Efficiency of ipo pricing mechanisms: Comparison of book-Building and fixed price methods in China

JOURNAL OF SCIENCE, Hue University, Vol. 70, No 1 (2012) pp. 121-132 121 EFFICIENCY OF IPO PRICING MECHANISMS: COMPARISON OF BOOK-BUILDING AND FIXED PRICE METHODS IN CHINA Ho Tu Linh1, Wang Yixia2, Nguyen Dinh Chien2 1College of Economics, Hue University, Vietnam 2School of Management, Huazhong University of Science and Technology, China Abstract. This paper compared the efficiency of IPO pricing mechanisms namely Fixed Price (FP) and Book-Building (BB) in China by analyzing IPOs' underpricing level. Findings include (1) the FP regime was more efficient than the BB one because the BB did not reduce the underpricing level in China as expected; and (2) reasons for this were (i) information transparency of the BB has not reduced other external effects, for example, impacts of some firm quality and ex- ante uncertainty proxies were still tight or even stronger since 2005; and (ii) investors were optimistic toward market conditions as higher quality underwriters invited in the BB procedure as well as the fact that underwriters often lowered offer price after collecting information from the BB process, which made the underpricing level rise. Keywords: IPOs, Fixed Price, Book-Building, underpricing. 1. Introduction Initial Public Offerings (IPOs) are priced by distinct mechanisms in different nations. Before 1990, Auction and Fixed Price (FP) were commonly adopted. In the recent two decades, Book-Building (BB) and its hybrids have nearly become dominant all over the world (Jagannathan et al., 2010). Different IPO pricing mechanisms have different efficiency levels of underpricing. Kucukkocaoglu and Sezgin (2007) found some mixed results supporting this point. BB was considered as the most effective method because it resulted in more efficient pricing for sellers compared to FP and Auctions almost half of the studies listed in their research. The underpricing phenomenon in Chinese stock market has been studied since the 1990s. The dramatic high level of IPOs underpricing in the early years of this equity market is often researched and quoted as an emerging-market problem in many studies. According to Ritter and Welch (2002), average Initial Returns (IR) in China from 1990- 2000 was the highest at 256.9%. In 2005, the BB pricing mechanism was introduced for 122 Efficiency of IPO pricing mechanisms: comparison of book-building and Chinese IPO pricing regulation by the China Securities Regulatory Commission (CSRC), which is considered as a milestone of transforming from FP to BB in this country. The first purpose of this transformation is to make offer price more exact and deals with the high level of underpricing issue which nearly all IPOs listed on China’s A-stock market suffer on the first trading day. However, the transformation's results were unexpected because while most recent studies found a drop of the underpricing level after 2005, Fei Jiehui (2009) pointed out a different trend in China’s IPO market. Hence, in order to find out a uniform answer for the current IPO pricing mechanism’s efficiency, this paper attempted to explore two research questions: (1) in China, whether the current pricing mechanism - the BB was more efficient than the previous one - the FP; and (2) why the BB mechanism was more or less efficient in China. To discover those questions, a set of 709 IPOs listed on Shanghai and Shenzhen stock market between 2001 and 2009 was used, among which 409 IPOs were listed after Jan. 1st, 2005 - called the post-BB regulation period. The period before 2005 was the pre-BB regulation one. Methods for analyzing factors believed significant in determining the underpricing level were regression models’ establishment and statistical analysis by a software namely Statistical Package for the Social Sciences (SPSS). 2. Study models In this paper, Market Adjusted Initial Return (MAR) were calculated in order to explain the underpricing level of IPOs. MAR was used as dependent variable in empirical study because it kept away the influence of market conditions and therefore was more accurate for detecting the efficiency of IPO pricing. The dependent and independent variables used in this paper were mainly chosen based on past empirical studies on China's IPOs and proved significant in explaining the underpricing phenomenon. Their definitions were listed in Table 1. Three study models used in this paper are: Model 1: MARi = β0 + β1BKBDi + β2EXCHi + εi Model 2: MARi = β0 + β1GOVOWNi + β2CNTOWNi + β3ONEYRPRFi + β4EPSi + β5ROEi + β6NICAGRi + β7INDUSTRYi + β8LGIPOPRCi + β9DEi + β10PRIEPSi + β11AGEi + β12PRMKTRETi + β13TIMELAGi + β14LOTTODDi + β15UNDRWRTRi + β16EXCHi + εi Model 3: MARi = β0 + β1GOVOWNi + β2CNTOWNi + β3ONEYRPRFi + β4EPSi + β5ROEi + β6NICAGRi + β7INDUSTRYi + β8LGIPOPRCi + β9DEi + β10PRIEPSi + β11AGEi + β12PRMKTRETi + β13TIMELAGi + β14LOTTODDi + β15UNDRWRTRi + β16EXCHi + β17PRIRANi + β18PRIADJi + εi HO TU LINH, WANG YIXIA, NGUYEN DINH CHIEN 123 Table 1. Definition of Variables Variables Definition Dependent Variable: MAR The percentage increase of IPO’s first-trading day closing price from its offer price, adjusted by market index return in this same period. 100(%) 0 1 0 1 ×      −= M M P PMAR Independent Variables: Proxies for firm quality GOVOWN Percentage of government or state-owned enterprise ownership of IPO post offering CNTOWN Percentage of controlling stockholder ownership post offering ONEYRPRF Market-adjusted cumulative return of IPO stock one year post offering. (P1 and MI1 are IPO’s price of the first listing date of each month and market index of the same date respectively. P21 and MI21 are values in the 21 listing date of the month i = 11 ,1 ; Yearly Initial Return-YIR; Yearly Market Return-YMR): 1 121 i (%) P PPIR −= ; 1 121 i (%) MI MIMIMR −= ; )MR...MR(MR)IR...IR(IRYIMYIRONEYRPRF 11211121 +++−+++=−= EPS Historical average earnings per share in last three years before IPO offering: 3 321 −−− ++ = EPSEPSEPSEPS ROE Historical average return on equity in last three years before IPO offering: 3 321 −−− ++ = ROEROEROEROE NICAGR Compound Annual Growth Rate of Net income in last three years before offering 2 2 21 3 32 − −− − −− − + − = NI NINI NI NINI NICAGR 124 Efficiency of IPO pricing mechanisms: comparison of book-building and INDUSTRY Industry dummy: if IPO belongs to manufactoring industry, INDUSTRY=1; otherwise 0 Proxies for ex-ante uncertainty LGIPOPRC Logarithm of IPO proceeds DE Historical average Debt/Equity in the last three years before IPO offering ii i EquityDebt DebtDE + =i ; 3 321 −−− ++ = DEDEDEDE PRIEPS Offer Price/earnings per share before offering AGE Time between the date of IPO firm filing registration with China’s Commerce Bureau and listing date Proxies for controlling BKBD Book-building period dummy: if IPO listed after 2005, BKBD=1; otherwise 0 PRMKTRET Market index’s cumulated return 30 days before offering TIMELAG Lag of time between the date of IPO offering and listing LOTTODD Odd of winning the lottery of online IPO allocation, equal to reciprocal of oversubscription ratio UNDRWRTR Underwriters’ ranking, UNDRWRTR=0 if IPO’s total capital (TC) > 1000 billion RMB, 1 if 100 < TC ≤ 1000, 2 if 10 < TC ≤ 100, 3 if 1 < TC ≤ 10, otherwise 4 EXCH Exchange dummy: EXCH=1 if IPO is listed on Shanghai Stock Exchange, otherwise 0 Other proxies PRIRANGE Percentage of offer price range settled in preliminary book-building process. PRIADJUST Percentage adjustment of final offer price from the expected offer price (arithmetic average of high-end and low-end prices in price range) implied in preliminary BB. 3. Hypothesis development In this section, hypotheses were developed according to previous researches. Regarding the theme of (1) whether BB was more efficient than FP in China or not, although the efficiency of IPO pricing mechanisms is on debates, a lot of international HO TU LINH, WANG YIXIA, NGUYEN DINH CHIEN 125 researchers have believed that the BB efficient was over others’. Beside this, excluding one domestic scholars of China, most authors found evidence proving the BB’s ability in significantly reducing the IPO underpricing level. Moreover, Fei Jiehui (2009) who found the negative results of this method predicted that there would be a better scenario for its underpricing level in the long run. So, we assumed Hypothesis 1 to be tested by Model 1 (Table 2). In terms of the second question of (2) why the BB mechanism was more or less efficient in China, two key possible reasons were information asymmetry related to firm characteristics and uncertainty and investors’ optimistic toward market conditions. Basing on reviewing a literature, other 5 hypotheses were assumed and tested through Model 2 and Model 3 (Model 3 equals Model 2 plus PRIRANGE and PRIADJUST). Table 2. Research Questions, Hypotheses and Study Models Question Hypothesis Model (1) Whether BB was more efficient than FP in China or not? 1. Book-Building was more efficient than Fixed Price in China. 1 (2) Why the BB mechanism was more or less efficient in China? 2. Effect of firm characteristics on underpricing significantly decreased after BB was introduced in China. 2 3. Effect of ex-ante uncertainty on underpricing significantly decreased after BB was introduced in China. 4. Private companies significantly explained the underpricing level in China, especially after using the BB mechanism. 3 5. Higher quality underwriters in the BB procedure significantly reduced the underpricing level on both SHSE and SZSE. 6. In preliminary BB process, higher percentage of offer price range settled and higher percentage adjustment of final offer price from the expected offer price implied would associated with higher underpricing on both SHSE and SZSE. 126 Efficiency of IPO pricing mechanisms: comparison of book-building and 4. Data analysis and findings After conducting the Komogorov – Smirnov test, the normality condition was assured among the dependent and independent variables before they were regressed against each other. This part shows results from analyzing the regression models. 4.1. Comparison of Book-Building and Fixed Price methods in China Hypothesis 1 was tested by Model 1 conducted on three groups: full sample, sub-sample on SHSE and sub-sample on SZSE. Table 3 represents results of regression for effectiveness of BB on the three groups. For the group of full sample, BKBD had a positive co-efficient with MAR but it was not significant. This does not support Hypothesis 1 that the underpricing level significantly decreased after the BB method was introduced. In addition to this, EXCH marked a significant negative co-efficient with the underpricing level at -.262, which indicates that IPOs listed on SHSE seemed to have lower underpricing level than those listed on SZSE. Table 3. Regression Result for the Effectiveness of Book-Building, 2001-2009 Coefficient Model 1 (Full Sample) Model 1 (Sub- SHSE) Model 1 (Sub- SZSE) β0 1.215*** 1.035*** .676*** β1(BKBDi) .007 -.274*** .617*** β2(EXCHi) -.262*** - - No. of IPOs 709 369 340 Adjusted R2 .021 .027 .039 (Note: * p < 0.1, ** p < 0.05, and *** p < 0.01). Table 3 showed the co-efficients of BKBD for both groups of SHSE and SZSE which were really significant on 5% level. However, while the underpricing level decreased by 27.4% on average after BB was used for IPOs on SHSE, it grew with a faster speed of 61.7% on SZSE. Therefore, the effect of BKBD on MAR of the full sample as a whole was weaker than that of SHSE or SZSE separately. To sum up, the finding was obviously against Hypothesis 1 that with the introduction of BB regime, IPO underpricing on China’s A-stock market has not decreased on SZSE as well as on the full sample of 709 IPOs, which means that the BB mechanism was not more efficient than the previous method of FP in China. 4.2. Reasons why Book-Building was more efficient or not 4.2.1. Information asymmetry related to firm characteristics and uncertainty To test Hypothesis 2 and 3, the significance of theory proxies before and after HO TU LINH, WANG YIXIA, NGUYEN DINH CHIEN 127 BB were compared by Independent Sample T-test. Table 4 displays the comparison results of theory proxies on the 3 groups of full sample, SHSE and SZSE sub-sample. The result of the comparison is represented by p-value. Low level of p-value indicates a significant change in the effect of coefficient after BB was introduced. a) Information asymmetry related to firm characteristics The first 7 variables of GOVOWN, CNTOWN, ONEYRPRF, EPS, ROE, NICAGR and INDUSTRY revealing characteristics of firms in Table 4 indicated a limited support for Hypothesis 2. On the full sample, although the proxies of ONEYRPRF and INDUSTRY provided a significant fall in the significance of coefficients, others showed opposite results. For example, CNTOWN, EPS and NICAGR indicated that Hypothesis 2 cannot be accepted due to unexpected increases in their effect on underpricing level instead of an assumed downward tendency. Excluding NICAGR, such change was significant with p-values of 0.023 and 0 for CNTOWN and EPS respectively. Table 4. Comparison of the Significance of Explanatory Variables before and after Book- Building, 2001-2009 Coefficient Model 2 (Full Sample) Model 2 (Sub-SHSE) Model 2 (Sub-SZSE) Pre-BB Post-BB p- value Pre-BB Post-BB p- value Pre-BB Post-BB p- value β0 11.939*** 8.426*** 13.015*** 7.553*** 20.972** 11.481*** β1(GOVOWNi) .181 -.119 .422 .393* -.394 .000 -1.545** .157 .411 β2(CNTOWNi) -.211 .349 .023 -.488 .414 .000 1.610 .504* .112 β3(ONEYRPRFi) -.320** .012 .000 -.247 -.679*** .000 -1.974 -.792*** .000 β4(EPSi) .047 -.057 .001 .046 .044 .010 1.256 -.022 .023 β5(ROEi) -1.358*** .155 .061 -1.213*** -.939 .000 .924 .246 .093 β6(NICAGRi) .011 .059 .975 .014* -.071 .114 .797 .069 .015 β7(INDUSTRYi) .075 -.040 .071 .098 -.573 .673 .171 -.035 .000 β8(LGIPOPRCi) -1.276*** -.764*** .000 -1.332*** -.621** .000 -2.598** -1.179*** .018 β9(DEi) -.265 .078 .000 -.352 -.402 .000 -2.219* .449 .762 β10(PRIEPSi) .000 .008 .013 -.001 -.004 .000 .031 -.021** .046 β11(AGEi) -.056*** -.012 .000 -.050*** .005 .000 -.053 -.006 .000 β12(TIMELAGi) .015*** -.040*** .853 .014*** -.002 .008 -.003 -.010 .002 128 Efficiency of IPO pricing mechanisms: comparison of book-building and β13(PRMKTRETi) 1.836*** 1.784*** .000 2.297*** .859 .202 - 11.118*** 2.261*** .000 β14(LOTTODDi) .182* -.206 .000 .174 .046 .000 4.683** .169 .001 β15(UNDRWRTRi) -.096*** .036 .085 -.101** -.054 .253 -.137 .080* .751 β16(EXCHi) .486*** .434 .000 - - - - - - No. of IPOs 300 409 709 260 109 369 40 300 340 Adjusted R2 .357 .178 .374 .469 .074 .409 (Note: * p < 0.1, ** p < 0.05, and *** p < 0.01). On SHSE, the firm quality effect was stronger post-BB, which was confirmed by increases in the significance of GOVOWN, ONEYRPRF, NICAGR and INDUSTRY. Obviously, GOVOWN did not satisfy Hypothesis 2 because its significance for two groups of full sample and SZSE has insignificantly decreased after BB with high level of p-Value. Besides, the SHSE group experienced a significant rise with a zero p-value of the effect of GOVOWN post-BB. For the SZSE, the effect of the 7 firm quality proxies on underpricing fallen post-BB but only ONEYRPRF and INDUSTRY really marked significant changes. b) Information asymmetry related to firm uncertainty Hypothesis 3 received a mixed approval by 4 variables of LGIPOPRC, DE, PRIEPS and AGE in Table 4. LGIPOPRC and AGE provided strong supports for Hypothesis 3 that the ex-ante uncertainty effect on underpricing significantly decreased post-BB on both SHSE and SZSE. DE and PRIEPS ratio gave limited supports due to the drop of significance was not obvious for DE and the increase in significance was clear for PRIEPS. Again, except for an insignificant fall of DE’s effect, the effect of firms’ ex-ante uncertainty on SZSE was significantly weaker, whereas only 2 of 4 variables described the similar situation on SHSE. 4.2.2. Investors’ optimistic toward market conditions The regression result of Model 3 was based on the post-BB subsample (Table 5). In order to examine Hypothesis 4, we take a look at Table 4 first, then Table 5. Table 5. Regression Result on Information Variables in Book-building Process Coefficient Model 3 (Full Sample) Model 3 (Sub- SHSE) Model 3 (Sub- SZSE) β0 6.882*** 8.067*** 9.550*** β1(GOVOWNi) -.175 -.419 -.009 HO TU LINH, WANG YIXIA, NGUYEN DINH CHIEN 129 β2(CNTOWNi) .317 .471 .354 β3(ONEYRPRFi) -.028 -.639*** -.712*** β4(EPSi) -.121 .015 -.049 β5(ROEi) .803 -.956 .605 β6(NICAGRi) .061 -.082 .082 β7(INDUSTRYi) -.028 -.601* -.045 β8(LGIPOPRCi) -.620*** -.670*** -.916*** β9(DEi) .071 -.316 .472 β10(PRIEPSi) .013** -.004 -.004 β11(AGEi) -.001 .008 -.002 β12(TIMELAGi) -.034*** -.003 -.020 β13(PRMKTRETi) .979** .838 1.746*** β14(LOTTODDi) -.201 .021 .221 β15(UNDRWRTRi) .088 -.090 .126** β16(EXCHi) .401 - - β17(PRIRANGE i) -.483* .060 -.968*** β18(PRIADJUST i) -4.270*** -1.370 -3.676*** No. of IPOs 409 109 300 Adjusted R2 .304 .480 .436 (Note: * p < 0.1, ** p < 0.05, and *** p < 0.01). From Table 4, it is clear that EXCH decreased from more to less positive coefficience with MAR. Although the change was significantly marked with p-Value of 0, the coefficient between EXCH and MAR was only significant pre-BB. This meant that after BB, the underpricing level of SHSE was still higher than that of SZSE but with an insignificant relationship, which did not support Hypothesis 4. Besides, for the full sample and SHSE, coefficient of CNTOWN with MAR changed from negative to positive post-BB, whereas it decreased from high to low positive coefficient on SZSE, which meant that IPOs had higher percentage of controlling stockholder ownership post offering associated with higher underpricing level. However it only marked a significant rise in its affect on underpricing for full sample while one significant decrease was revealed on SHSE. On SZSE, there was also a fall though insignificant. Hence, after BB was introduced, higher CNTOWN caused higher underpricing level on the full sample 130 Efficiency of IPO pricing mechanisms: comparison of book-building and and SHSE. To sum up, EXCH and CNTOWN obviously offered a limited support to Hypothesis 4. Similarly, Table 5 supplied more evidence to deny Hypothesis 4. First, the positive relationship between CNTROWN and MAR of SZSE sub-sample, which was 0.354, was stronger than that of full sample but weaker than the SHSE sub-sample with 0.317 and 0.471 respectively. Second, a coefficient index of EXCH and MAR was also positive, which meant the underpricing level of SZSE was lower than that of SHSE post-BB. Overall, Hypothesis 4 received a weak support. Private companies did not tend to affect the underpricing level in China. Although SZSE has known as the exchange which listed more private companies than SHSE did, the underpricing level of SZSE was nearly as high as that of SHSE, especially after applying the BB. Hypothesis 5 was not supported by UNDRWRTR. Table 5 showed that for the full sample and the SZSE sub-sample UNDRWRTR sign was positive with MAR but with insignificant coefficiency while their relationship was significant negative in the SHSE. Because UNDRWRTR was marked between 0 and 4 from the first cla