performance: New Evidence from IPO Fraud of Chinese Listed Firms Qiuyue Zhang Xueyong Zhang * Current Version: January, 2017 * Qiuyue Zhang, School of Finance, Central University of Finance and Economics, Beijing, 100081; e-mail: zhangqiuyue817@foxmail.com; phone: +86 13261395551. Xueyong Zhang, School of Finance, Central University of Finance and Economics, Beijing, 100081; e-mail: zhangxueyong@cufe.edu.cn; phone: +86 13701249262. 1

Underwriter Reputation and Post-IPO Price Performance: New Evidence from IPO Fraud of Chinese Listed Firms ABSTRACT We examine the impact of underwriter reputation on post-ipo price performance of China s A-share market during the period from January 1, 2000 to March 31, 2015. Our research proposes a new measure for underwriter reputation using IPO fraud events. We find that firms managed by underwriters with a damaged reputation exhibit higher underpricing levels and poorer long-run performance. In addition, the new reputation proxy used in our research is powerful, even after controlling for three traditional reputation proxies based on relative market share, the number of IPOs managed, or the registered capital of the underwriter. Our results are robust after controlling for potential endogeneity using a difference-in-differences approach. JEL Classification: G12 G24 Keywords: IPO fraud Underwriter reputation Post-IPO price performance Difference-in-differences 1

1. Introduction Underwriter reputation provides not only a comprehensive record of each underwriter s equity-marketing history, but also a guarantee of its future product quality (Booth and Smith, 1986; Chemmanur and Fulghieri, 1994). The effect of underwriter reputation on the price performance of initial public offerings (IPOs) has been examined in a substantial body of research (see Beatty and Ritter, 1986; Booth and Smith, 1986; Carter and Manaster, 1990; Carter, 1992; Michaely and Shaw, 1994; Dong, Michel and Pandes, 2011). Empirically, a major challenge for testing the impact of underwriter reputation is that it is an elusive concept and consequentially, not directly observable. Prior IPO research has widely documented two proxies to measure underwriter reputation. Carter and Manaster (CM) (1990) first grade underwriter reputation by its relative placements in tombstone announcements. Megginson and Weiss (MW) (1991) use the proportion of each underwriter s market share as another influential proxy for underwriter reputation. Carter, Dark and Singh (1998) provide further evidence that the CM ranking and the MW measure are highly correlated. Besides these two widely used reputation measures, some studies also use the capital scale of the underwriter (Michaely and Shaw, 1994; Migliorati and Vismara, 2014), IPO mispricing (Nanda and Yun, 1997), and the number of IPOs underwritten (Su and Bangassa, 2011; Agathee, Sannassee and Brooks, 2012; Migliorati and Vismara, 2014) as additional measures of underwriter reputation. In this paper, we take a different, creative approach examining how IPO fraud events can be used to construct a totally new and effective proxy for underwriter reputation. Existing reputation proxies are closely related to underwriter reputation, but there exists some misalignment, especially when exogenous events damage underwriter reputation. A quintessential example is Citigroup. When WorldCom files for bankruptcy protection in July 2002, it is the largest accounting fraud scandal in the history of the United States and it knocks global stock markets down and raises investors doubts about the credibility of third-party certifiers. Citigroup, the key backer underwriter of WorldCom, agrees to pay $2.65 billion to settle a suit brought by WorldCom investors. Despite a great loss in its reputation, Citigroup issues more than $414 billion shares and manages 1,300 IPOs in the same year, far exceeding their competitors. In this case, relative market share as a measure for underwriter reputation may not be a good fit because it can only represent 2

the number of IPOs an underwriter has managed, which cannot be equated with the occupational ethics of the underwriter or the accuracy of the information produced. In this research, we define a bad reputation as an attempt to engage in short-run opportunistic behavior. A prestigious reputation is the willingness to avoid higher risk offerings. Empirically, we first identify all public IPO fraud events. The disclosure of IPO fraud can represent an opportunity to detect reputation damage due to underwriter s mis-certification during IPO. Evidence supporting this implication is provided by Beatty and Ritter (1986) that mispriced offerings significantly damage underwriter reputation. Beatty, Bunsis and Hand (1998) find indirect penalties to underwriters due to Securities and Exchange Commission (SEC) investigations. Song and Uzun (2004) document a negative return of underwriter s stock subsequent to a lawsuit filing. External certifier failure causes investors concerns of underwriter reputation. This allows us to classify an underwriter s reputation as damaged after its former client is reported IPO fraud. We examine the relationship between underwriter reputation and post-ipo performance using the new reputation measure. We first find a negative effect of underwriter reputation on the initial returns in both a univariate comparison and a multivariate analysis framework, and this result is consistent when controlling for the influence of both industry and market. Our finding of a negative relationship between underwriter reputation and underpricing uniquely complements earlier research that seeks to understand why a reputable underwriter can inhibit the tendency of underpricing. Information asymmetry-related theories view an underwriter as the certifying agent and information producer. An issuer hires an underwriter to certify the consistency of the offering price with available inside information (Booth and Smith, 1986; Cooney, Kato and Schallheim, 2003; Yung and Zender, 2010). Meanwhile, the underwriter produces information to reduce the heterogeneity of investor beliefs (Dong, Michel and Pandes, 2011). Reputation is a guarantee of the underwriter s quality assurance for investors and issuers. Rents earned on reputation provide incentives for underwriters to maintain their reputation (Beatty and Ritter, 1986; Fang, 2005). Therefore, reputable underwriters handle IPOs less underpriced (Carter and Manaster, 1990; Carter, Dark and Singh, 1998). The majority of studies utilize the CM ranking or MW proxy to measure underwriter reputation. Our empirical results provide strong evidence supporting those theoretical models from another perspective. 3

In addition, we conduct a deep investigation linking different long-run returns to underwriter reputation finding that when an underwriter s reputation is damaged, the firms managed by these underwriters exhibit lower long-run returns. Our findings are robust to various checks including the buy-and-hold raw returns as well as three market-adjusted buy-and-hold returns. The strong price explanation of a new underwriter reputation similarly complements early studies. Firms with favorable information about their value choose a higher-quality underwriter, creating an increasing function of firm value and underwriter quality (Titman and Trueman, 1986). Accordingly, a prestigious underwriter carefully performs due diligence and selects firms of high quality to avoid high risk issuing (Carter and Manaster, 1990; Chemmanur and Fulghieri, 1994; Booth, 2004; Hsu, Reed and Rocholl, 2010). As a result, IPO firms that hire more prestigious underwriters are likely to exhibit better long-run returns (Carter, Dark and Singh, 1998; Chan et al., 2008). It is interesting to compare how well the new reputation proxy performs relative to alternative existing underwriter reputation proxies. Based on the prior IPO-related literature, we consider three traditional proxies for underwriter reputation: relative market share (Megginson and Weiss, 1991), the number of IPOs managed by each underwriter (Su and Bangassa, 2011; Agathee, Sannassee and Brooks, 2012; Migliorati and Vismara, 2014), and the registered capital of each underwriter (Michaely and Shaw, 1994; Su and Bangassa, 2011; Migliorati and Vismara, 2014). We find that the new reputation proxy is still significant when each of these traditional reputation proxies is used as a control. One concern about our strategy, which uses disclosure of IPO fraud to decide underwriter reputation, is that the selection of firms is not random. Expected changes in underwriter reputation may cause limited choices of its future IPO clients. To tackle this endogeneity problem, we use a difference-in-differences (DID) approach as our next set of robustness tests. We further analyze how stock price performance changes with underwriter reputation surrounding the IPO fraud disclosure time. Similarly, we examine the impact of underwriter reputation on the first-day returns and long-term returns subsequent to an IPO. We document significant higher initial returns and poorer long-run performance of IPOs when an underwriter with a damaged reputation is hired. The remainder of this study is organized as follows. The following section discusses the construction of the underwriter reputation proxy. Section 3 provides summary statistics of the data, 4

Section 4 explores the impact of underwriter reputation, and Section 5 reports the robust test of an underwriter s reputation using the DID approach. Our final section concludes this study. 2. Measures of underwriter reputation In this section, we first provide the method for constructing the new underwriter reputation measure. We then analytically compare it with the CM ranking and MW measure to understand their differences. 2.1. New reputation measure Our idea is to exploit the IPO fraud event information of Chinese listed firms to obtain a new proxy for the purpose of measuring underwriter reputation. We use PingAn Securities as an example; 2010 and 2011 are the peak periods for PingAn Securities underwriting business. Both the gross proceeds and the number of IPOs managed by PingAn Securities reach the top place 1. PingAn Securities should enjoy a prestigious reputation, according to its relative market share, but in fact, the direct opposite is true. Past IPO clients of PingAn Securities, like WanMa Cable, DangSheng Tecnology, AnNi, QunXing Toy, and ShengJingShanHe allege IPO fraud, one after another, from 2009 to 2011. All these public IPO fraud events heavily damage investors confidence, raising doubt about the information PingAn Securities produced. The list of similar examples in China can easily be extended. Our research finds that, compared to the severe market discipline imposed on guilty issuers or underwriters in more mature western country markets, China fails to adequately penalize fraudulent underwriting behavior. The US Securities and Exchange Act of 1934 states that any person making a statement that is false or misleading, with respect to any material fact, shall upon conviction be fined not more than $5 million or imprisoned for not more than 20 years, or both, except that when such person is a person other than a natural person, a fine not exceeding $25 million may be imposed. It is worth noting that these US prison sentences and fines can be accumulated. On the other hand, 1 PingAn Securities ranks number one in 2010 and number two in 2011 for underwriting services and manages 39 and 34 IPOs in each of these years, respectively. The ratio of gross proceeds raised by PingAn Securities is 7.3% in 2010 with a rank of third, and 11.11% in 2011 capturing the largest relative market share. For two years running, PingAn Securities places first in the gross spread list (Wind database). 5

the Securities Law of the People's Republic of China states that where an issuer does not meet the conditions for issuance and has not started the issuance of securities, although it has obtained approval by fraudulent means, it shall be fined not less than 300,000 yuan but not more than 600,000 yuan; and, if it has started such an issuance it shall be fined not less than one percent but not more than five percent of the amount of the illegally raised funds. China s securities law continues, describing that any person directly in charge and other persons directly responsible shall be fined not less than 30,000 yuan but not more than 300,000 yuan each. Subject to the provisions of Article 189, of the Securities Law of the People's Republic of China, the chairman of WanFuShengKe, following the infamous IPO fraud scandal, is sentenced to the 300,000 yuan penalty and only imposed an administrative penalty of life of securities market exclusion, while in the US, the chairman of Enron is charged a much greater fine of $90 million and faced 235 counts of fraud. In China, the benefits from illegal actions far exceed the costs of breaking the laws, which is one of the main reasons for the higher frequency of fraud cases in the country s IPO market. And, although the China Securities Regulatory Commission (CSRC) strengthens its inspection law enforcement work to combat the frequency of IPO fraud, according to CSRC data, there are 46 firms registered and investigated on the suspicion of fraud for only 10 months from January to October 2013. Rampant IPO fraud cases in China s security market provide us with the opportunity to measure underwriter reputation. We manually collect all IPO fraud events from official announcements. Firms that commit fraud, when going public, are mainly revealed by the following two means: firstly, through official announcements published by the CSRC, Shanghai Stock Exchange (SSE), the Shenzhen Stock Exchange (SZSE), or the Securities Association of China; and secondly, through the financial media, such as Sina Finance and Netease Finance, which can effectively describe the process and results of IPO fraud based on an official announcement. To ensure the authenticity of IPO fraud cases, our research samples are based on all official announcements. We double-check IPO fraud events in the CRSC of each province, municipality, and autonomous region. The confirmation of fraud time is the earliest time when official announcements are made public. 6

The Securities Law of the People's Republic of China, which legally defines IPO fraud, did not come into effect until July 1, 1999. Our IPO fraud sample period therefore begins on July 1, 1999 and continues through to March 31, 2015. We manually collect 106 IPO fraud events with clear evidence, where 41 firms acquire the qualifications of listing mainly through false records, 24 companies though major omissions, 12 firms though fictitious profit, 5 due to false disclosure of information, and 24 for other reasons. We define an underwriter s reputation as damaged after its former client is reported the IPO fraud. 2.2. Traditional reputation measures CM ranking is estimated as the relative placement in tombstone announcements. The MW measure is taken by the proportion of its market share. However, there is no channel to provide a tombstone announcement in China or in European countries, which makes a CM ranking less directly applicable. MW is therefore widely used to measure underwriter reputation in these countries. However, Carter, Dark and Singh (1998) show that the relative market share is highly correlated with tombstone rank. To thoroughly compare underwriter reputation proxies, we select three typical, traditional reputation proxies as follows: (1) Relative market share. According to Megginson and Weiss (1991), we compute the percentage of gross spreads raised by each of the underwriters with the total gross proceeds of the entire market in a different period. When more than one underwriter jointly undertakes an IPO, the gross spreads are equally distributed among these underwriters if the actual quota is not published. (2) The number of IPOs managed. Following Su and Bangassa (2011), Agathee, Sannassee and Brooks (2012), and Migliorati and Vismara (2014), we take the number of IPOs managed by each underwriter as another reputation proxy. Similarly, when more than one underwriter jointly undertakes an IPO, the number of IPOs is equally distributed among these underwriters if the actual quota is not published. (3) Registered capital. The third reputation proxy is based on the theories of Michaely and Shaw (1994), Su and Bangassa (2011), and Migliorati and Vismara (2014) where the underwriter with greatest capital puts more attention into maintaining its reputation to avoid any loss. Therefore, we utilize registered capital as a proxy for underwriter reputation. 7

3. Data We select all Chinese firms listed in the A-share stock market from January 1, 2000 to March 31, 2015 as our initial sample, excluding IPOs in the financial industry. We utilize each IPO s trading data including: the listing date; offering price; closing price at the end of the first trading day; monthly return; market value from the CSMAR database; and financial data including asset, profit, book-to-market ratio, as well as the Shenwan industry classification index from the Resset database. Other data related to underwriter, such as gross proceeds and registered capital is obtained from the Wind database; we exclude firms with incomplete or unavailable information. The Wind database includes 108 IPOs with missing underwriter information; to supplement the corresponding information we download the official listing prospectuses from the SSE and SZSE. 3.1. Underpricing and long-run stock returns Underpricing reflects positive abnormal returns from the offering price to the close price at the first trading day (Booth and Smith, 1986; Rock, 1986; Welch, 1989; Loughran and Ritter, 1995). To study the impact of underwriter reputation on the underpricing of an IPO, we define the initial return as follows: IR = (P 1 P 0 ) P 0, where P 1 is the close price at the first trading day and P 0 is offering price. Following the research of Carter, Dark and Singh (1998), we also adjust the initial return to control for the influence of both industry and market. We define the market-adjusted initial return as adjir = (1 + IR) (1 + R M ), where R M is the value-weighted market returns of the same industry as the corresponding IPO firm from prospectus date to the first day of listing. The Shenwan industry classification index is used here. The long-term returns of the listing firms are further defined. We use the buy-and-hold strategy to test the association between underwriter reputation and the long-run performance of IPOs. Similar to Ritter (1991) and Loughran and Ritter (1995), we then compute the buy-and-hold T i=1 return, BHR i = (1 + R it ), where R it is monthly raw return of firm i in month t. Here, T represents 60 months. The buy-and-hold return is computed from the second month after an IPO. 8

In order to reflect the long-run performance of the IPO firm objectively, we next follow the calculation introduced by Loughran and Ritter (1995) and Carter, Dark and Singh (1998), using different market benchmarks to adjust for the buy-and-hold return. The market-adjusted T i=1 buy-and-hold return is defined as BHAR i = (1 + R it ) (1 + MR it ), where MR it is the contemporaneous return of the market benchmark. Here, we use an equally weighted return of all China A-share stocks and a circulating market value weighted return of all these A-share stocks to compute the equal-weighted buy-and-hold return and value-weighted buy-and-hold return. To further strengthen the robustness of the above tests, we next construct a benchmark return from 25 (5 5) equally weighted portfolios. According to Fama and French (1992; 1993), a firm s size and book-to-market ratio largely explain its stock return. However, excluding the influence of size and the book-to-market ratio can intuitively and clearly reflect the impact of an underwriter s reputation on stock price. Therefore, we divide all stocks into quintile by size and book-to-market ratio respectively to form 25 portfolios. Then we match each IPO on those two dimensions to the corresponding portfolio, using the matching portfolio s return as benchmark. Portfolio-adjusted buy-and-hold return is buy-and-hold return minus equally weighted corresponding portfolio return. T i=1 3.2. Additional variables Following previous IPO-related studies, we control the characteristics of the issuing firms and market condition in order to assess the marginal impact of underwriter reputation. The regression model includes seven additional control variables: return on asset (ROA), fee, age, leverage, scale, Growth Enterprises Market (GEM), and market. [Insert Table 1 Here] Issuer and issue characteristic variables included in our empirical analyses are defined as follows: ROA is calculated as profits divided by assets at the end of the filing year, which is measured as a percentage. Fee, is the ratio of the actual underwriter fees to the actual gross proceeds of the IPO, which is also measured as a percentage. Age, is defined as the interval year prior to the IPO minus the founding year. Leverage, is the ratio of total liabilities to total assets at 9

the end of the IPO year. Scale (i.e., the size of a firm), is the logarithm of total assets at the beginning of the IPO year. GEM takes on the value one if the IPO firm goes public on the GEM board, or zero otherwise. The market condition variable (market) represents the market sentiment, which is the number of IPOs issued in the corresponding firm s listing month. The specific definition and descriptive statistics for all variables are shown in Table 1. 4. Empirical results 4.1. Cross-section estimation results We test the relationship of underwriter reputation and post-ipo price performance of the listed firms using the following regression model: performance i = β 0 + β 1 deterioration i + β 2 traditional proxy i + β 3 ROA i + β 4 fee i + β 5 age i + β 6 leverage i + β 7 sacle i + β 8 GEM i + β 9 market i + ε i (1) where performance represents the initial returns and long-term returns of the listed firms, respectively. deterioration is the new underwriter reputation proxy, according to the reputation of the underwriter that IPO firms hire, which takes on the value one if the former IPO client of the underwriter is charged with IPO fraud, or zero if not. For comparison, we also consider three traditional underwriter reputation proxies, and each of these traditional proxies is used as a control. We perform four regressions for each dependent variable using Model (1). The first is regressed on deterioration, then we add the traditional reputation proxies separately, that is, relative market share, the number of IPOs managed, and the registered capital. For IPOs managed by underwriters whose past clients have alleged IPO fraud (labeled as IPOs that hire underwriters with a damaged reputation ), we use a propensity score matching algorithm to select the IPOs that list in the same period but hire underwriters without fraud disclosure (labeled as IPOs that hire underwriters with an intact reputation ). [Insert Table 2 Here] 10

Table 2 compares the initial returns of these two groups from January 1, 2000 to March 31, 2015. Over the entire sample period, the average initial returns of IPOs that hire underwriters with a damaged reputation is 102 percent, which is significantly higher than that of matching IPOs, 71 percent, at the 1 percent level of significance (t-stat = 3.55). Controlling for the influence of both the industry and market, the return differential between these two groups is 28.9 percent, with t-statistic 3.39, which is still statistically significant and economically crucial. The results indicate that disclosure of IPO fraud aggravates information asymmetry between the issuers and investors, and that the market generates doubts about underwriter reputation, which further induces a higher underpricing level. The validity of underwriter reputation has therefore been preliminarily verified. It is worth noting that the SSE and SZSE limit the growth of the IPO price on its first day, to prevent new shares speculation on December 13, 2013. The SSE has set an effective bidding price below 144% and above the 64% offer price. After price restrictions on the first day of trading, 114 out of 118 IPO firms draw direct trading. In order to exclude the effect of this special sample on the results, we further remove the IPO sample listed after December 13, 2013. Lines 5 and 6 of Table 2 compare the underpricing of firms listed between January 1, 2000 and December 13, 2013 which are free from the effects of the particular trading system. The results hold true after the exclusion of the special sample. [Insert Table 3 Here] Results of the t-test, by groups, shows that the IPO firms that hire underwriters with a damaged reputation exhibit higher initial returns. To further examine the marginal impact of underwriter reputation on the initial returns and to yield a more robust t-statistic, we employ a regression model and control the characteristics of firms as well as market condition. Columns (1) and (5) of Table 3 present the empirical results of Model (1) with the initial returns and adjusted initial returns as dependent variables, respectively, which includes deterioration i as the proxy to measure underwriter reputation only. The results show that the coefficient of the new reputation proxy is significantly positive, which is consistent with the theory of Carter and Manaster (1990) that the worse the underwriter reputation is, the higher the initial return. 11

Columns (2) to (4) and Columns (6) through (8) of Table 3 use the regression presented in Model (1) with the addition of the traditional underwriter reputation proxy, respectively, that is, the relative market share, the number of IPOs managed, and registered capital. The statistically insignificant coefficients for most of the regression results of these three traditional reputation proxies imply that the general way to measure underwriter reputation has limited explanatory power in the price performance of IPO firms. These results could be due to the ignorance of market discipline after underwriter s mis-certification. Investors will judge underwriter reputation at the time of the IPO fraud announcement. Therefore, it s effective and more direct to measure underwriter reputation through the disclosure of IPO fraud. We also remove IPOs that list subsequent to December 13, 2013 and regress Model (1) with the new sub-sample. We find that all the results are consistent. Based on the research of Carter, Dark and Singh (1998), firms going public and hiring an underwriter with a more prestigious reputation can earn higher long-term returns over the three-year period examined. Importantly, whether the new reputation measure constructed by the IPO fraud events can have a significant impact on the long-run performance of IPO firms is not examined in previous research. [Insert Figure 1-4 Here] [Insert Table 4 Here] Similarly, we divide the IPO firms by the reputation of the underwriters they hire. For firms that hire underwriters with a damaged reputation and firms that hire underwriters with an intact reputation, different long-term returns are conducted; these results are shown in Figures 1 to 4 and Table 4. All the figures suggest that firms managed by underwriters whose former clients are reported fraudulent substantially underperform a comparison group of matching firms over 60 months, after the first month of trading. This is based on a sample of 414 IPOs listed on China s A-share market from January 1, 2000 to March 31, 2015. In addition, when excluding the impacts of size and the book-to-market ratio, the results are consistent. The t-statistic results, shown in Table 4, illustrate that the differences in both the buy-and-hold return and market-adjusted 12

buy-and-hold returns are statistically significant. Other results imply that the IPOs that hire underwriters with a damaged reputation greatly underperform from 2 to 60 months after listing, as the equal-weighted and portfolio-adjusted buy-and-hold returns show. The mean equal-weighted and portfolio-adjusted 60 months buy-and-hold returns are -41.4 percent and -54.9 percent, respectively, compared with 20.5 percent (t-stat=2.85) and 22.1 percent (t-stat=3.05) of the matching group. We further formally use multiple regressions to investigate whether the results above are due to factors other than underwriter reputation. Meanwhile, we sequentially add the traditional reputation proxies to examine their explanation power on the long-term returns of IPO firms. Table 5 presents the empirical results of ordinary least square (OLS) regressions with buy-and-hold return, market-adjusted, and portfolio-adjusted buy-and-hold returns as dependent variables. [Insert Table 5 Here] As expected, the impact of the deterioration proxy on long-term returns is significantly negative after adjusting for other variables, which indicates that the price reaction of IPO firms is negatively correlated to underwriter reputation. The results of the OLS regressions on deterioration and other reputation proxies are shown in Columns (2) to (4), (6) to (8), (10) to (12), and (14) to (16) of Table 5, where the relative market share, the number of IPOs managed, and registered capital are added consecutively as Model (1) suggests. Not surprisingly, all coefficients for the new proxy remain statistically significant, while most of the traditional reputation proxies are no longer significant. These results confirm that the deterioration proxy constructed from IPO fraud can explain the long-term returns, even after controlling for the traditional reputation proxies. 4.2. Time-series estimation results We have examined the reputation of underwriters on the basis of time after the IPO firm has alleged fraud. But, this lacks considering the reputation of underwriters before their clients are reported IPO fraud. Since we classify an underwriter s reputation as damaged after its former 13

client is reported the IPO fraud, the reputation of this underwriter is intact before fraud disclosure. In this subsection, we conduct a time-series comparison to thoroughly study the reputation change of the same underwriters. The regression model is as follows: performance i = β 0 + β 1 post i + β 2 traditional proxy i + β 3 ROA i + β 4 fee i + β 5 age i + β 6 leverage i + β 7 sacle i + β 8 GEM i + β 9 market i + ε i (2) where post is the new measure for the reputation of the same underwriter, which takes on the value one after the underwriter s prior client has alleged IPO fraud, or zero before the disclosure of the IPO fraud. As discussed, once an IPO fraud event is disclosed, the reputation of the underwriter that handles this fraudulent IPO is damaged. We select this underwriter s past and future IPO clients and further divide them into two groups: IPOs that list within 12 months after the fraud disclosure (labeled as IPOs that list after fraud disclosure ) and IPOs that list within 12 months before the fraud disclosure (labeled as IPOs that list before fraud disclosure ). [Insert Table 6 Here] [Insert Table 7 Here] Table 6 and 7 analyze the initial returns and long-run market performance of these two groups. Our results show that the average initial returns and adjusted initial returns for IPOs that list after fraud disclosure are 102 percent and 100.5 percent, respectively, which are significantly higher than 54 percent for their corresponding group. In addition, compared with IPOs that list after fraud disclosure, IPOs that list before fraud disclosure can earn higher both buy-and-hold returns and market-adjusted buy-and-hold returns. These results further support the conclusions above and imply that the disclosure of IPO clients fraud damages underwriter reputation, which will affect the market performance of the underwriter s future IPO clients. In addition, when we remove the IPOs that list after December 2013, our results are still robust as Lines 5 and 6 of Table 6 suggest. 14

[Insert Table 8 Here] We next fix other control variables to examine how underwriter reputation impacts stock price reactions. Table 8 offers the results of the adjusted initial return and value-weighted buy-and-hold returns as dependent variables only. Although not reported, we conduct similar empirical tests with other post-ipo performance variables. All of the results are qualitatively similar to those presented in Tables 2 through 5 above. Specifically, the relationship between post-ipo price performance and the traditional reputation proxies remain insignificant, but importantly, the statistically significant coefficients of post imply that the worse the underwriter reputation is, the higher the initial return and the lower the long-term returns of its IPO clients. As the key gatekeeper in the IPO process, the underwriter fails in truth-telling. Compared with relative market share, the number of IPOs managed and the registered capital, the post proxy demonstrates the market s punishment to the underwriter, due to the underwriter s dereliction of duty. Overall, the empirical results show that it is effective to judge underwriter reputation by publicly disclosed formal announcements of IPO fraud. 5. DID analysis A potential concern about the new underwriter reputation proxy is the endogeneity problem. Once the IPO fraud is disclosed, the reputation of the underwriter that handles this fraudulent IPO is damaged. Expected changes in underwriter reputation may cause limited choices of its future IPO clients. Meanwhile, a low-quality firm would pick an underwriter with a bad reputation to gain the qualification of listing. To solve this endogeneity problem, in this section we perform a robust test adopting the DID approach. For each firm managed by a bad reputation underwriter, we select a matched firm managed by a good reputation underwriter. Specifically, once an IPO fraud event is made public, we consider the firms that hire this particular underwriter with alleged fraud as the treatment group. For each IPO in our treatment group, we select a matched IPO hiring underwriter without fraud as the control group using a propensity score matching algorithm. The propensity score matching method effectively reduces our sample selection bias and eliminates the endogeneity problem. 15

The specific matching steps are as follows: (1) Estimate a probit model based on a vector of firms and underwriter characteristics that may affect a firm s post-ipo performance. Firm characteristics include: ROA, fee, age, leverage, scale, and GEM. Meanwhile, underwriter characteristics include the relative market share, number of IPOs, and registered capital, as previously defined. All of the variables in the probit regression are used to ensure the satisfaction of the parallel trends assumption of the DID approach. (2) For each IPO in the treatment group, we first generate a list of other IPO firms hiring an underwriter without fraud whose listing dates are in the 12 months before and after the sample IPO listed as alternative samples. We then match each IPO in the treatment group to an IPO in the control group, with the nearest neighbor matching method. (3) We repeat the above two steps until matched samples are determined for all IPOs in the treatment group. [Insert Table 9 Here] Table 9 reports the univariate comparisons in the firm and underwriter characteristics between the treatment group and the control group, as well as their corresponding t-statistics. The average means of the treatment and the control group are close, which indicates no significant difference between these two groups. The propensity score matching process removes the meaningful observable differences between the treatment and control groups and the parallel trends assumption is not violated. [Insert Table 10 Here] Table 10 reports the univariate DID test results. Columns (2) through (5) present the average means before and after fraud disclosure for the treatment group and the control group, respectively. Columns (6) and (7) report the DID estimation with its corresponding t-statistics. As shown in panels A and B, the DID estimates of the initial return and adjusted initial return are all positive and statistically significant at the 1% level in both the total sample and the sub-sample. Meanwhile, the DID estimates are all negative and statistically significant at the 1% level for all of the buy-and-hold returns presented in Panels C, D, and E. These findings suggest that the increase in 16

initial returns and the decrease in long-term returns are larger for the treatment group than for the control group after the fraud disclosure, which further illustrates the deterioration of underwriter reputation after the IPO fraud event is disclosed. We then perform the DID test in a multivariate regression framework, which is constructed as follows: performance i = β 0 + β 1 post i + β 2 deterioration i + β 3 post i deterioration i + β 4 traditional proxy i + β 5 ROA i + β 6 fee i + β 7 age i + β 8 leverage i + β 9 sacle i + β 10 GEM i + β 11 market i + ε i (3) where the dependent variable performance i represents the initial returnand long-term returns of the listed firms, respectively. β 3 is the coefficient estimate of post i deterioration i, which captures the causal effect of firms that hire a damaged reputation underwriter. [Insert Table 11 Here] Table 11 presents the estimation results from Model (3). We offer the results of the adjusted initial return and five-year value-weighted buy-and-hold returns only, but all of the results are consistent. The coefficient estimate of post i deterioration i is the DID estimate that captures the causal effect of underwriter reputation on post-ipo performance. The coefficient estimates of the interaction terms are positive for the initial return and negative for the buy-and-hold returns at the 1% level, which is consistent with the univariate analysis result. The magnitudes of the DID coefficient estimates suggest that, compared to the post-ipo performance prior to the fraud disclosure, the treatment group exhibits a 23.86% larger increase in its adjusted initial return and a 98.28% decrease in its five-year value-weighted buy-and-hold return than the control group after fraud disclosure as Columns (1) and (5) show. These findings suggest that the underwriter reputation expectation generated by the IPO fraud disclosure appears to have a significant effect on the post-ipo performance. 6. Conclusions 17

In this paper, we examine the effect of underwriter reputation on post-ipo price performance with a totally new underwriter reputation measure. We conduct a thorough analysis of prior research on underwriter reputation, which utilizes the CM ranking and MW measure as typical proxies of underwriter reputation, that unfortunately cannot straightforwardly reflect the damage on underwriter reputation from exogenous events. The major contribution of this paper is to measure underwriter reputation from a groundbreaking perspective and directly judge the underwriter s reputation from the latest news of its past IPO client. We first manually collect all IPO fraud events in China s A-share market, and then we construct a new underwriter reputation using these events. We define a damaged underwriter reputation after its corporate client is caught IPO fraud. Consistent with Carter, Dark and Singh (1998), we demonstrate that firms going public and hiring a damaged reputation underwriter exhibit higher underpricing levels and lower long-term returns compared to firms that hire good reputation underwriters. In addition, we develop three traditional reputation proxies based on the relative market share, number of IPOs managed by each underwriter, and the registered capital of each underwriter. The addition of traditional reputation measures suggests that they have little influence on the level of IPO underpricing and long-run performance, while there is still a significant relationship between the new reputation proxy and post-ipo performance. Finally, to address the endogeneity problem between underwriter reputation and post-ipo performance, we employ a DID approach and find that IPO fraud disclosure has a negative effect on underwriter reputation. 18

References Agathee, U. S., R. V. Sannassee, and C. Brooks. 2012. The underpricing of IPOs on the Stock Exchange of Mauritius. Research in International Business and Finance 26: 281-303. Beatty, R. P., H. Bunsis, and J. R. M. Hand. 1998. The indirect economic penalties in SEC investigations of underwriters. Journal of Financial Economics 50: 151-186. Beatty, R. P., and J. R. Ritter. 1986. Investment banking, reputation, and the underpricing of initial public offerings. Journal of Financial Economics 15: 213-232. Booth, J. R., and R. L. Smith. 1986. Capital raising, underwriting and the certification hypothesis. Journal of Financial Economics 15: 261-281. Booth, L. C. 2004. Underwriter reputation and aftermarket performance of Closed-End funds. Journal of Financial Research 27: 539-557. Carter, R. B. 1992. Underwriter reputation and repetitive public offerings. Journal of Financial Research 15: 341-354. Carter, R. B., F. H. Dark, and A. K. Singh. 1998. Underwriter reputation, initial returns, and the Long-Run performance of IPO stocks. Journal of Finance 53: 285-311. Carter, R., and S. Manaster. 1990. Initial public offerings and underwriter reputation. Journal of Finance 45: 1045-1067. Chan, K., J. W. Cooney, J. Kim, and A. K. Singh. 2008. The IPO derby: Are there consistent losers and winners on this track? Financial Management 37: 45-79. Chemmanur, T. J., and P. Fulghieri. 1994. Investment bank reputation, information production, and financial intermediation. Journal of Finance 49: 57-79. Cooney, J. W., H. K. Kato, and J. S. Schallheim. 2003. Underwriter certification and Japanese seasoned equity issues. Review of Financial Studies 16: 949-982. Dong, M., J. Michel, and J. A. Pandes. 2011. Underwriter quality and Long-Run IPO performance. Financial Management 40: 219-251. Fama, E. F., and K. R. French. 1992. The Cross-Section of expected stock returns. Journal of Finance 47: 427-465. Fama, E. F., and K. R. French. 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33: 3-56. Fang, L. H. 2005. Investment bank reputation and the price and quality of underwriting services. 19

The Journal of Finance 60: 2729-2761. Hsu, H. C., A. V. Reed, and J. Rocholl. 2010. The new game in town: Competitive effects of IPOs. The Journal of Finance 65: 495-528. Loughran, T., and J. R. Ritter. 1995. The new issues puzzle. Journal of Finance 50: 23-51. Megginson, W. L., and K. A. Weiss. 1991. Venture capitalist certification in initial public offerings. Journal of Finance 46: 879-903. Michaely, R., and W. H. Shaw. 1994. The pricing of initial public offerings: Tests of Adverse-Selection and signaling theories. Review of Financial Studies 7: 279-319. Migliorati, K., and S. Vismara. 2014. Ranking underwriters of European IPOs. European Financial Management 20: 891-925. Nanda, V., and Y. Yun. 1997. Reputation and financial intermediation: An empirical investigation of the impact of IPO mispricing on underwriter market value. Journal of Financial Intermediation 6: 39-63. Ritter, J. R. 1991. The Long-Run performance of initial public offerings. Journal of Finance 46: 3-27. Rock, K. 1986. Why new issues are underpriced. Journal of Financial Economics 15: 187-212. Song, W. L., and H. Uzun. 2004. The impact of clients' alleged financial reporting fraud on underwriter reputation. SSRN Electronic Journal 438-457. Su, C., and K. Bangassa. 2011. Underpricing and long-run performance of Chinese IPOs: The role of underwriter reputation. Financial Markets and Portfolio Management 25: 53-74. Titman, S., and B. Trueman. 1986. Information quality and the valuation of new issues. Journal of Accounting and Economics 8: 159-172. Welch, I. 1989. Seasoned offerings, imitation costs, and the underpricing of initial public offerings. Journal of Finance 44: 421-449. Yung, C., and J. F. Zender. 2010. Moral hazard, asymmetric information and IPO lockups. Journal of Corporate Finance 16: 320-332. 20

Table 1 Descriptive statistics of the variables This table presents underwriter characteristics in Panel A, issuer and issue characteristics in Panel B, and market condition in Panel C. N, is the number of IPO firms that exist at any time during the sample period. We construct and define underwriter characteristic variables as follows: deterioration as the new measure of underwriter reputation, according to the reputation of the underwriter that IPO firms hire, which takes on the value one if the former IPO client of the underwriter is charged with IPO fraud, or zero if not. Other underwriter reputation proxies are traditional measures. Specifically, relative market share is the relative market share of the IPOs managed by each underwriter, which is measured as a percentage. The number of IPOs managed, is the number of IPOs managed over the period January 1,1991 to December 31, 2015 by each underwriter. Registered capital, is the registered capital of each underwriter. Issuer and issue characteristic variables are defined as follows: ROA is calculated as profits divided by assets at the end of each filing year, which is measured as a percentage. Fee, is the ratio of actual underwriter fees to the actual gross proceeds of the IPO, which is measured as a percentage. Age, is defined as the IPO year minus the year of the firm s founding. Leverage, is the ratio of total liabilities to total assets at the end of the IPO year. Scale (i.e., the size of a firm), is the logarithm of total assets at the beginning of the IPO year. GEM takes on the value one if the IPO went public on the GEM board and zero otherwise. The market condition variable represents the market sentiment, which is the number of IPOs issued in the corresponding firm s listing month. Variables N Mean Std Min Max Panel A: Underwriter Characteristics deterioration 912 0.500 0.500 0.000 1.000 reputation1 (%) 912 4.231 2.988 0.092 12.571 reputation2 912 58.191 45.733 1.000 162.583 reputation3 912 22.083 0.849 18.421 23.123 Panel B: Issuer and issue characteristics ROA (%) 912 6.630 2.790 1.306 16.274 fee (%) 912 6.244 2.808 1.736 16.439 age 912 7.203 4.582 0.822 20.060 leverage 912 0.264 0.156 0.025 0.728 scale 912 20.934 0.843 19.589 25.213 GEM 912 0.246 0.431 0.000 1.000 Panel C: Market condition Market 912 19.182 10.050 1.000 43.000 21

Table 2 Underwriter reputation and initial returns: a cross-sectional test This table reports IR (initial return) and adjir (adjusted IPO initial return) of IPO samples where an underwriter is hired with a damaged and an intact reputation. Each IPO that hire underwriter with a damaged reputation is matched with a single IPO that hire underwriter with an intact reputation, using a propensity score matching method. The sample periods include both January 1, 2000 to March 31, 2015 and January 1, 2000 to December 13, 2013. IR, is calculated as the percentage change in price from the offering price to the closing price of the stock on the first day of trading. adjir, is the initial return adjusted by the industry and market, which is calculated as adjir = (1 + IR) (1 + R M ), where R M is the market return of the same industry as the corresponding IPO firm from the prospectus date to the first day of listing; the Shenwan industry classification is used. Mean values and the return differences of IPOs that hire underwriters with different reputations are provided, along with the associated t-statistics. *, **, and *** denote the statistical significance at the 10, 5, and 1 percent levels, respectively. IPO sample period January 1,2000 to March 31, 2015 January 1, 2000 to December 13, 2013 IPOs that hire underwriters with a damaged reputation IPOs that hire underwriters with an intact reputation With- Without T- statistic IR 1.020 0.710 0.310 3.55*** adjir 1.005 0.717 0.289 3.39*** N 207 207 IR 1.059 0.727 0.332 3.61*** adjir 1.045 0.737 0.308 3.46*** N 195 195 22

Table 3 Regression results of underwriter reputation on the initial returns of IPOs: a cross-sectional test This table presents the results of OLS regressions explaining the initial returns of IPOs using the new underwriter reputation measure as well as the traditional reputation measures. The entire sample for the initial return from January 1, 2000 to March 31, 2015, consists of 414 IPO firms. We estimate with only the new underwriter reputation measure first performance i = β 0 + β 1 deterioration i + β 2 ROA i + β 3 fee i + β 4 age i + β 5 leverage i + β 6 sacle i + β 7 GEM i + β 8 market i + ε i and then with both the new and traditional underwriter reputation measures performance i = β 0 + β 1 deterioration i + β 2 traditional proxy i + β 3 ROA i + β 4 fee i + age i + β 6 leverage i + β 7 sacle i + β 8 GEM i + β 9 market i + ε i where IR and adjir are dependent variables, respectively. The independent variable, deterioration, takes on the value one if the IPO firms hire underwriters whose clients are disclosed IPO fraud, or takes on the value zero if the IPO firms list in the same period but the underwriter is hired without fraud disclosure. The traditional proxies for reputation are: reputation1, reputation2, reputation3 in three models, which are defined as the relative market share of the IPOs managed by each underwriter; the number of IPOs managed over the period January 1, 1991 to December 31, 2015 by each underwriter; and the registered capital of each underwriter, respectively. The first column for all dependent variables is the result of OLS regressions using only the new reputation measure, while the last three columns add the traditional reputation measures separately. All other control variables are as defined in Table 1. All the regressions control for industry fixed effects, which are based on the Shenwan industry classification. *, **, and *** denote the statistical significance at the 10, 5, and 1 percent levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) IPO initial return (IR) adjusted initial return (adjir) deterioration 0.2659 0.2651 0.2684 0.2659 0.2441 0.2435 0.2470 0.2442 (3.51)*** (3.51)*** (3.54)*** (3.51)*** (3.34)*** (3.34)*** (3.37)*** (3.33)*** reputation1 0.0267 0.0218 (1.86)* (1.57) reputation2-0.0009-0.0010 (-0.93) (-1.08) reputation3 0.0339 0.0210 (0.74) (0.48) ROA 0.0729 0.0749 0.0726 0.0723 0.0674 0.0690 0.0670 0.0670 (5.14)*** (5.28)*** (5.11)*** (5.09)*** (4.92)*** (5.03)*** (4.89)*** (4.88)*** fee -0.0226-0.0212-0.0220-0.0218-0.0260-0.0248-0.0253-0.0255 (-1.30) (-1.22) (-1.26) (-1.25) (-1.54) (-1.48) (-1.50) (-1.51) age -0.0328-0.0308-0.0320-0.0328-0.0324-0.0307-0.0314-0.0324 (-3.62)*** (-3.39)*** (-3.51)*** (-3.62)*** (-3.70)*** (-3.49)*** (-3.57)*** (-3.70)*** leverage 1.7976 1.8956 1.7586 1.7878 1.6746 1.7546 1.6310 1.6685 23