Pre-IPO Hype by Affiliated Analysts: Motives and Consequences

Transcription

1 Pre-IPO Hype by Affiliated Analysts: Motives and Consequences Yiming Qian University of Iowa Xinjian Shao University of International Business and Economics Jingchi Liao Shenzhen Stock Exchange April 2018 Abstract There have been debates on the pros and cons of the recent US regulatory changes about pre-ipo analyst coverage and involvement in the IPO process. We contribute to this debate by examining underwriter-affiliated analyst coverage prior to IPOs using data from China. We document that affiliated analysts make highly overoptimistic forecasts about IPO clients. More optimistic forecasts are associated with poorer post-ipo stock performance. Neither individual nor institutional investors are very sensitive to the analyst hype since the offer price is set sufficiently lower than short-term post-market prices. More aggressive hype helps an underwriter to gain or retain more investment banking business in the future.

2 1. Introduction Until recently, underwriter-affiliated analysts in the US had long been prohibited from issuing research reports about their IPO client firms prior to the initial public offerings (IPOs). In fact, the so-called quiet period rules prevent an issuing firm and its underwriters from making earnings forecasts or any forward-looking statements that are not already contained in the prospectus, between the period of IPO filing and shortly after the IPO. These rules have been relaxed for all firms since 2015 and lifted for some (small) firms since But the debate about the net effects of these rules is reignited, and there is little direct evidence for either side of the arguments. The rationale of restricting affiliated analysts pre-ipo coverage is to prevent hyping prior to the offering. There are indeed reasons to question affiliated analysts objectivity. Existing studies document that affiliated analysts almost always issue favorable reports and recommendations for their IPO clients immediately after the quiet period ends, and that their recommendations are less valuable than those of unaffiliated analysts (Bradley, Jordan, Ritter 2003; 2008; Michaely and Womack, 1999; George, Derrien and Womack, 2007). 2 Underwriters have incentives to provide overoptimistic coverage for the purpose of obtaining future investment banking business from issuing firms and/or increasing brokerage revenues (Jackson, 2005; Ljungqvist et al., 2009; Niehaus and Zhang, 2010). Such behavior will likely influence stock prices because investors might not fully recognize the extent of their bias. 3 But restricting hype for a period extending 1 In June 2002, the quiet period expiration day was extended from 25 to 40 calendar days after the IPO as part of the global settlement between the SEC and ten large investment banks. In September 2015, the quiet period was reduced to end 10 calendar days after the IPO. Moreover, Emerging Growth Company (EGC) IPOs are not subject to the quiet period rule after the JOBS ACT was signed into law in April Other studies document evidence that investment banks issue more positively biased research reports and investment recommendations around seasoned equity offerings or in general (Lin and McNichols (1998), Dechow, Hutton and Sloan (2000), O Brien, McNichols and Lin (2005), Barber, Lehavy and Trueman (2007) and Agrawal and Chen (2008)). 3 For example, Michaely and Womack (1999) document that stocks recommended by underwriter analysts perform more poorly than those recommended by unaffiliated analysts. Barber, Lehavy and Trueman (2007) compare recommendations of investment banks and independent research firms and also document that stocks 1

3 after the IPO at least gives the market a reasonable amount of time to establish a level of price without undue influences from the issuer and affiliated analysts. In recent years, the criticizing voice has become louder about the regulation. The main argument is that allowing the issuer and its underwriters to communicate more information can reduce information asymmetry about the IPO firm and thus lead to more efficient pricing of the stock. The argument also contends that the quiet period rules have become irrelevant in the internet age where information is abundant and information dissemination is fast. More information is always better because the market can distinguish truthful from false or misleading information. For example, Heyman (2013) states, Today, if Groupon or Facebook predicts, during the offering process, that it will be wildly profitable, a plethora of news outlets and analysts stands ready to pick the statement apart. (p189) In addition, the quiet period rules might put retail investors at more of an information disadvantage. Despite the recent relaxation and lifting of the regulation, we do not observe pre-ipo analyst coverage on the US market: affiliated analysts do not issue research reports until 25 days after the IPO and unaffiliated analysts hardly initiate coverage before affiliated analysts do. 4 Thus we can t empirically examine the effects of pre-ipo analyst coverage in the US. Indirect evidence, however, raises concerns about the relaxation of these rules. Dambra et al. (2018) show that following the JOBS Act that allow certain analysts to be more involved in the IPO process (e.g., by attending pitch meetings and interacting with potential investors), these analysts provide more optimistically biased and less accurate coverage after IPO, and that investment banks benefit from this increased bias. In this study, we directly examine underwriter-affiliated analysts pre-ipo coverage and its effects on stock prices and investor behavior, using a sample of Chinese IPOs. In China, affiliated analysts are allowed to issue research reports before IPOs and they always do so. A recommended by the former group underperform. On the other hand, Agrawal and Chen (2008) show evidence that the market recognizes analysts conflicts and properly discount their opinions. 4 Jia, et al. (2018) conjecture that the industry practice is path dependent and that changes will happen slowly. 2

4 firm typically also receives pre-ipo coverage from multiple unaffiliated analysts. We obtain from Shenzhen Stock Exchange (SZSE) the Research Report on Investment Value issued by lead-underwriter-affiliated analysts for 773 IPO stocks that went public during , and hand collect from these reports the analyst forecasts of the firm s future operating performance. Using this unique dataset, we study the nature and effects of pre-ipo coverage by affiliated analysts. Specifically, we examine the following research questions: (1) are affiliated analysts pre-ipo coverage optimistically biased? Are they more optimistically biased than forecasts for similar public firms? Are they more optimistically biased than unaffiliated analysts reports? (2) How do affiliated analysts forecast bias influence the stock price? (3) If the pre-ipo coverage is biased, are investors able to recognize and adjust to these biases? (4) Do underwriters benefit from hyping the IPO stock? We first document that affiliated analysts forecasts of an IPO firm s future performance are highly overoptimistic, i.e., they are positively biased compared to actual performance. The forecasted revenues and income are larger than the actual performance 70-90% of the time. The magnitude of the bias increases with more distant forecasts, and is larger for income than for revenues. For example, the average bias (measured by the percentage difference between forecasted and actual performance) for revenues is 6.5%, 14.8%, and 20.4% for the IPO year, the year after and the 2 nd year after, respectively; and the average bias for net income is 13.2%, 33.0% and 48.1% for the three years respectively. We recognize that there is analyst optimism in earnings forecasts in general (e.g., Fried and Givoly, 1982; O Brien, 1988; Butler and Lang, 1991). To ensure what we find in the pre-ipo context is above and beyond the general bias, we compare analysts pre-ipo forecasts with those for similar public firms. For each IPO firm, we find a set of matching firms in the same size and market-to-book deciles. We then use the median forecast bias of the matching portfolio as the benchmark. The abnormal bias for each accounting measure has a significantly positive mean and median, and the difference is economically large. We further compare the pre-ipo forecasts 3

5 made by affiliated vs. unaffiliated analysts for a same IPO firm. Both univariate comparison and multiple regressions show that affiliated analysts are more positively biased than unaffiliated analysts. We then test whether the biased forecasts influence the price of the IPO stock. We document strong and robust evidence that more optimistic forecasts are associated with more negative post- IPO long-run returns, relative both to the first trading day closing price and to the IPO offer price. This suggests that analyst hype with biased forecasts increase both the offer price and the postmarket prices immediately after the IPO. Investigation into investor behaviors reveals that investors inadequately adjust to analyst hype. Neither retail nor institutional investors demand of the stock decreases with the extent of the forecast bias. Institutional investors slightly but inadequately adjust their bid prices downward when facing more biased forecasts. They assign top ratings to the underwriter s research report most of the time, and hardly change to lower ratings in response to analyst hype. Investors are insensitive to analyst hype because the IPO offer price is set sufficiently lower than short-term post-market prices: the IPO initial return does not vary with analyst hype. This is consistent with the sentiment theory of Ljungqvist, Nanda and Singh (2004) and Derrien (2005) who argue that underwriters will increase the IPO offer price when investor sentiment is high, but still below what sentiment investors are willing to pay (i.e., the immediate post-market price). Finally, we find evidence that underwriters have monetary incentives to hype IPO stocks. Underwriters will receive higher RMB fees from the current IPO offering since aggressive hype will lead to higher offer price. Underwriters will also receive more investment banking business from the current client and more IPO business from other firms in the future. Our paper is the first to examine pre-ipo coverage by affiliated analysts. We therefore contribute to the ongoing debate about the relative importance of additional information provision and the governance of information quality around security issuance. Similar as in the US, underwriters in China have strong monetary incentives to hype IPO stocks to generate 4

6 demand, maximize fees, and increase future investment banking business. Our study shows that when unregulated, underwriters do try to hype the IPO stock, and that the market does not fully recognize and adjust to such hype in pricing the stock. After all, forward-looking statements are not so easy to be picked apart even in the internet age since they are not verifiable ex ante, and issuers and their underwriters are deemed to have private information that other market participants do not have access to. Although IPO investors are guaranteed with good returns due to the IPO underpricing relative to short-term post-market prices, analyst hype results in less efficient stock prices and leave investors on the open market bear the loss. Our evidence points to the merits of regulatory restrictions on pre-ipo communications and research coverage by affiliated analysts. In a study complementary to us, Jia, Ritter, Xie and Zhang (2018) analyze unaffiliated analysts pre-ipo coverage in China. They document that more analyst coverage is associated with better post-ipo long-run performance, suggesting that unaffiliated analysts research prior to IPO provides valuable information. This poses interesting contrast to our results but is not inconsistent with our message. Together the two papers show that affiliated analysts, due to its partial position, try to hype the stock, whereas unaffiliated analysts generate more useful information. Our paper adds to a growing literature on IPOs in emerging markets. Doidge et al. (2013) show that an increasing share of IPOs worldwide are occurring in emerging markets, yet evidence on these markets is limited. The different institutional features of these markets can provide interesting counterfactuals to the US and offer opportunities to test theories that otherwise can t be tested. Chang, Chiang, Qian, and Ritter (2017) examine the pre-ipo market in Taiwan and IPO pricing given the pre-market. Chiang, Qian, and Sherman (2010) and Chiang, Hirshleifer, Qian, and Sherman (2011) examine investor behavior in IPO auctions in Taiwan where unlike in the US, different IPO methods are commonly used. Similar as these papers, we 5

7 use unique data from an emerging market to shed light on fundamental questions relevant to markets worldwide. 2. Institutional Background and Data 2.1 Institutional Background Firms can be listed and traded on one of the two stock exchanges in China: Shanghai Stock Exchange (SSE) and Shenzhen Stock Exchange (SZSE). By the end of year 2012, there were 954 listed firms on SSE with a total market cap of RMB15.9 trillion (i.e., $2.5 trillion); and there were 1,540 listed firms on SZSE with a total market cap of RMB7.2 trillion (i.e., $1.1 trillion). 5 Today SSE is the 4 th -largest stock exchange in the world and SZSE the 8 th by market cap. Firms listed on SSE tend to be large state-owned enterprises whereas more private firms are listed on SZSE. There are three sections of listings on Shenzhen Stock Exchange (SZSE): the main board, the Small and Medium Enterprise (SME) board, and ChiNext board (also known as Growth Enterprise Market, or GEM) with the last board featuring high-growth high-tech but likely smaller and younger companies. 6 There are many more IPOs on SZSE than on SSE. During our sample period of July 2009-November 2012, there were 783 IPOs on SZSE, and only 102 IPOs on SSE. Since our key data affiliated analysts research reports are obtained from Shenzhen Stock Exchange (SZSE), our sample focuses on IPO firms listed on SZSE IPO application and approval. Firms need to obtain approval from China Securities Regulatory Commission (CSRC), the counterparty of SEC in the US, to go IPO. To apply for an IPO, a firm must satisfy some minimum requirements on size and profitability. For one thing, it must have positive net income prior to the filing. 7 Satisfying these conditions does not guarantee 5 The exchange rate by 2012 year end was $1= RMB In 2000, the regulators decided that SSE will mainly host large bluechip stocks and SZSE will cater to medium and small size firms. Since then, there have been no IPOs on SZSE Main Board. 7 SSE has generally stricter listing requirements than SZSE SME board, which in turn has stricter requirements than GEM boards. For example, SZSE SME board requires firms to have positive income in each of the last three years (as opposed to just the last year) prior to IPO filing. 6

8 the approval of an IPO. CSRC has the discretion of deciding on individual cases or imposing stricter conditions across board during certain periods. Appendix A shows the key steps for IPO application and approval Hybrid IPO auctions. Once an IPO approval is officially granted, the offering process starts. IPO shares are sold via a hybrid auction method consisting of an offline auction tranche and an online fixed-price tranche. 8 For most offerings (over 90%) in our sample, about 20% of the IPO shares are sold in offline tranche and 80% in the online tranche. Appendix A illustrates the timeline of the IPO process. In the industry, the day on which investors submit order deposits is customarily called Day T. The offering starts on Day T-6, when the issuer publishes IPO announcement and IPO Bookbuilding [Auction] and Road Show Announcement. ( Bookbuilding is a misnomer here: Chinese IPOs do not use the bookbuilding method because underwriters have no allocation discretion.) The offline auction tranche and the road show start on Day T-5 and last for three business days. The auction tranche caters to institutional investors, in which investors submit bids, i.e., combinations of price and quantity. Each investor can bid no more than the total number of shares sold in the tranche. The auction method used is a dirty Dutch auction, i.e., the offer price can be set lower than the market clearing price at which the demand equals to the supply. Bids above the offer price are called valid bids. Among valid bids, allocation is made proportionally during July 2007-October 2010 and determined based on a lottery during November November Once the offer price for the IPO is determined based on the auction, investors can submit orders to the online tranche. Both individual and institutional investors can participate in the tranche (but an investor cannot participate in both tranches). Each investor can bid no more than 8 We focus on the IPO process on firms listed on Shenzhen Stock Exchange. IPOs on Shanghai Stock Exchange also use a hybrid auction method, but with differences on specific mechanisms and processes. Perhaps the biggest difference is that for the auction tranche, SSE uses a two-stage auction method, with the first auction producing a price range and the second auction producing the final offer price. 7

9 0.1% of the shares sold in the online tranche. Allocation is determined using a lottery. The IPO stock starts public trading typically 7 to 14 days after the IPO allocation Lock-up period. During our sample period, online IPO investors are not subject to any lock-up period and can sell their shares immediately after the stock starts public trading. Offline investors, on the other hand, are subject to a 3-month lock-up period before November, 2011, and no lock-up period after that Pre-IPO analyst coverage. When the offering process starts (on Day T-6), the underwriter will submit a Research Report on Investment Value for the IPO stock written by its analyst, to the electronic EIPO System operating by Shenzhen Stock Exchange. Institutional investors qualified to participate in IPO auctions can access the report from the EIPO System. 9 The report discusses the firm s competitive position in the industry, growth and profitability prospects, and its risk factors. It makes projections of the firm s pro forma financial statements, and provides valuation analysis. In particular, the report provides the forecasts for the next three year s sales and income, as well as a suggested price range. The suggested price range is not available in the prospectus. For IPOs after November 2011, the suggested price range as well as the detailed auction bids are publicly disclosed on day T+2, in the document The Announcement of Offline Allocation Results. Bidding investors are required to rate the research report before they can submit bids. 10 They are asked to give three ratings on the objectivity of the report, the rationality of the valuation, and the overall quality of the research, respectively. The ratings range between 1 (the worst) and 5 (the best). Some unaffiliated analysts also issue pre-ipo research reports. In our sample, unaffiliated analysts reports are all issued after the affiliated analyst. 9 Qualified investors must meet certain criteria set by China Securities Regulatory Commission, including varying criteria on size, investment type and activeness for different types of investors. Qualified investors must have registered with Securities Association of China. 10 All qualified Investors have access to the report but those who do not bid cannot rate. 8

10 2.2 Sample and Data Our sample includes 773 IPOs listed on Shenzhen Stock Exchange during the period of July 2009 and November 2012, for which the key data the underwriter s affiliated analysts Research Report on Investment Value is available from the stock exchange (out of 783 IPOs on SZSE during the period). We hand collect from each report the analyst forecasts for the next three year s sales, operating income and net income, as well as the suggested price range for the IPO stock. 11 SZSE also provides the institutional investors ratings of these reports. For unaffiliated analysts forecasts, we use a union set of such forecasts from both CSMAR and WIND. We also obtain bids of institutional investors in the auction tranche and their IPO allocations from the stock exchange. Most of the IPO background information is from WIND Database including firm characteristics such as firm age and financials, and offering information such as underwriter name, offer price, shares sold, fees and online subscription ratio. We collect additional information from IPO prospectuses such as underwriting fees and marketing fees for the offering. Finally, we retrieve post-ipo financial data and stock price and returns from CSMAR database. Table 1 reports the distribution of our sample by year and by industry. Table 2 reports the descriptive statistics of the IPO firms in our sample. The median firm is 10 years old, has an asset value of RMB 427 million, and is profitable with a ROA of 14%. Most of these firms are private firms (only 8% are SOEs), which is very different from firms listed on Shanghai Stock Exchange where 49% of the IPO firms during the same period are SOEs. Forty-seven IPOs are listed on the GEM Board and the rest are on the SME Board. During our sample period, the average initial (first-day) return of the IPO is 36.6% and 85% of the initial returns are positive. In contrast, the average market return over the three months prior to the IPO is virtually zero. The price revision defined as the percentage change 11 Some reports provide forecasts for a subset of these variables and/or for one or two years in the future. 9

11 from the mid point of the suggested price range to the offer price, has a small average (0.7%), but exhibits substantial variation (the standard deviation is 20.6%). The IPOs are usually highly oversubscribed: the average offline subscription ratio is 105 times the supply and the offline subscription ratio is even higher at 161. We measure the media attention a firm received by searching Baidu News for the number of news articles with the firm name in the title during the year prior to its IPO. On average a firm appears in news headlines 14 times in the year prior to the IPO, although this variable is highly skewed with the median count being only one. Each IPO has one underwriter with a single exception. The average fee paid to the underwriter is 5.7% of the proceeds. There are 68 unique underwriters (securities firms) in our sample. Each year, Securities Association of China (SAC) evaluates each securities firm for its performance on regulatory compliance, risk management, and its competitiveness in the industry and assigns a rating ranging from 1 (the worst) to 11 (the best). We classify a security firm as having high reputation if it receives a rating of 10 or 11. Using this definition, 52.7% of IPOs have underwriters with high reputation. We also look at whether the underwriter-affiliated analyst is a start analyst in the IPO year based on the financial magazine New Fortune. In 28% of the cases, the research report was provided by a star analyst. We calculate the quantity-weighted average bids of institutional investors in the auction tranche. The average bid-to-suggested-price ratio (i.e., weighted average bid relative to the mid point of the price range) is 94%. For post-ipo stock performance, we calculate the buy-and-hold abnormal returns (BHARs) using the weighted average market index return as the benchmark. The average buy-and-hold returns in the one, two, and three years post IPO are -10.2%, -0.8%, and 41.2% (the median returns are -11.3%, -6.8% and 1.4%), all significant at the 1% level with one exception: the mean value of BHAR2Y is not statistically significant. 10

12 3. Affiliated analysts forecast bias This paper examines whether affiliated analysts try to hype IPOs. The first thing we investigate is whether their pre-ipo forecasts of the firm s operating performance is positively biased. We measure the forecast bias as the negative of (the actual performance minus the actual performance, divided by the forecasted performance). The higher the measure, the more positively biased the forecast is. There are three operating performance measures: sales, operating income and net income; and we look at forecasts for the IPO year, the year after, and the 2 nd year after IPO. Results based on operating income are very similar to those on net income. For brevity, we present results using sales and net income only. Table 3 Panel A reports the summary statistics of these measures of forecast bias. Forecasts on both sales and net income are highly positively biased: all measures have significantly positive means and medians, and the percentage of positive values ranges between 66%-88%. Perhaps not surprisingly, the forecasts for the more distant future are biased. For example, the mean forecast bias for sales is 6.5%, 14.8%, and 20.4% for years 0, 1 and 2, respectively, with year 0 being the IPO year. The frequencies of positive forecast bias for the three measures are 66%, 74% and 79%. We also observe that forecasts on net income tend to be more biased than those on sales. The mean forecast bias for net income is 13.2%, 33.0%, and 48.1% respectively for years 0, 1 and 2 and the frequencies of positive values of these three measures are 73%, 85% and 88%. Figure 1 depicts the distributions of the six measures of forecast bias. With about 70-90% of forecasts higher than the actual performance across measures, it is hard to argue that these forecasts are unbiased. We recognize that sell-side analysts exhibit optimistic bias in general (e.g., Fried and Givoly, 1982; O Brien, 1988; Butler and Lang, 1991). However, the magnitude of the bias we document suggests that they are even more biased in the pre-ipo setting. We examine whether that is the case by comparing the forecast bias of affiliated analysts for the IPO stock and the bias exhibited in forecasts for similar publicly-traded stocks. For each IPO stock, we identify a matching 11

13 portfolio as follows. First, we include as matching candidates firms that have been publicly traded for at least three years by the time the affiliated analyst s report is published, which we use as the matching day. Second, we identify matching firms that are in the same size and market-to-book (M/B) deciles as the IPO firm. We measure the IPO firm s size as the market cap based on the mid point of the suggested price range, i.e., shares outstanding times the mid point of the suggested price range, and its market-to-book ratio is the market cap divided by the book value of equity prior to IPO. We measure matching candidates size as the market cap on the matching day. Third, we require there are at least 3 firms in the matching portfolio. We then collect analyst forecasts for these matching firms within 120 days around the matching day and calculate their forecast bias as similarly defined for the IPO stock. If an analyst makes multiple forecasts about the same firm s same performance measure during the period, we use her last forecast. Table 3 Panel B reports the affiliated analysts pre-ipo abnormal forecast bias using the analyst forecasts for matching firms as the benchmark. For each matching firm, we obtain the median analyst forecast bias. We then take the median across matching firms (within the matching portfolio) as the benchmark. The abnormal forecast bias is a raw measure of the bias (e.g., error_sales0) minus the benchmark. Panel B shows that all abnormal bias have positive means and medians (all statistically significant at the 1% level), and their magnitude remains high. For example, after accounting for general analyst optimism, the abnormal forecast bias for net income is 6.4%, 23.0% and 30.1% for the IPO year, the year after and the second year after, respectively. The results thus show that affiliated analysts optimistic bias is over and beyond the general analyst optimism, possibly due to the acute monetary incentives related to the underwriting business and/or to the high information asymmetry surrounding the IPO stock. We also examine whether affiliated analysts forecasts are more optimistic than unaffiliated analysts for the same IPO stock. Interestingly, for each IPO in our sample, unaffiliated analysts reports are published 12

14 after the affiliated analyst. It is plausible that unaffiliated analysts forecasts are therefore influenced by the affiliated analyst, either for benign reasons such as updating beliefs given the information, or for more self-interested reasons such as currying favor (so that their clients might obtain IPO allocation) or reciprocating favor (so that the affiliated analyst will do the same for them when their firms are acting as underwriters). Nonetheless, they might still have less direct monetary incentives to hype the stock. We examine this in Table 4. Panel A reports the affiliated analysts abnormal forecast bias using the median value of unaffiliated analysts bias as the benchmark (using the mean value does not change the results qualitatively). These alternative measures of forecast bias again all have positive means and medians, suggesting that affiliated analysts are more optimistic than unaffiliated analysts. Similar as before, the abnormal bias increases with the forecast horizon. The frequency of positive abnormal bias is also very high at around 60-80%. One notable difference compared to Table 3 is that the abnormal bias benchmarked against unaffiliated analysts for the same IPO stock is much smaller than that benchmarked against the analyst forecasts for non-ipo stocks. For example, the forecast bias for net income is 5.0%, 6.0% and 7.3% for years 0, 1, and 2 (versus 6.4%, 23.0% and 30.1% for the corresponding variables in Table 3 Panel B). This suggests that both affiliated and unaffiliated analysts are more positively biased in their pre-ipo forecasts than for non-ipo stocks, with affiliated analysts over optimism being the highest. Table 4 Panel B examines whether the affiliated status affects the forecast bias by estimating a multiple regression. Specifically, we use observations at the IPO-analyst level and regress a measure of forecast bias on affiliated dummy which equals to one if the analyst is an affiliated analyst, controlling for firm and IPO characteristics. In addition to year and industry fixed effects, we also include investment bank (with which the analyst is affiliated with) fixed 13

15 effects. 12 The standard errors are clustered by IPO. Consistent with Table 4 Panel A, we find a significantly positive coefficient on affiliated dummy for all six measures of forecast bias, suggesting that affiliated analysts are more positively biased. Moreover, the magnitude of each coefficient is similar to the corresponding abnormal forecast bias in Panel A. In summary, Tables 3 and 4 show that affiliated analysts pre-ipo forecasts are highly overoptimistic. The positive bias is way beyond the general optimism for similar non-ipo stocks, and is also higher than the bias of unaffiliated analysts for the same IPO stock. The evidence suggests that underwriter-affiliated analysts try to hype the IPO stock. 4. The impact of analyst hype 4.1 The impact of analyst hype on prices In this section, we examine the impact of affiliated analyst s hype on the prices of the IPO stock on both the post-market price and the offer price. Our approach is to investigate the postmarket long-run stock returns. Under the premise that stock prices approach to the fair value in the long term, post-market stock returns measure how efficiently the offer price and the first-day market price are set. If analyst hype leads to overvaluation of the stock in and immediately after the IPO, then the stock should suffer poor performance in the long run. In other words, we expect to see long-run stock returns are negatively related with pre-ipo analyst hype. Alternatively, if investors fully recognize the hype and therefore the prices are not affected, then there should be no relationship between hype and long-run returns. We calculate IPO stocks buy-and-hold abnormal returns (BHARs) during the three years after the IPO, using the value-weighted market return as the benchmark. We compute two return variables: the 3-year BHAR relative to the closing price on the first trading day, BHAR3Y, and the BHAR relative to the IPO offer price, BHAR3Y_offerPrc. They measure the extent of 12 In an alternative specification, we include IPO fixed effects instead of year and industry dummies in the regressions, results are robust. 14

16 overvaluation of the first-day trading price and the offer price, relatively. Results using 1-year or 2-year BHARs are robust (not reported). Table 5 reports the results of regressions of post-market BHARs on the affiliated analyst s pre-ipo forecast bias, controlling for firm and IPO characteristics including year and industry fixed effects. The dependent variable is BHAR3Y in Panel A and BHAR3Y_offerPrc in Panel B. In both panels, we observe a significantly negative coefficient on each of the six forecast bias measures, suggesting that analyst s hype in their pre-ipo forecasts push up the stock prices both in the IPO (the offer price) and immediately after the IPO (the first-trading day price), leading to adjustments in the long run. The economic significance of the impact is big. For example, according to the estimate in Panel A, a one-standard-deviation increase in the forecast bias in sales for years 0, 1, and 2 lead to a decrease in the 3-year abnormal return by 13.0%, 30.2%, and 36.8%, respectively. This is substantial compared with the median value of the 3-year BHAR of 1.4%. The forecast bias in net income has impact of similar magnitude. The results thus suggest that the more aggressive the affiliated analyst hype the IPO stock, the more inflated the IPO offer price and the first-trading day price are, hence the lower the longrun stock returns. 4.2 The impact of analyst hype on IPO investor behavior The previous subsection documents that analyst hype pushes up the IPO offer price, which suggests that IPO investors do not fully adjust to the hype. We now investigate in depth how they react to analyst hype. We do so by examining IPO investor demand, institutional investors bids in the auction tranche and the resulting price revision, and institutional investors rating of the analyst research report. Table 6 reports the results of regressions of investor demand on affiliated analyst forecast bias, controlling for firm and IPO characteristics including year and industry fixed effects. In 15

17 Panel A, the dependent variable is retail investors demand by the subscription ratio in the online tranche. In Panel B, the dependent variable is institutional investors demand by the subscription ratio in the auction tranche. Table 6 Panel A show that the coefficients on the three measures of forecast bias in sales are statistically insignificant; and that the coefficients on the three measures of forecast bias in net income, however, are significantly positive. A one-standard-deviation increase in net-income forecast bias for years 0, 1, and 2 lead to an increase in online subscription ratio by 10.3, 7.9, and 9.2 times. In comparison, the median subscription ratio for the sample is 139. This table shows that retail demand does not decrease with hype, and actually increases with hype to earnings. In Panel B, all the coefficients on the six measures of forecast bias are statistically insignificant, suggesting that institutional demand does not change with hype. In summary, Table 6 indicates that neither retail investors nor institutional investors reduce demand when hype is high; in fact, retail investors demand more when forecasts about net income are more positively biased. There are some interesting contrasts between retail and institutional demands. Retail demand is positively related to the recent market return, whereas institutional demand is not. This is consistent with the finding of Chiang, Qian and Sherman (2010) who examine Taiwanese IPO auctions and document that individual investor are return-chasers and institutional investors are not. Another difference is retail demand does not depend on the firm s SOE status, whereas institutional demand increases if the firm is a SOE. Next, we examine how institutional investors adjust their bid prices in response to the affiliated analysts forecast bias. If they realize that the underwriter tries to hype the stock, then they might adjust downward their bids relative to the suggested price range. Table 7 reports the results of regressions of bid adjustment on analyst forecast bias, controlling for firm and IPO characteristics including year and industry fixed effects. In Panel A, we use IPO level observations and the dependent variable is the quantity-weighted average bid price divided by 16

18 the mid point of the price range. In Panel B, we use IPO-investor level observations and the dependent variable is each investor s quantity-weighted average bid price divided by the mid point of the price range. For these regressions, we include in addition investor fixed effects, and cluster standard errors by IPO. Table 7 Panel A shows that institutional investors bid prices are negatively related to the forecast bias: the coefficients on all six measures of forecast bias are significantly negative. The economic magnitude of the effect, however, is small. A one-standard-deviation increase in sales forecast bias for years 0, 1, and 2 lead to a decrease in standardized average bid price, avgbid_midprc, by 3.1, 1.9, and 1.8 percentage points, compared with a median value of 94.3 percentage points. In other words, although a one-standard-deviation increase in sales forecast bias is associated with 20-50% return decrease in long-run returns relative to the offer price (BHAR3Y_offerPrc), institutional investors only adjust downward their bid prices by 2-3%. Panel B, using investor level data, reports similar results about the effect of forecast bias on investors bid prices. That is, institutional investors adjust downward their bid prices slightly when analyst hype is aggressive, but not adequately. If investor bid prices inadequately adjust to analyst hype, we expect the final offer price will exhibit similar pattern, i.e., the offer price will be adjusted downward, but not enough to nullify hype s impact on prices. Table 8 Panel A estimates regressions of the price revision on forecast bias. Price revision is defined as the offer price relative to the mid point of the price range, minus one. The variable is negatively related to four out of six forecast bias measures. Where the coefficient is statistically significant, a one-standard-deviation increase in forecast bias leads to a downward price revision by percentage points. Consistent with the results in Table 7, this suggests that the offer price is adjusted downward when investors adjust their bid prices downward in response to aggressive hype. However, this adjustment is hardly enough given the previous result that a one-standard-deviation increase in forecast bias decreases long-run returns (relative to offer price) by about 20-50%. 17

19 Table 8 Panel B reports the results of regressions of IPO initial return on analyst forecast bias. We find statistically insignificant coefficient for each measure of forecast bias. This is consistent with the notion that both the offer price and the first trading-day price are similarly inflated by analyst hype, hence rendering no significant changes to the first-day return. In the initial return regressions, we include price revision as an additional control variable. It is well documented that for US bookbuilding IPOs, there is a positive relationship between price revision and the initial return, which is known as the partial adjustment phenomenon (Hanley, 1996). In our sample, the univariate relationship between the two variables is also highly positive. However, price revision becomes insignificant in the multiple regression, in particular after including the year fixed effects. The coefficients on analyst forecast bias remains insignificant with and without including price revision. Lastly, we conduct a direct test on how sensitive institutional investors are to analyst hype and/or how much they care about the issue. Specifically, we examine whether and to what extent their ratings of the affiliated analyst research report vary with analysts forecast bias. For brevity, we focus on the ratings of the overall quality of the research report. Results using the other two ratings (of the objectivity of the report, and of the rationality of the valuation) are qualitatively similar. Among the 54,271 IPO-investor who submit ratings, 64.1% of them give the top rating (5), 29.4% give the rating of 4, 6.1% at 3, and 0.4% at 1 and 2. In other words, about two-thirds of the ratings are at 5, which gives a quick indication that ratings are not very discriminative. Table 9 reports the results of regressions of investor ratings on forecast bias, controlling for other firm and IPO characteristics including year and industry fixed effects. We also add investor fixed effects and cluster standard errors by IPO. The dependent variable is a dummy equal to one if the investor gives a rating of 5 and zero otherwise. We observe that in two of the six regressions, the coefficient on forecast bias is significantly negative. The economic significance of these two coefficients is small: a one-standard-deviation 18

20 increase in the forecast bias leads to a decrease in the likelihood of a top rating by percentage points. Thus we find some weak evidence that investor give lower ratings to the research report when analyst hype is aggressive. We also consider the average investor rating for an affiliated analyst s report (i.e., the average rating in an IPO). We find the variation of the average rating is quite small. Across the 773 IPOs, the average rating has a 10 percentile of 4.40, a 90 percentile of 4.65, and a standard deviation of This again suggests that investors do not give very discriminating ratings to reports with different qualities. When we use IPO level observations and regress the average rating on analyst forecast bias, we find similar results (not reported). That is, there is weak evidence that the average rating decreases with forecast bias with small economic significance: a one-standard-deviation of forecast bias decreases the rating only by less than In summary, we find evidence that retail demands are positively influenced by analyst hype. Institutional investors demands are not significantly changed by the hype; they adjust downward their bid prices to the hype but the adjustment is incomplete. One reason that investors are insensitive to the analyst hype is that the IPO offer price is sufficiently underpriced relative to the post-market prices. As shown in Table 2, the average (median) initial return is 36.6% (27.0%) with only 15% of the IPOs with nonpositive initial returns. During our sample, online IPO investors are not subject to any lockup period. These investors can earn a high return within a couple of weeks (from IPO pricing to listing). Offline (institutional) investors are subject to a 3-month lock-up period before November 2010, and no lock-up period after that. If we calculate the return for institutional investors assuming they sell immediately after the lock-up period, the average (median) return is 26.8% (15.8%) and they make positive profits 71% of the time. This is consistent with Ljungqvist, Nanda and Singh (2004) and Derrien (2005) who argue that when there is positive investor sentiment, the IPO offer price will be set higher, but still below the short-term aftermarket price. Analyst hype can increase investor sentiment, which 19

21 increases both the offer price and the short-term aftermarket prices. Due to sufficient underpricing relative to the short-term aftermarket prices, IPO investors will still receive lucrative returns, which make them insensitive to the hype. The consequence, however, is less efficient IPO prices. 5. Underwriters motives to hype The analyses in previous sections shows that by hyping an IPO stock, the underwriter is able to increase both the offer price and short-term aftermarket prices, which benefits the issuer but do not hurt IPO investors who only hold the stocks for the short run. This suggests that underwriters can gain monetary benefits from hyping the stock. First, underwriters receive higher RMB fees from the issuer given that fees are contracted as a percentage of the IPO proceeds. Table 5 Panel B suggests that a one-standard-deviation increases in forecast bias can increase the offer price by 20-50%, which can in turn increases the underwriting fee by 20-50%. Second, we hypothesize that an underwriter is more likely to retain the IPO firm as a future investment banking client if it hypes the IPO stock more aggressively. To test this hypothesis, we examine the likelihood that the investment bank will continue to serve as the IPO firm s underwriter in its post-ipo financing activities. We obtain from WIND Database the sample firms financing activities over the three years following their IPOs, including SEOs (seasoned equity offerings), PIPEs (private investments in public equities), stock dividends and bond issuances. We find 216 firms had 294 financing events. We define a dummy undrwrt retention equal to one if the IPO underwriter acts as an underwriter in any of the firm s post-ipo financing activities. The average value of undrwrt retention is 52%, meaning that about half of the 216 firms use the same underwriter in at least one of its post-ipo financing events. Table 10 reports the results of regressions of undrwrt retention on the affiliated analyst s pre- IPO forecast bias, controlling for firm and IPO characteristics including year and industry fixed effects. In four out of the six regressions, the coefficient on forecast bias is significantly positive. 20

22 For these significant estimates, a one-standard-deviation increase in forecast bias is associated an increase of retention probability by 6-8 percentage points, which is substantial compared to the average probability of 52%. Third, we hypothesize that by pleasing the current IPO client with more aggressive hype, an underwriter is able to attract more IPO underwriting business in the future. Table 11 estimates regressions of an underwriter s future IPO market share on its affiliated analyst s forecast bias in the current IPO. We measure an underwriter s market share of the total proceeds of the IPOs in the year after the current IPO. The average market share is 3.8%. We control underwriter reputation (high undrwrt repu) and the current IPO s stock performance in the subsequent year (BHAR1Y). Table 11 shows that the coefficients on all but one measures of forecast bias are significantly positive. Consistent with our hypothesis, this suggests that more aggressive hype wins the underwriter more IPO market share in the future. The economic magnitude of the impact is relatively small but still reasonable: a one-standard-deviation increase in forecast bias is associated with an increase of the market share by percentage points, which is an increase of 4-5% of the average market share. Not surprisingly, high undrwrt repu is the most important determinant of underwriters market shares: being a high-reputation underwriter increases the market share by about 2 percentage points. Interestingly, an IPO s post-market stock performance has no impact on an underwriter s market share. Together the evidence suggests that underwriters receive monetary rewards for hyping IPO stocks. The rewards include increased underwriting fees from the current IPO, as well as more future investment banking businesses from both the current client firm and other IPO firms. 6. Conclusions In this paper, we empirically examine underwriter-affiliated analysts pre-ipo research coverage for Chinese IPO firms. We document that affiliated analysts forecasts of firm 21

23 performance is highly positively biased. This positive bias is way beyond the normal analyst optimism for similar firms, and is higher than the optimism of unaffiliated analysts for the same IPO firm. Investigation of the long-run stock performance suggests that affiliated analysts hype increases both the IPO offer price and the first-trading-day price. More aggressive hyping is associated with higher current prices, and therefore lower long-run returns. A one-standarddeviation increases in (positive) analyst forecast bias leads to a 13-38% decrease in the threeyear return relative to the first trading day price, and 20-50% decrease in the three-year return relative to the offer price. IPO investors, however, do not appear too concerned about analyst hype. Retail demand for IPO shares does not decrease and actually increases with analyst hype. Institutional demand does not vary with hype. Institutional investors adjust downward their bid prices when hype is aggressive, but the adjustment is not nearly adequate. In addition, they give top ratings to the analyst research report two-thirds of the time. There is only weak evidence that ratings decrease with the extent of forecast bias. IPO investors insensitivity to analyst hype is explained by the underpricing of the IPO stock relative to the short-term post-market prices. This is consistent with the sentiment theory of Ljungqvist, Nanda and Singh (2004) and Derrien (2005) who argue that underwriters increase the IPO offer price with investor sentiment, but still below what sentiment investors are willing to pay (i.e., the immediate post-market price). So IPO investors still enjoy high returns if they sell the stock soon after the IPO, despite the inflated offer price and the short-term post-market prices. We also find evidence that underwriters have monetary incentives to hype IPO stocks. More aggressive hyping will give them higher RMB underwriting fees from the current IPO, and also more future investment banking business from both the current client and other IPO firms. 22