Do experience and luck affect the behavior of institutional investors. in IPO markets? Abstract

Transcription

1 Do experience and luck affect the behavior of institutional investors Seth Armitage Business School, University of Edinburgh Youyan Fu Business School, University of Edinburgh Ufuk Güçbilmez Business School, University of Edinburgh in IPO markets? WORK IN PROGRESS, PLEASE DO NOT CITE Abstract We use a unique set of bookbuilding data to examine the impact of personal experience and luck on the behavior of institutional investors in an IPO market. We find that, when deciding to participate in future IPOs, institutions take into account initial returns of the IPOs they participated in the past, regardless of the fact that they received a share allocation or not. While this type of behavior is consistent with Bayesian learning, we also find that institutions participate more often in future if they personally experienced a large gain from the past IPOs and if they were lucky with the share allocations in the past. In the field of financial economics, rational Bayesian belief updating and naive reinforcement learning are the two leading theories on an agent s learning behavior. Naïve reinforcement learning refers to a strengthening of the behavior that is followed by an experienced-appropriate consequence (Dinsmoor,2004). Moreover, naïve reinforcement learners pay attention only to past payoffs, not to the causations that generate those payoffs. With respect to rational Bayesian belief learning theory, it premises that learners keep track of forgone outcomes as well as personally experienced ones, and then form their beliefs based on the updated information.

2 Afterwards, learners tend to make future decision according to the rationally formed beliefs (Camerer and Ho (1999)). Therefore, the first distinction between these two theories is that naïve reinforcement learner gives more weight to personally experienced outcomes than observations whereas Bayesian belief-learner allocates equal weight to these two aspects. Put differently, directly experienced outcomes have more impact on future decision under naïve reinforcement, but for rational Bayesian belief learning, actual and observed outcomes are equally influential. With respect to the learning process, the second difference is that Bayesian belief-learners rationally learn from their past experiences but naïve reinforcement learners do not. The original idea of naïve reinforcement learning, the law of effect, is due to Thorndike (1911). Through a famous experience named the puzzle box, he investigates the learning process of animals and suggests that behaviors that generate good outcomes are likely to be repeated in the future. Build on the law of effect, Skinner (1938) and Zeiler (1968) formally establish the model of naïve reinforcement learning in the psychological realm. Cross (1973), Arthur (1991) and Roth and Erev (1995) introduce this psychological theory to the economics literature. They show that naïve reinforcement learning theory is more powerful in describing human behavior in experiments than predictions based on rational human beings. Researchers not only pay attention to naïve reinforcement learning but also construct theoretical models to investigate Bayesian learning process (Crawford (1995),Cheung and Friedman (1997), Crawford and Broseta (1998)). However, naïve reinforcement learning and rational Bayesian belief learning are mostly treated as fundamentally different in these researches. To integrate these two learning theories, Camerer and Ho (1999) build a hybrid learning model named experience-weighted attraction (EWA). They introduce a parameter to measure the payoffs that would have yielded relative to the payoffs that are actually received, by which EWA model shows better explanatory power than pure naïve reinforcement model and rational Bayesian learning model in most of game experiments. In this research, we use a unique set of bookbuilding data to explore the impact of experience and luck on institution s future investment behavior in IPO market and

3 further find out which learning pattern(s) institutions are subject to. The unique IPO allocation mechanism in our sample explicitly generates actual and forgone payoffs, which are the key elements for Bayesian learning and naïve reinforcement learning. Based on these two types of returns, we distinguish one learning theory from the other and examine the hybrid learning model of Camerer and Ho (1999) by empirical data instead of experiment. There are two researches that particularly investigate investor s learning behavior in IPO market. The pioneer one comes from Kaustia and Knüpfer (2008). They reveal that individual investor s bidding behavior is consistent with naïve reinforcement learning theory using a data set of 183,000 individual investors in 57 Finnish IPOs from Jan 1995 to Dec However, they do not test the behavior of institutional investors. Following Kaustia and Knüpfer (2008), Chiang et al. (2011) do a further research based on 17,000 winning bids in 84 Taiwanese IPO auctions during 1995 to Beside the positive relationship between past return and the likelihood of participating in future IPOs, they find individual investor s auction selection ability deteriorates as more experiences are obtained. According to this finding, they conclude that individual investor s behavior is consistent with naïve reinforcement learning rather than rational Bayesian learning, where individual investors become irrationally optimistic after experiencing good outcomes. In addition to individual investors, they also investigate the behavior of institutional investors but do not find any learning pattern for institutions. There are three differences between this paper and Kaustia and Knüpfer (2008) and Chiang et al. (2011). First of all, the unique allocation mechanism in Chinese IPO market results in two types of returns, namely actually experienced return and missed return. Since the fundamental difference between Bayesian learning and naïve reinforcement learning is the various weights apportioned to experienced payoffs and forgone payoffs, and the two kinds of returns in our data are consistent with the theoretically defined payoffs, it allows us to explicitly identify which learning theory(ies) is(are) followed by institutional investors. On the contrary, Kaustia and Knüpfer (2008) and Chiang et al. (2011) conclude investor s learning pattern as naïve

4 reinforcement learning merely based on the result they got without explicit identifications. In addition, the actual and missed payoffs in our sample produce psychological effect. Therefore, this research is able to test whether luck affects investor s investment decision. Secondly, Kaustia and Knüpfer (2008) and Chiang et al. (2011) mainly focus on individual investors, whereas we specifically test the learning behavior of institutions. Compared to individual investors, institutional investors are likely to be better informed and regarded as being sophisticated (Cohen et al. (2002), Nagel (2005), Chiang et al. (2010)). This implies the learning process of institutional investors could be different from individual investors due to their distinct characteristics. Thirdly, the IPO mechanism in our research is unique. The sample used in Chiang et al. (2011) is IPO data in Taiwan where auction is used as pricing mechanism. However, the pricing mechanism in our sample is bookbuilding which is same as Kaustia and Knüpfer (2008). Comparing with Kaustia and Knüpfer (2008), one advantage of our dataset is that institutional investors play price-setting role and underwriters have no discretion over allocation in our dataset. Hence, our result will not be driven by the effect of underwriter discretion. On the hand, the implications of learning theories are studied in finance literature as well. Feng and Seasholes (2005) and Dhar and Zhu (2006) find that sophistication and trading experience can lower the disposition effect 1. Seru et al. (2010) also examine the learning patterns of investors. They find two types of learners: some investors become better at trading as they gain experience, while others cease trading when they realize their investment ability is poor. To the best of our knowledge, this is the first research that explicitly explores the learning behavior of institutions in IPO market. We find that, some institutions such as security firms have inherently high propensity of participating in IPOs. Institutions take into account initial returns of the IPOs they participated in the past, regardless of the fact that they received a share allocation or not. This implies Bayesian learning plays role in institution s learning process. Institutions participate more often in future 1 The behavioral bias of reluctance to realize losses and diminishing the tendency to realize gains.

5 if they personally experienced a large gain from the past IPOs, which hints that naïve reinforcement learning has impact on institution s future decision as well. Finally, institutions are more likely to participate in future IPOs if they were lucky with the share allocation in the past. The remainder of this paper is organized as follows. In section 1, we introduce the background of Chinese IPO market. Section 2 describes the data. Section 3 addresses the hypotheses and Section 4 presents the methodologies and the results. Section 5 concludes. 1. The Background of Chinese IPO Market In China, IPOs typically include two separate tranches: offline offering and online offering. Normally, 20% 2 of the total issued shares are placed to offline offering where only institutional investors are eligible to subscribe. The most important function of offline offering is to set offer price. Bookbuilding is used as the primary pricing mechanism in our IPO sample. Through bookbuilding, underwriter and issuer collect information of bid price and bid amount from institutional investors and then set the offer price. Once the offer price is settled, individual investors start subscribing for the rest of 80% 3 issued shares through online offering by which individual investors inform underwriter how many shares they would like to buy. In other words, individual investors have nothing to do with setting the offer price. Another significant difference between retail and institutional offering is related to share allocation mechanism. For retail offering, share allocation is conducted by a lottery process where underwriter ballot to decide the amount of shares individual investor receive in one IPO. This means that whether individual investors could obtain shares totally depend on how lucky they are. It is worth noticing that only institutional investors whose bids at or above the offer price, hereinafter referred to as qualified 2 Due to decree No.78, China Securities Regulatory Commission (CSRC) increases the percentage of shares allocated to institutions to 50% from 20% since May 18, In our sample, only 36 IPOs are affected by this regulatory revision. 3 This proportion changes to 50% after regulatory revision on May 18, 2012.

6 bid 4, are eligible for the lottery process. The share allocation mechanism of institutional offering is same as retail offering until now since the IPO regulatory revision on 1st Nov Before the regulation revision, institutional investors pro rata obtain shares according to their bid amounts in bookbuilding. It means that institutional investors will definitely receive shares if their bids are not less than the offer price. This research focuses on IPOs after 1 st Nov 2010 when lottery allocation is used instead of rationing. A typical timeline of Chinese IPO is as follows, where T is the day of offering. T-6: T-5: T-2: The preliminary prospectus and a notice for the bookbuilding are issued. Bookbuilding takes place. The offer price is set. T-1: The final prospectus is issued. The order book is published. An online roadshow takes place for retail investors. T: Subscriptions for the institutional and retail offerings take place. T+1: T+2: T+3: Balloting for the institutional offering takes place. Balloting for the retail offering takes place. Allocations to institutions are announced. Allocations to individuals are announced. 2. Data We hand-collect institutions bids information from 214 IPOs in ChiNext, a part of the Shenzhen Stock Exchange designed for high-technology Chinese firms, from Nov 2010 to Sep In total, there are 353 unique institutions that take part in these 214 IPOs. The bid data is collected from legally required document released by issuing firms, in which we can observe institution name, account name 6, bid amount 7 and allocation amount 8. The descriptive statistics of bookbuilding are presented in Panel A of Table 3. 4 Oppositely, bids with prices lower than the offer price are defined as unqualified bids. 5 There also has 141 IPOs fulfilled in ChiNext before IPO regulatory revision on 1st Nov We exclude these IPOs from our research sample because we cannot fully know that one specific bid is submitted by which institution before this revision. In addition, the allocation method in the former period is rationing such that it is infeasible to distinguish Bayesian learning and naïve reinforcement learning. 6 It refers to investment products under the management of institutions. 7 The amount of shares subscribed by each account in bookbuilding of one IPO. 8 The amount of shares obtained by each account in one IPO.

7 With respect to the IPO data, offer price, the number of shares issued, proceeds, issuing firm s profitability and price-to-earnings ratio are obtained from the official website of Shenzhen Stock Exchange. Issuing firm s founding year, issuing date and trading date are collected from Thomson One Banker. In addition, we collect data from the SDC to double check the accuracy of the data in IPO date, trading date, offer price, number of shares and proceeds. The definitions of other variables are as follows. The unadjusted initial return is the change between the offer price and the closing price on the first trading date. Considering the market effect between the day when offer price is set and the first trading day, we use the change of Shenzhen A-Share Stock Price Index during the waiting period to control it and get adjusted initial return. Firm age is calculated by the number of years between the foundation year and the IPO year. P/E ratio is measured by the scale of offer price to the average price-to earnings ratio of peers. The descriptive statistics of IPOs are given in Panel B of Table Hypotheses In terms of naïve reinforcement learning theory, decision makers would like to repeat actions that generated favorable experience in the past. This theory implies that institutions are more likely to participate in future IPOs once they experienced high returns. Attentively, the return herein must be actually experienced by institutions based on the definition of naïve reinforcement learning. Because of the lottery allocation mechanism in our sample, returns can be realized only if institutions are lucky enough to get shares. Therefore, the impact of actually realized return on institution s future decision can reflect the extent to which naïve reinforcement learning explains institution s learning behavior. With respect to Bayesian belief learning, investors update their beliefs about IPOs through observing past experience and then make future investment decision according to the updated beliefs. Particularly, Bayesian learners do not only review their personally experienced outcomes but also those that would have been occurred,

8 namely actual return and missed return 9. Although Bayesian learning theory also suggests favorable past return will motivate institutions to participate in more IPOs, the motivation equally comes from actual and missed return rather than actual return itself, which is the key characteristic comparing with naïve reinforcement learning. Since the fundamental difference between rational Bayesian learning and naïve reinforcement learning is the weights allocated to experienced payoffs and forgone payoffs. Thus, the extent to which actual return and missed return affect future decision can be used to explicitly identify which learning theory(ies) is(are) followed by institutional investors. Hypothesis 1: If institution s learning behavior is subject to Bayesian learning, both high actual and missed high return in the past will motivate institutions to participate in IPOs and the effect from these two returns are equal. Alternatively, if institution s learning behavior is consistent with naive reinforcement learning, the impact of actual return should be more than missed return. If Bayesian and naïve reinforcement learning work interactively, it implies that the hybrid learning model from Camerer and Ho (1999) is valid. In China, IPO shares are heavily oversubscribed such that the chance of getting shares is extremely low. For the 214 IPOs in our sample, the average allocation rate to institutional investors, which is calculated as the number of shares offered to institutions divided by the number of shares demanded by institutions, equals to 3.13%. This fact raises a question that whether luckiness sway institution s future investment decision. For instance, one institution took part into ten IPOs in the past and only got shares from one of them due to the lottery allocation. Relative to the other nine IPOs, whether the return from the only one winning IPO is high or low is randomly determined by institution s luck. Therefore, we use the difference between experienced return and missed return to measure institution s luckiness and test if luck matters. If luckiness dose affect institution s future decision, we will further examine whether institutions overweight their luckiness when deciding to participate in future. 9 Missed return results from bids that do not get allocation in lottery.

9 Hypothesis 2a: If high difference between experienced return and missed return makes institutions to participate in more IPOs, it shows luck have impact on their future decisions. Hypothesis 2b: If the weight allocated to luckiness more than to other returns, it implies institutions overweight their luckiness when they make future decisions. Apart from the past experience about returns rate, monetary gain generated from previous IPOs could also affect institution s future investment behavior. For example, one institution can earn enormous money from IPOs even if the return rate is relatively low. In this case, institution may think of more about absolute gain rather than return rate when making future investment decision. We propose that high monetary gain in the past will prompt institutions to participate in more IPOs later. Hypothesis 3: If institutions participate more often in future after experienced a large gain from the past IPOs, it indicates that monetary gain matters for decision makers. 4. Methodology and Results 4.1. Univariate Tests First of all, we divide our sample into sub-period A and B that have the same number of IPOs. The sub-period A includes 107 IPOs with 9584 bids and the sub-period B involves 107 IPOs with 9567 bids. For each institution, we calculate the times of participation in each sub-period and, the times of getting allocation in sub-period A and the nominal average IPO returns in sub-period A. Due to the lottery allocation, can be decomposed into two parts which are the actually experienced average return 10 and the missed average return 11 respectively, where and are weighted by the scaled times of winning lottery and the scaled times of losing lottery. Therefore, we have the following equation: 10, where is the adjusted initial return for IPO i in which institution obtains shares. 11, where is the adjusted initial return for IPO i in which institution do not obtain shares.

10 12 In this test, we investigate whether the past experience influence institution s future investment propensity which is measured by. Beside the past experience, inherent investment appetence could also affect institution s behavior. For example, institutions such as fund and security companies have naturally high propensity to take part in IPOs. To control this effect, we divide institutions into three groups according to based on. Within each group, we further split institutions into two sub-groups. In addition, we exclude 80 institutions that do not participate in sub-period A, i.e., because they do not have any experiences in the past. Figure 1 illustrates the relationship between the nominal return in sub-period A and the times of participating in IPOs in sub-period B after controlling natural investment propensity. For the middle 13 and high 14 participation group, we find clearly positive pattern between and for both mean and median. However, the mean of marginally increases by 0.12 from the low return group to the high return group for low-frequency participators, which results from these institutions participate in IPOs occasionally 15 so that past experience does not matter so much for them. According to the univariate test, we discover that institution with more participation in sub-period A are more likely to participate in sub-period B IPOs, namely inherent investment propensity dose affect institution s investment decision. Having controlled the natural investment propensity, we find past nominal return has positive impact on the frequency of participating in IPOs in the future. Although this relationship has been revealed, how the decomposed returns and affect institution s future investment decision is not clear. Therefore, multivariate test is conducted in the following part in order to explore the significance of their impacts. 12 Lichtenberg (1900) and Lach (1933) suggest that, the OLS model suffers from omitted variable bias when the aggregate variable is used as explanatory variable but the true model is constructed by decomposed variables. 13 The mean and median of for middle participation group are 9.92 and 8 respectively. 14 The mean and median of for high participation group are and 41 respectively. 15 The mean and median of for low participation group are 1.64 and 1 respectively.

11 4.2. Multivariate Test Past return rate and future behavior For testing to what extent and influence institution s future behavior, we conduct the following regression: where is and is Since and are ordinal variables, we use logarithm format Log ( ) and Log ( in this regression. The dependent variable is proxy for future participating propensity; is used to control institution s inherent investment tendency. Importantly, the coefficients of and are the key measurements used to detect how institutions pay attention to experienced return and missed return respectively. Recall hypothesis1, if institutional investors weight the returns they have experienced and those they have not equally ( ), their learning behaviors are consistent with rational Bayesian belief learning. If institutional investors care more about the returns they have experienced than those missed ( ), institutions are subject to naïve reinforcement learning. Alternatively, if institutions give more weight to missed return, it indicates that the hybrid learning model from Camerer and Ho (1999) works. 错误! 未找到引用源 presents the result of the multivariate tests. In model 1, we regress only on the nominal return of sub-period A. We can see that has significantly positive impact on the frequency of participating in future IPOs, which indicates that past experience does affect institutions future decisions, which is consistent with univarite tests. This result supports both rational Bayesian learning and naïve reinforcement learning. To distinguish these two learning theories, we split the nominal return into experienced return and missed return and regress on these two returns in model 2. In the last row, it presents the p-value of equality test where the null hypothesis is. In model 2, the p-value of testing is 0.732, which implies the null hypothesis of

12 cannot be rejected. Referring to hypothesis 1, this result reveals that institutions give equal weight to the experienced and missed return. However, the experienced return is not significant in model 2, which could due to many institutions do not get any shares 16 in sub-period A so that the variance of is too low. In model 3, therefore, we exclude 144 institutions with regression as model 2. The t-value of from our sample and run the same increases from 0.67 to 1.07 although the impact of is still insignificant. Meanwhile, the result of becomes more strong. On the other hand, we find the coefficient of and are much closer to each other than they are in model 2, which supports Bayesian learning. In model 2 and model 3, we weight the experienced return and missed return based on the scaled times of getting shares. The extremely low probability of getting shares leads to a quite low and high such that most of the nominal return is attributed to the missed part 17. In reality, however, institutions may not treat these two returns as the way we did. They could just calculate the simple-averaged return for experienced return and missed return without any weighting. In model 4, we simply use and instead of and to measure past returns. Similar as model 2 and model 3, we get close estimations of and with and respectively. Meanwhile, the t-test cannot reject the null that. According to the four models, we conclude that, when deciding to participate in future IPOs, institutions take into account initial returns of the IPOs they participated in the past, regardless of the fact that they received a share allocation or not. This type of behavior is consistent with Bayesian learning. 16 In China, shares issued in primary market are heavily oversubscribed. As a result, the chance of getting shares is extremely low. If one institution does not get any share, we make its. In model 2, there are 144 institutions with. 17 In model 2, the mean and median of are 94% and 100%. In model 3, the mean and median of are 87% and 92%. In addition, we regress on % variance of are explained by if we keep institutions with. After excluding those unlucky institutions, the explanatory power of is still quite high with R-square of 91.2%.

13 4.2.2 Luckiness and future behavior With respect to hypothesis2, we wonder whether luckiness matters when institution make future investment decision. For testing this hypothesis, we rearrange the decomposition of as follows and use to measure the luckiness of institutions. Then, we run the following regression: According to hypothesis 2, if has significantly positive impact on it suggests luckiness affects institution s future decision. Furthermore, if, it implies institutional investors overweight their luck when deciding to participate in future IPOs. Regression results are exhibited in model 5 of 错误! 未找到引用源. We find that the luckiness term positively affect institution s decision at 1% significant level, which is consistent with hypothesis 2a. On the other hand, the coefficient of luckiness term is equal to which is much higher than the coefficient of missed return with For the equality test, the null hypothesis of is rejected. This result supports hypothesis 2b that institutions overweight their luckiness when they make future decisions. In model 6, again, we exclude institutions that do not get any shares in sub-period A and conduct the same test as model 5. The coefficient difference between and becomes wider comparing with model 5. Meanwhile, the equality test of is rejected more strongly at 5% significant level relative to the significant level of 10% in model 5. One may argue that the results could be driven by the division point of the two sub-periods. Therefore, we split our sample in another way by which we observe institution s experience in one year period. Based on the new division point, sub-period A covers 135 IPOs from Nov 2010 to Nov 2011 and sub-period B consist of 79 IPO from Dec 2011 until now. We implement the same six models as before and the results are displayed in Table 1 Panel B. The alternative results are quite similar to the results in Table 2 Panel A. This robustness test ceases the worry of division effect.

14 4.2.3 Monetary gain and future behavior In this test, we explore whether monetary gain generated from previous IPOs influences institution s future investment behavior. Firstly, we calculate the nominal monetary gain in the sub-period A 18 for each institution and split into actually experienced gain 19 and missed gain 20 to measure naïve reinforcement learning and Bayesian learning respectively. Similar as previous tests, we use to control inherent investment propensity and to measure the likelihood of participating in future IPOs. Table 3 exhibits the seven models that test the impact of monetary gain on future investment tendency. In model 1, we regress Log ( ) on Log ( ) and only. We find that favorable nominal monetary gain makes institution to participate IPOs more often in the future. In order to further distinguish Bayesian learning and naïve reinforcement learning, we used the decomposed terms and as explanatory variables in model 2. The outcome reveals that actual gain has significantly positive impact on future participating desire but missed gain doses not. In model 3, we exclude institutions without any allocation in sub-period A and the new results are consistent with model 2. Interestingly, we herein find only actual gain influence institution s future decision, but previous tests show that only the missed part of return rate has effect on it. To examine the influence of return rate and monetary gain simultaneously, are included in model 4 and model 5 as explanatory variables. Same as the independent tests before, we find the actual gain and missed return positively affect future participating frequency. In addition, both of the two variables turn to be more significant after excluding those unlucky institutions. The interesting results imply that both Bayesian learning and naïve reinforcement learning play roles in institution s decision making but from different aspects. 18 The division of sub-periods is same as test 1 with 107 IPOs in sub-period A and 107 IPOs in sub-period B. 19, where is adjusted initial return for IPO i; is the offer price for IPO i; is the number of shares obtained in IPO i. is the times of getting shares in sub-period A. 20, where is adjusted initial return for IPO i; is the offer price for IPO i; is the number of shares missed in IPO i. is the times of having missed shares in sub-period A.

15 In model 6 and model 7, we investigate the luckiness effect on future decision with controlling of monetary gain. For both models, the coefficient of is significantly higher than the coefficient of at 1% significant level. This result shows that luckiness still matters while actual monetary gain influences institution s future participating frequency. According to the results in Table 3, we conclude that institutions participate more often in future after experienced a large gain from the past IPOs, which sustains hypothesis 3. Therefore, we infer that the learning processes of institutions are subject to both Bayesian and naïve reinforcement learning, which support the hybrid learning model of Camerer and Ho (1999). Consistent with hypothesis 2, luckiness still plays role in decision process after considering monetary gain.

16 5. Conclusion This research explores the impact of personal experience and luck on the behavior of institutional investors in an IPO market. First of all, we find that some institutions have inherently high propensity of participating in IPOs. Secondly, we disclose that institutions take into account initial returns of the IPOs they participated in the past, regardless of the fact that they received a share allocation or not. This finding implies Bayesian learning plays role in institution s learning process. Thirdly, we find that institutions participate more often in future if they personally experienced a large gain from the past IPOs, which hints that naïve reinforcement learning has impact on institution s future decision as well. Hence, we conclude that institutions are subject to both Bayesian and naïve reinforcement learning but in different aspects. In addition, we find that institutions are more likely to participate in future IPOs if they were lucky with the share allocation in the past.

17 References ARTHUR, W. B Designing Economic Agents that Act like Human Agents: A Behavioral Approach to Bounded Rationality. The American Economic Review, 81, CAMERER, C. & HO, T.-H Experienced-Weighted Attraction Learning in Normal Form Games. Econometrica, 67, CHEUNG, Y.-W. & FRIEDMAN, D Individual Learning in Normal Form Games: Some Laboratory Results. Games and Economic Behavior, 19, CHIANG, Y.-M., HIRSHLEIFER, D., QIAN, Y. & SHERMAN, A. E Do Investors Learn from Experience? Evidence from Frequent IPO Investors. The Review of Financial Studies, 24, CHIANG, Y.-M., QIAN, Y. & SHERMAN, A. E Endogenous Entry and Partial Adjustment in IPO Auctions: Are Institutional Investors Better Informed? The Review of Financial Studies, 23, COHEN, R. B., GOMPERS, P. A. & VUOLTEENAHO, T Who underreacts to cash-flow news? evidence from trading between individuals and institutions. Journal of Financial Economics, 66, CRAWFORD, V. & BROSETA, B What Price Coordination? The Efficiency-Enhancing Effect of Auctioning the Right to Play. The American Economic Review, 88, CRAWFORD, V. P Adaptive Dynamics in Coordination Games. Econometrica, 63, CROSS, J. G A Stochastic Learning Model of Economic Behavior. The Quarterly Journal of Economics, 87, DHAR, R. & ZHU, N Up Close and Personal: Investor Sophistication and the Disposition Effect. Management Science, 52, DINSMOOR, J. A The etymology of basic concepts in the experimental analysis of behavior. Journal of the Experimental Analysis of Behavior, 82, FENG, L. & SEASHOLES, M. S Do Investor Sophistication and Trading Experience Eliminate Behavioral Biases in Financial Markets? Review of Finance, 9, KAUSTIA, M. & KN PFER, S Do Investors Overweight Personal Experience? Evidence from IPO Subscriptions. The Journal of Finance, 63,

18 LACH, S Decomposition of Variables and Correlated Measurement Errors. International Economic Review, 34, LICHTENBERG, F. R Aggregation of Variables in Least-Squares Regression. The American Statistician, 44, NAGEL, S Short sales, institutional investors and the cross-section of stock returns. Journal of Financial Economics, 78, ROTH, A. E. & EREV, I Learning in extensive-form games: Experimental data and simple dynamic models in the intermediate term. Games and Economic Behavior, 8, SERU, A., SHUMWAY, T. & STOFFMAN, N Learning by Trading. Review of Financial Studies, 23, SKINNER, B. F The behavior of organisms : an experimental analysis, New York : Appleton-Century-Crofts ; c1938. THORNDIKE, E Animal intelligence: Experimental studies., New York: Macmillian. ZEILER, M. D Fixed and variable schedules of response-independent reinforcement. Journal of the Experimental Analysis of Behavior, 11,

19 Table 3 Panel A: Descriptive Statistics for Bookbuilding # of IPOs 214 # of Bids # of Unique Institutions 353 Average # of Bids per IPO Average # of Institutions per IPO Average # of Bids Offered by Institutions in an IPO 1.79 Panel B: IPO Description N Mean Median SD Unadjusted Initial Return (%) Adjusted Initial Return (%) Number of Days between Issuing Day and Trading Day Firm Age Gross Proceeds ($ mil) P/E Ratio Sales One Year before IPO($ mil) Net Profit One Year before IPO($ mil)

20 Table 2 Panel A The effect of past returns on future decision Dependent variable: Log ( ) Independent variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Log ( ) 0.793*** (2.92) 0.975*** 0.970*** 1.088*** 0.943*** 0.981*** 1.063*** (25.28) (23.3) (14.37) (21.14) (25.48) (14.35) (0.67) (1.07) 0.784*** (2.88) 2.507*** (3.35) (1.44) 0.695*** (2.65) 0.812*** (2.96) 2.725*** (3.64) 4.354** (2.26) 6.894*** (3.13) Constant *** *** *** *** *** *** (-4.22) (-4.18) (-3.78) (-3.89) (-4.23) (-3.77) Obs R-sq 62.48% 62.49% 58.49% 62.61% 62.75% 59.42% Comparison of * **

21 Table 2 Panel B The effect of past returns on future decision (Alternative sub-period division) Dependent variable: Log ( ) Independent variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Log ( ) 0.703** (2.31) 0.807*** 0.797*** 0.866*** 0.772*** 0.812*** 0.832*** (21.13) (18.28) (10.45) (15.77) (21.3) (9.79) (0.9) (0.97) 0.689** (2.26) 1.826* (1.91) (1.25) 0.613** (2.08) 0.721** (2.34) 2.158** (2.2) 4.241** (2.01) 6.486** (2.34) Constant *** *** *** *** *** *** (-4.32) (-4.2) (-2.84) (-3.8) (-4.31) (-2.68) Obs R-sq 57.38% 57.44% 45.67% 57.53% 57.64% 46.5% Comparison of * 0.065*

22 Table 3 The effect of past monetary gain on future decision Dependent variable: Log ( ) Independent variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Log ( ) 0.009* (1.69) 0.925*** 0.905*** 1.010*** 0.915*** 0.938*** 0.915*** 0.989*** (18.23) (17.77) (11.54) (17.94) (9.85) (17.97) (10.91) 0.912** (2.58) 0.884** (2.42) 1.200** (2.58) 1.302*** (2.81) 0.750** (2.14) 0.553* (1.89) (-0.13) (-0.71) (-0.54) (-0.92) (0.34) (0.37) (-0.77) (-0.92) 0.807*** (2.9) 2.877*** (3.82) 0.797*** (2.89) 2.631*** (3.52) 3.944* (1.9) 6.536*** (2.71) Constant ** ** ** *** *** *** *** (-2.58) (-2.35) (-2.1) (-3.39) (-2.75) (-3.43) (-3.28) Obs R-sq 61.71% 62.16% 56.46% 63.28% 60.25% 63.35% 60.27% Comparison of 0.007*** 0.002***

23 Figure 1 Past nominal returns and frequency of participating in following IPOs Low Return High Return Low Return High Return Low Return High Return Low Participation Middle Participation High Participation Mean Median