Why Are Stock Exchange IPOs So Underpriced and Yet Outperform in The Long Run? A Test of the Signaling Hypothesis

Why Are Stock Exchange IPOs So Underpriced and Yet Outperform in The Long Run? A Test of the Signaling Hypothesis Abstract: Isaac Otchere Sprott School of Business Carleton University Ottawa, Canada [This version: November 2009] We investigate underpricing and long term stock market performance of stock exchange IPOs and find that they are more underpriced than regular IPOs. Despite being more underpriced, the financial exchanges significantly outperform the market indexes. Thus, contrary to the findings in prior studies, we find that stock exchange IPO underpricing is positively related to their long run stock returns. The lack of managerial ownership of shares in the stock exchange firms creates a situation where management is not constrained to significantly underprice their initial issues in order to signal their firm s prospects since they do not directly bear the cost of leaving too much money on the table. Interestingly, both underpricing and long run returns of stock exchange IPOs are positively related to proxies of the signaling hypothesis. On the basis of additional tests involving the stock exchange IPOs and a sub-sample of regular IPOs that also outperformed the market in the long run, we are able to rule out the quasi-monopoly argument as a possible explanation for the strong long run performance of stock exchange IPOs. Key Words: JEL Classification: Stock exchange, self-listed IPOs, underpricing, performance G15, G32 Email: iotchere@sprott.carleton.ca Tel: (613) 520-2600 ext 2731: Fax (613) 520-4427 1

Why Are Stock Exchange IPOs So Underpriced and Yet Outperform in The Long Run? A Test of the Signaling Hypothesis 1. Introduction Prior studies have documented that initial public offerings are underpriced (see Ritter and Welch, 2002) and many theories have been developed to explain why owners of firms sell their shares to the public for less than what they could have realized. The underpricing phenomenon has been attributed, among others to signaling hypothesis (Allen and Faulhaber, 1989; Welch, 1989; Grinblat and Hwang, 1989), winners curse hypothesis (Rock, 1986; Ko and Walter, 1989), underwriter s reputation hypothesis (Loughran and Ritter, 2002), and lawsuit avoidance hypothesis, among others. Another regularity that has been documented in the IPO literature is the negative aftermarket long run returns for IPOs (see Ritter (1991), Aggarwal and Rivoli (1990), Loughran and Ritter (1995), and Lewis (1993). Ritter (1995), for example, finds significantly negative abnormal returns of 29.1% after 36 months of trading. This literature also shows that underpriced IPOs underperform the market in the long run, suggesting that investors who are initially overly optimistic about the prospects of the firm become more realistic through time. The negative relationship between the level of underpricing and long run performance for IPOs has also been attributed to investor over optimism and ex-ante uncertainty (Shiller, 1990 and Miller, 1977), overoptimistic accounting (Teoh et al, 1998) and temporary overvaluation by investors (Aggarwal and Rivoli, 1990). Recent evidence on stock exchange IPOs suggests a positive relationship between underpricing and long run performance. In the past decade, there has been a wave of financial exchange conversion IPOs as a growing number of these exchanges have demutualized and corporatized their structure by going public and listing their shares, sometimes on their own stock exchanges, a phenomenon that Otchere (2006) has dubbed self-listing. Interestingly, unlike 2

regular IPOs, which on average underperform in the long run, stock exchange IPOs outperform in the long run both in terms of returns to shareholders and operating performance. Aggarwal and Dahiya (2006) and Otchere (2006) provide preliminary evidence that shows that despite a market adjusted underpricing of 23%, stock exchange IPOs outperform the market. Given the evidence in extant literature concerning the underpricing phenomenon and long term stock market performance of IPOs, we ask why stock exchange IPOs are significantly underpriced and yet outperform in the long run? Using a sample of financial exchange IPOs and a control sample of firms that also went public in the same period as the stock exchanges, we first document the extent of underpricing for the sample firms and find that the stock exchange IPOs, on average, are significantly underpriced by about 30%, while the regular IPOs are underpriced by 14%. Despite being significantly more underpriced, the stock exchange sample significantly outperforms the market index and the regular IPOs irrespective of the return-generation method employed. We find that the market-adjusted cumulative abnormal returns (CAR) for stock exchanges after 12, 36 and 60 months are as high as 34%, 63% and 90% respectively. In contrast, the market-adjusted cumulative average returns for the regular IPOs are -1.89%, -32% and -39% for one, three and five years respectively. Having established that stock exchange IPOs are significantly more underpriced and yet outperform the market and regular IPOs, we examine why the initial returns of financial exchange IPOs are positively related to their long term performance We find that the performance of the financial exchange IPOs is related to proxies of the signaling hypothesis. Specifically, underpricing of stock exchange IPOs is positively related to share retention by the owners, strong pre-ipo operating performance, use of reputable underwriters, demand, and 3

overallotment. Similarly, the long run stock market performance of the financial exchanges is also related to their pre-ipo operating performance, underpricing and the percentage of the firm s share retained by the owners. These results are consistent with the signaling theory proposed by Allen and Faulhaber (1989), Welch (1989) and Grinblat and Hwang (1989), i.e. good performing firms signal their type by underpricing their IPOs and the owners retaining a significant percentage of the shares. On why stock exchange IPOs are more underpriced than regular IPOs, we argue that unlike the managers/owners of regular IPO firms, the managers of financial exchanges do not usually own shares in the firm prior to the initial public offering, since the shares of financial exchanges are usually issued to members who own seats on the exchange. This lack of managerial ownership in the stock exchange firm may explain the significant underpricing. Since usually there are no shares to be sold by management in the IPO process, there is no incentive on their part to monitor underwriters to reduce the level of underpricing because they do not directly bear the cost of leaving money on the table. Consequently, if management wants to signal information about the future prospects and quality of the firm, they will not be constrained to significantly underprice. On the other hand, for regular IPOs, since the owners/managers (including venture capitalists) own a significant proportion of the pre-ipo shares and want to maximize the proceeds, they would be relatively more concerned about leaving too much money on the table; hence, they will monitor underwriters to reduce underpricing. Our results, which are robust to the measurement of underpricing and long run returns, are consistent with this assertion. The rest of the paper is organized as follows. In Section 2, we discuss the developments in the financial exchange industry that have induced the conversion of financial exchanges to 4

publicly traded entities. A description of the data and methodology is presented in Section 3. Section 4 presents our preliminary results concerning underpricing and long term performance of the sample. In Section 5, we perform cross sectional analysis of the determinants of stock exchange IPOs underpricing and long term performance and then link the results to the existing theories. A summary and conclusion appear in section 6. 2. Trends in Stock exchange IPOs Stock exchanges have traditionally been organized as mutual entities. Under this governance structure, the exchange is owned by members; therefore, ownership and the right to consume the services of the exchange are bundled (Hart and Moore, 1996; Di Noia, 1998). It is the members, and therefore the customers, who have the residual rights of control over the exchange. The mutual form has prevailed for centuries because exchanges were able to maintain their monopoly status. Hitherto, the state of technological development restricted communication across physical distances and curbed any real competition among exchanges. Consequently, until more recently, the mutual structure had been the favored governance structure. The operating environment for financial exchanges has changed markedly in recent years as shocks to the industry have reshaped both how exchanges compete and with whom they compete. Recent shocks that have affected the industry include globalization, technological developments, and the attendant growth in electronic communication networks (ECNs) all of which have heightened the level of competition for order flow among exchanges. The development of sophisticated trading capabilities has literally changed the face of stock exchange industry, with electronic trading platforms replacing the traditional, physical trading floors. In the developed world, virtually all exchanges have replaced trading floors with electronic trading systems. This is not surprising given the well-documented improvements in efficiency associated with electronic trading system 5

(see Bessembinder & Kaufman, 1997 and Domowitz & Steil, 1999). Advancements in communication, information, and network technology have also facilitated the break-down of barriers to entry and have further increased competition among stock exchanges and electronic communication networks (ECNs). ECNs offer investors expanded trading opportunities and undermine the traditional monopolistic stronghold of stock exchanges by offering speedy and cost-efficient services made possible by technologically-advanced electronic trading platforms. In the current competitive environment, exchanges are forced to operate with much more efficiency and flexibility, but the user-owned nature of the mutual form constrains management s ability to respond to the needs of competition (Domowitz and Steil, 1999; Hart and Moore, 1996). In fact, empirical evidence (e.g. Krishnamurti, et al., 2003, and Otchere, 2006) suggests that the mutual ownership structure does not optimize operational performance of the exchange. The developments in the stock exchange industry have forced exchanges to change their governance structure by demutualizing in order to better adapt to changing operating conditions. According to the World Federation of Exchanges, in 1998, 38% of WFE members operated as demutualized entities. However, by 2006, 75% had demutualized. Demutualization involves replacing mutual ownership of seats on the exchange with shares. The exchange becomes a for-profit (commercial) entity, a significant departure from the not-for-profit mandate of the mutual form. However, a demutualized exchange still remains a private entity, with a clear separation between the trading rights (now owned by the exchange) and ownership rights. Increasingly however, exchanges have taken a further step by corporatizing their structure, going public and most of the time self-listing their shares on the publicly traded exchanges. By corporatizing their structure, owner and user interests are completely decoupled, thereby bestowing on managers the freedom to pursue business strategies 6

and opportunities that increase the firm s competitiveness and maximize value for shareholders. According to the World Federation of Exchange (WFE), as of December 2008, 25 stock and derivative exchanges had gone public. The changing trend in the exchange industry is better visualized in Figure 1. In the early 1990s, almost all the WFE members were mutual exchanges. Since then, the number of mutual exchanges has fallen, while the number of profit-oriented exchanges has increased markedly. Also, the number of stock exchange going public has increased from 2 in 1998 to 25 as of 2008. The major stock exchanges, including NYSE, the London Stock Exchange, Hong Kong Stock Exchange, Toronto Stock Exchanges, Australian Stock Exchange, and the NASDAQ have all undertaken initial public offerings. 1 The few studies that analyze the performance of these publicly listed exchanges (e.g., Aggarwal and Dahiya (2006), Otchere (2006), and Serisfoy (2007)) document significant improvements in the financial and operating performance of these exchanges. Since the conversion of financial exchanges to publicly traded entities is a relatively new phenomenon, research into the extent of stock exchange IPO underpricing is scarce. Aggarwal and Dahiya (2006) provide some preliminary empirical findings on this issue. They found that the median underpricing of 17.3%. Otchere (2006) also shows that stock exchange IPOs outperform in the long run. Therefore, the question that emanates from the IPO literature and the recent evidence is why are the stock exchange IPOs underpriced and yet outperform in the long run? Do they perform better in the long run because they are underpriced? Is the underpricing and strong long run performance a reflection of signaling? In this study, we provide answers to 1 The NYSE s was a backdoor listing through its merger with the then publicly-traded Archipelago Holdings. 7

these interesting questions by first documenting the extent of underpricing and long run performance of regular IPOs and a control sample of non exchange IPOs, and then proceed to analyze why they outperform in the long run while the ordinary IPOs underperform. 3. Data and Methodology 3.1 Data We obtain the list of stock exchanges IPOs from Otchere (2006) and supplemental data from the Securities Data Company (SDC) new issues database. For each stock exchange IPO, a sizematched control sample of non-stock exchange financial and non-financial services firms that went public in the same country and during the same period of time (6 months) is created. The matched firms had market capitalization in the range of 90%-110% of that of the listed exchange at the time of the IPO. The number of matched firms per stock exchanges ranges from a minimum of 10 to a maximum of 20. This procedure generated a control sample of 339 firms of which 41% consists of financial services firms. Stock price data were obtained from Datastream, while the IPO transaction data including information on underwriters, auditors, fees, and pre and post-ipo financial data were obtained from SDC database. 3.2 Methodology 3.2.1 Estimation of the level of underpricing The degree of underpricing, defined as the difference between the offer price and the first closing price, is estimated as:, 100%..(1) where IR i is the initial return of firm i, P i,t is the closing price of firm i on the first trading day on the secondary market and E i is the offering price of firm i. The above unadjusted initial returns 8

do not take into account returns on alternative investments. We therefore estimate marketadjusted underpricing as follows:,,, 100% where M t is the value of the index on the first trading day of the IPO and M t,0 is the value of the major market index in the respective countries at the offering date of the IPO. An alternative estimation of the level of underpricing is based on log returns is presented as follow: ln, ln, 100% Habib and Ljungqvist (2001) argue that underpricing is of concern to the issuing party to the extent that it impacts directly on their net wealth position. Some entrepreneurs choose to retain a large proportion of the capital in the firm whereas others see the IPO as an opportunity to cashout their investment in the firm. The owners are likely to be more concerned about the level of underpricing. Consequently, an alternative measure of underpricing, adjusted for the percentage of shares retained by the owners (RetOwn i ), is also estimated as follow: 1,,, 3.2.2 Long-run abnormal returns We measure the long-term performance of IPOs using the market-adjusted returns method where the returns of the major index of the respective country are used as benchmark returns. The standard event study method is employed to estimate the long run abnormal returns as:, where ar i,t is the average market-adjusted abnormal returns for the sample. As an alternative to the cumulative market-adjusted returns, we also compute buy and hold returns (BHAR) for the 9

sample firms. The market-adjusted buy and hold returns of stock i for period T is calculated as follow:, 1, 1, where R i,t is the return of stock i at time t; R m,t is the return of the respective stock index and (T) is the time period for which the buy-and-hold returns are calculated. The mean buy and hold return is estimated as the average returns for sample. The significance of the CAR and BHAR is tested using the cross sectional standard deviation of the sample returns. 4. Preliminary results 4.1 Descriptive Statistics Summary statistics for the sample firms are presented in Table 1. On average, stock exchanges are larger, raise more capital, and have relatively less debt than regular IPOs. While the profitability ratios of the financial exchanges are higher, they are not statistically different from those of the matched firms. Financial exchange IPO firms also employ the services of more reputable underwriters than regular IPOs. As reflected in the high overallotment amount, the demand for the shares of financial exchange firms is higher than that of regular IPOs. In addition, the owners of financial exchanges retain more of their firms shares than those of regular IPO firms. This perhaps reflects a higher level of confidence that the owners or former members have in the exchange s future. These statistics provide preliminary evidence about the quality of the firm as these are good proxies of signaling. 4.2. Underpricing of stock exchange IPOs The level of underpricing of stock exchanges IPOs and the regular IPOs is presented in Table 2. The average underpricing calculated using the initial return method for the stock exchange IPOs 10

is 34.31% (t-statistic= 3.53) while that of the regular IPOs is 13.45% (t-statistic = 6.09). Financial services firms are underpriced by 9.38% (t-statistic: 4.61) compared to 19.01% (tstatistic: 4.33) for non-financial firms. Similar results are obtained using the market-adjusted and log return methods, albeit the magnitude of the underpricing based on the log return and retention-adjusted methods is much smaller. Generally, the level of underpricing documented for the stock exchanges IPOs is higher than that of the regular IPOs. Only 3 of the 20 publicly traded exchanges realized negative first day returns and many of the stock exchange IPOs enjoyed first day returns in excess of 40%, with the Osaka stock exchange realizing an initial return of over 150%. On average, 70% of regular IPOs realized positive initial returns. An interesting finding is that the high initial returns associated with early financial exchange IPOs have made the more recent listings very attractive to investors. As exhibited in figure 2, the magnitude of the initial returns associated with stock exchange IPOs has increased significantly in the latter half of the study period but there is no significant increase in the level of underpricing for regular IPOs. We find that irrespective of the method used to estimate underpricing, stock exchanges IPOs are significantly underpriced than their regular counterparts (by 20% on average). The level of underpricing of ordinary IPOs is quite consistent with the evidence documented in previous research (e.g., Ritter, 1991; Ritter and Welch, 2002; Hahn and Ligon, 2006, and Garg al el, 2008). [Fix Table 2 here] 4.3 Long-run performance of IPOs In Table 3, we report the long run stock market performance of the sample firms. Panel A shows the market-adjusted returns, Panel B exhibits the BHAR returns and Panel C shows the unadjusted returns. The control sample exhibits a gradual decline in performance over the 60-11

month period following the first month of trading. Based on the Market-adjusted method, we observe that the one-year, two-year, three-year, four-year and five-year market-adjusted CARs of the matched firms of 2.01%, -9.25%, -15.78%, -26.37% and -36.39%, respectively are statistically significant, except the one year returns. The matched financial firms underperformed but to a lesser extent as compared to non-financial matching firms. The long run abnormal returns documented for the control sample is similar to the evidence reported in Ritter (1991). For the stock exchange IPOs however, we observe that their CARs increase at a steady rate over the five-year period. Using the market-adjusted approach, the CAR of 34.33%, 62.04%, 63.43%, 68.61% and 90.3% for the first, second, third, fourth and fifth year respectively are statistically and economically significant. Similar results are obtained for the stock exchange IPOs and control samples using the BHAR returns. In the five years following the initial public offering, the stock exchange sample realized average returns of 119.32% (t-statistic=7.94) while nonexchange IPO firms realized a return of -16.85% (t-statistic=3.03). A $1 invested in the financial exchange IPOs instead of ordinary IPOs would have generated an extra return of 135%. 4.4 Robustness test Barber and Lyon (1997) and other researchers show that long term stock return studies based on BHAR and CAR are subject to new listing bias, as during the period of analysis newly listed firms become part of the market index against which the sample firms performance is measured. By virtue of their sheer relative sizes (see Table 1), stock exchanges are likely to be included in the index sooner and are likely to significantly impact on the value of the index. Though in most cases, newly listed firms are not included in the benchmark index immediately after going public (as exchange listing rules usually require that candidates for inclusion in the index be listed on the stock exchange for at least 3 months before being added to the index) it is reasonable to 12

expect that over the five year period following the public listing, publicly traded financial exchange firms would be included in the index. Therefore, using the benchmark indexes to compute abnormal return could result in biased estimates. [Fix Table 3 here] A possible solution to this potential problem is to remove the returns of the sample firms that have been included in the index from the benchmark index returns. However, to be able to purge the index returns of the returns of the included firms, one need to know the constituents of the indexes and the dates the sample firms were included. It was not possible to confirm the constituents of the market indexes used in the study or whether and when any of the sample firms were included. Therefore, removing the returns of the sample firms from the returns of the market index would be difficult. To ensure that our results are not affected by this index inclusion bias, we estimate the unadjusted abnormal returns for the sample firms and report the results in Panel C of Table 3. The unadjusted returns for both the financial exchange IPOs and the regular IPOs are similar to what we obtained using the market-adjusted and BHAR methods. The financial exchanges realize significantly positively returns whereas the regular IPOs realized significantly negative abnormal return during the study period. The main conclusion that we draw from the foregoing discussion is that despite being significantly more underpriced, the stock exchange IPOs significantly outperform the market index and the regular IPOs irrespective of the return-generation method employed. The long run returns graphically presented in Figure 2, show a start difference in the long run performance of the sample firms [Fix Figure 2 here] 13

5. Why are stock exchange IPOs more underpriced than regular IPOs? Having documented that stock exchange IPOs are significantly underpriced and yet perform better than the market while the ordinary IPOs significantly underperform, in this section we examine why stock exchange IPOs are significantly underpriced than ordinary IPOs. To explore this question, we estimate the following model which includes factors that extant literature suggests can influence underpricing. UP i = α + β 1 Underwriter s Ranking + β 2 Fee + β 3 EPS -1 + β 4 Total Assets + β 5 Time Lag + β 6 Demand + β 7 Proceed + β 8 Retention + β 9 Selling Shareholders + β 10 Pricing Technique + β 11 Offering Techniques + ε i where UP i is percentage underpricing of firm I, underwriter s ranking represents the lead underwriter s reputation. To measure an underwriter s reputation, we use Rank, the tombstone underwriter s prestige ranking variable developed by Carter and Manaster (1990) and updated by Carter, Dark, and Singh (1998) and Loughran and Ritter (2004). The rankings range from one to nine, with a higher ranking implying a more prestigious investment bank. The more prestigious the underwriter is, the higher the level of confidence and optimism investors will have in the issuing company. The higher level of optimism could lead to a higher level of underpricing. In other words, good quality firms can signal their quality by employing a prestigious underwriter. Fees are the total amount incurred by the issuing firm as a percentage of total proceeds. EPS -1 is the earnings per share one year prior to the year of issuance. This variable measures the earnings potential of the firm and can also be used as a measure of the level of optimism that investors have for the issue. Total Assets is included as a proxy for size as larger firms may exude more confidence in the investors. Alternatively, a large firm could wield more power in negotiations with the underwriter and may be able to better influence the determination of the issue price. 14

Time Lag is the difference between the issue date and the settlement date. This variable is used to measure the degree of anticipation for the IPO. High anticipation should lead to high excitement, thus higher levels of underpricing. Demand is the amount of overallotment shares sold as a percentage of total proceeds. The greater the overallotments, the higher the level of demand for the shares and therefore, the higher the underpricing. Proceeds is also be used to measure the level of excitement that an issue creates. A larger issue would generate more excitement in the market as compared to a smaller issue. Consequently, a higher level of excitement can lead to a higher level of underpricing. Retention, defined as the percentage of shares retained by existing shareholders is used as a proxy for signaling. A high level of retention will suggest that the owners are confident about the future prospects of the firm. This will send a positive signal to the market and increase the demand for the issue, thus resulting in a higher level of underpricing. Extant literature also suggests that an issuer s ownership structure (i.e., the composition of shareholders and distribution of shares among them) prior to and after the IPO affects the level of underpricing (Ligon, 2005). We include Selling Shareholders to capture institutional sales. If the selling shareholders include institutional investors, then this variable coded as 1, and zero otherwise. Prior studies also show that the Offering techniques and Pricing techniques used by the underwriter could also affect the underpricing of IPOs; hence, we include these factors as control variables. 5.1.2 Results The results of the regression using different measures of underpricing are reported in Table 4. Panel A shows the results for the financial exchange IPOs, while Panel B presents the results for the regular IPOs. For the stock exchange IPOs, we observe that Demand, measured by the percentage overallotment, is a significant determinant of underpricing. This result is consistent 15

with the winner s curse hypothesis (Rock 1986) in which underpricing exists because underpriced IPOs are rationed to more investors than overpriced IPOs. Therefore, investors bid a higher price in the aftermarket trading to get the quantity they applied for during the subscription period. The coefficient for Demand of 1.38, 1.31 and 1.35 in the regression using the Initial return, Market-adjusted return and Log returns respectively as dependent variables are significant at 10%. The coefficient of Underwriter s ranking is significant at the 1% level and suggests that using reputable intermediaries is associated with higher levels of underpricing. This result is consistent with the findings of Dolvin (2005) and Hoberg (2007) on the impact of underwriter s reputation. Strong historical performance, as measured by EPS in the year prior to the offering, is positively related to underpricing. This result is consistent with the signaling hypothesis presented by Allen and Faulhaber (1989), Welch (1989) and Grinblat and Hwang (1989), i.e. strong performing firms signal their quality and future prospects by underpricing their IPOs. Similarly, the coefficient of Retention is positive and significant and it signals the owners/managers level of confidence in the firms future performance. The level of underpricing is lower for large firms as indicated by the coefficient of Total Asset, perhaps because larger firms wield more power in the negotiation process with the underwriters and thus are able to reduce the level of underpricing. This result is also consistent with the assertion that larger firms are associated with less ex ante uncertainty. The coefficient of Proceeds is also negative and significant at the 5% level for both the Initial return and Market-adjusted methods. Large capital raisings involve less uncertainty and hence require less of a pricing discount to compensate uninformed investors. Other factors such as Fee, Time Lag, Selling shareholders, Pricing Techniques, Offering Techniques have little power in explaining the level of underpricing of 16

stock exchanges IPOs. The adjusted R 2 of 69%, 71% and 72% for the Initial return, Marketadjusted return and Log return models respectively suggests that the models explain a large portion of the variability in the level of underpricing of stock exchange IPOs. Turning to the results for the matched firms, we observe that while some of the determinants of underpricing are similar, there are a few notable differences. Demand, Proceeds and underwriter rankings are significantly related to underpricing of regular IPOs. However, EPS in the year prior to the issue and Retention (proxies of the signaling hypothesis) do not significantly influence the level of underpricing of regular IPO firms. In summary, there are differences in the determinants of the level of underpricing of stock exchange IPOs and regular IPOs. The level of underpricing of ordinary IPO firms is mainly driven by Demand, Proceeds, Underwriter s ranking and Pricing Technique while underpricing of stock exchange IPOs is largely driven by Demand, Underwriter s ranking, stronger pre-ipo performance and Retention. That strong historical performance and share retention significantly influence underpricing of stock exchanges is indicative of the branding and prestige elements that stock exchanges enjoy and provide preliminary evidence consistent with the signaling hypothesis posits that good quality firms signal to the public information about their good future prospects by underpricing their IPOs, and they do so by employing the services of reputable underwriters. [Fix Table 4 here] 5.2 Why do stock exchange IPOs perform better than regular IPOs in the long run? The results reported in Table 3 clearly show that stock exchange IPOs have better long term performance than ordinary IPOs. To explore the question of why the financial exchange IPO 17

sample outperforms the market and ordinary IPOs in spite of being significantly underpriced, we estimate the following model: LP i,t = α + β 1 Underwriter s ranking + β 2 Fees + β 3 EPS -1 + β 4 Revenue growth + β 5 D/E + β 6 ROA + β 7 Demand + β 8 Proceeds + β 9 Retention + β 10 Selling Shareholders + β 11 UP i + β 12 change in Total Assets + ε i where LP i,t is the five-year market-adjusted cumulative abnormal returns of firm i and the independent variables are essentially those used in the underpricing regression that also affect long term performance. We include underpricing as an additional independent variable because of Levis (1993) argument that underpricing might be due to initial over optimism in the market, which is subsequently reversed in the after-market. The D/E ratio is also included as an independent variable since a healthier D/E ratio means better financial strength and, therefore, better long term performance. We include Revenue growth and change in Total Assets (investments) as additional measures of operating performance. The inclusion of these operating performance measures and ROA helps determine whether the abnormal returns are related to efficient use of assets. The results of the regression are presented in Table 4. Panel A shows the results for the financial exchanges, while Panel B presents the results for the regular IPOs. The superior performance of the stock exchange sample can be attributed to good prior operating performance. Consistent with the signaling argument, EPS in the year prior to the issue is positively related to long term performance, with the coefficient of 0.018 and 0.027 for Market-adjusted CAR and BHAR respectively being significant at 5%. ROA is also positively related to the long run abnormal returns. Interestingly the coefficient of Retention is positively related to the long term performance, suggesting that the higher the retention ratio, the better the long run abnormal returns realized by the stock exchange IPOs. The coefficient of change in Total Assets (investment) is also positively related to long term performance, suggesting that 18

management invested in positive NPV projects. The adjusted R-squared is about 68%, thus the model explains a significant portion of the variability in the long run stock returns of the stock exchange IPOs. [Fix Table 5 here] The determinants of the long term stock market performance of ordinary IPOs are different. From Panel B of Table 4, we observe that first, the direct cost of issue, measured by fees charged by the underwriter, is negatively related to the long-term performance of ordinary IPOs. The coefficient of -1.79 and -1.43 for the market-adjusted CAR and BHAR methods are significant at 5% and 10% respectively, suggesting that the higher the amount of fees incurred from the offering, the worse the long term performance. Second, operating performance as proxied by ROA at the time of the issue is negatively related to the long-term stock returns. This finding is consistent with the hypothesis that managers attempt to window-dress their financial accounts prior to going public and this leads to the pre-ipo performance being over-stated (Aharony et al., 2000). The negative coefficient estimate for underpricing also indicates that investors are initially overoptimistic about the firm s performance. As this over optimism dies down, the stock price drops in the long run. The results are also consistent with the findings of Levis (1993) and Paudyal, Saadouni, and Briston (1998). Comparing the results of the stock exchange IPOs and ordinary IPOs, once again, we observe striking differences in the determinants of the long term stock market performance of the sample. For regular IPOs, the evidence suggests that managers tend to window-dress their accounts prior to going public. On the other hand, the long run returns of financial exchanges are more related to proxies of signaling, namely, strong historical performance, share retentions at the time of the IPO, and underpricing. 19

In summary, we find that contrary to the findings in prior studies and unlike our results for the regular IPOs that show that underperformance is negatively related to long term stock returns, underpricing of the stock exchange IPOs is positively related to their long run stock returns, albeit the coefficient is not significant. In addition, we find that share ownership retention in financial exchanges following the IPO is positively related to their long term performance. This implies that the higher the owners level of confidence about the firm s future, the more likely it is that the firm performs better in the long run. The results presented in Tables 3 and 4 show that underpricing and strong long term stock market performance of stock exchange IPOs are attributed to the signaling hypothesis. In the subsequent sections, we provide further test of the signaling hypothesis. 5.3 Stock exchange IPO underpricing, long run stock market returns and existing theories. In this section, we relate our results to existing theories with the view to ascertaining which of the hypotheses is consistent with our results. Before doing so, a summary of our key findings is in order. As documented above, both the financial exchange IPOs and regular IPOs are underpriced, but the financial exchanges are significantly more underpriced than regular IPOs. Interestingly also, unlike regular IPOs, which on average underperform in the long run, the financial exchange IPOs generate substantial excess returns in the long run. Underpricing of stock exchange IPOs is strongly positively related to share retention, pre-ipo operating performance, the use of reputable underwriters, demand, and overallotment. Similarly, the long run stock market performance of the financial exchanges is also related to pre-ipo operating performance, underpricing and the percentage of the firm s shares retained by the owners. While the regular IPOs are also underpriced, albeit to a lesser extent, they underperform in the long run. 20

Extant literature offers a number of explanations for the underpricing and long-run underperformance of IPOs. For underpricing, some of the notable explanations are the winner s curse hypothesis, signaling theory, book building theory and costly information acquisition hypothesis. The winner s curse hypothesis (Rock, 1986) suggests that some investors are more informed about the value of the IPO firm than others. The informed investors only subscribe to issues whose offer prices are lower than the expected market prices. As a result, the uninformed investors stand a higher chance of being allocated overpriced issues. The uninformed will not participate in the IPO market unless the underwriter underprices the issue. For financial exchange IPOs however, the shares are usually issued to member/owners who are institutional investors and are presumable informed investors. Therefore, we rule out this hypothesis as a possible explanation for the significant underpricing of the stock exchange IPOs. The costly information acquisition hypothesis posits that underpricing compensates informed investors for their information production. Since the shares are sold to members who own seats on the exchange, the information production hypothesis does not apply to stock exchange IPOs. The signaling theory is based on the assumption that the firm knows about its prospects better than the investors. Allen and Faulhaber (1989) propose that good firms find it optimal to signal their type by underpricing their initial issue of shares. This hypothesis is further formalized by Welch (1989) in a model where high quality firms underprice, but low quality firms are not able to do so because of high imitation costs. Grinblat and Hwang (1989) also assert that high quality issuers signal their quality by underpricing as well as retaining a significant proportion of their firms shares. Our results primarily support the signaling hypothesis. For example, the significance of retention ratio in the regression with underpricing as 21

the dependent variable is suggestive of the higher level of confidence in the firm s future by the owners or former members. The strong historical performance lends credence to this assertion. On the long run performance of IPOs, the underperformance has been attributable to the divergence of opinion hypothesis, the windows of opportunity hypothesis, and the market optimism hypothesis. Ritter (1991) for example, asserts that the long run underperformance of regular IPOs is consistent with an IPO market in which investors are periodically overoptimistic about the earnings potential of young growth companies, and firms take advantage of these "windows of opportunity to go public. These hypotheses posit that the degree of underpricing is negatively related to the post-issue price, suggesting that initial public offerings will underperform in the long run. While these hypotheses hold for our regular IPO sample, we rule them out as possible explanations for the long run stock market performance of the financial exchange IPOs on account of the strongly positive abnormal returns. Aggarwal and Rivoli (1990) also offer an explanation that links the long-term performance to the initial returns of IPOs. They postulate that the aftermarket is not immediately efficient in valuing newly issued securities and that the abnormal returns earned by IPO investors are the result of temporary overvaluation by investors in early trading. This view is consistent with the fads hypothesis (Shiller, 1990; DeBondt and Thaler, 1985) which predicts that the greater the initial returns at the IPO date, the greater the degree of subsequent correction of overpricing by investors, and the lower the long run returns. We rule out this hypothesis as a possible explanation of our results because of the strong positive cumulative abnormal returns realized by the stock exchange sample in spite of the underpricing at the time of the IPO. 22

Schultz (2001) also argues that when the community of entrepreneurs sees a group of IPOs become successful, they issue more IPOs to follow those successful ones. As a result, the last group of IPOs which would form a relatively large fraction of the sample would underperform. Schultz contends that the high-volume periods issues would, on average, carry a larger weight resulting in underperformance. To test this hypothesis, we split both our stock exchange sample and regular IPO sample into two, early and later issuers, and examine the differences in the level of underpricing and long term performance. The results, presented in Table 6 and figure 3 do not support this hypothesis as far as the stock exchange sample is concerned. For regular IPOs, we find confirmatory evidence that later issuers underperform the market in the long run, albeit there is no perceptible difference in their level of underpricing. For stock exchange IPOs however, the success of the earlier issues generated significant interest in these IPOs which resulted in significant increases in underpricing of later issues (see figure 3). While the long run stock market performance of later issues is less than that of earlier issuers, the returns are nonetheless significantly positive. Therefore, we rule out the subsequent high volume hypothesis as a possible explanation for the performance of the stock exchange IPOs. [Fix Table 6 here] Other studies attempt to find variables that result in cross-sectional predictability of the long run underperformance. Teoh, Welch, and Wong (1998) attribute the poor post-ipo stock performance to over optimistic accounting early in the life of the firm. They assert that firms are eager to look good when they conduct their IPOs, and that the market has difficulties in disentangling carefully hidden warning signals. They suggest that at least part of the poor longrun performance is due to unduly optimism of IPOs and the firm s inability to properly forecast tougher times ahead. We rule out this explanation because of the strong post-ipo performance of 23

the stock exchange sample. The positive relationship among retention, underpricing and long run performance suggests that signaling is a probable explanation. In the following sections, we provide additional test of this hypothesis. 5.4 Financial exchange IPO underpricing and long term performance: Further test of the Signaling Hypothesis Rooted in asymmetric information argument, the signaling hypothesis posits that issuers know more about the firms prospects than outsiders. High value firms signal their quality by deliberately selling their shares at a discount and the cost of signaling deters lower quality issuers from imitating (Allen and Faulhaber (1989)). Grinblatt and Hwang (1989) also argue that issuers signal higher quality in IPOs by retaining a significant proportion of the firms shares. We argue that not only will high quality firms signal their quality by underpricing, they will also use reputable underwriters in the process. Based on the foregoing discussion, we identify proxies of signaling hypothesis to include high post-issue ownership of shares (retention), strong historical operating performance, significant underpricing and the use of firm commitment method. Owners of good quality firms with these characteristics would signal their quality by retaining a significant percentage of the firms shares and underpricing their issues, and they will do so by employing the services of top investment bankers. Consistent with the signaling hypothesis, we find that the financial exchange IPO firms employ the services of more reputable underwriters than regular IPOs. There is a higher demand for the shares of financial exchange IPOs than regular IPOs as reflected in the higher overallotment amount. More importantly, the financial exchanges have stronger historical operating performance, and the owners retain more of the firms shares than the regular IPOs. This evidence is consistent with the signaling hypothesis. 24

Moreover, if indeed underpricing and retention are used to signal quality and good prospects, then these variables would be significantly related to the long term performance of the firms. That is precisely what the stock exchange IPO sample s long term stock market performance indicates- the managers know that the underlying business model is attractive and they signal that quality by underpricing. We find that the variables representing firm s quality such as EPS and ROA, together with retention and underpricing are positively related to the long run stock market performance of the stock exchanges. 5.5 Additional test: Signaling vs. Monopoly explanation Some analysts have argued that the exchanges enjoy quasi-monopoly status, with competitors facing significant barriers to entry and that this unique advantage can explain the strong stock market performance of the stock exchange IPOs (Aggarwal and Dahiya, 2006). The conjecture that the strong stock market performance is due to the exchanges being quasi monopolies may not hold. In this era of globalization, competition is no longer localized. Globalization and technological developments have increased competition and significantly affected the hitherto monopoly rents that the exchanges earned. The increasing internationalization of financial markets has set national exchanges in direct competition with each other. Technological advancement has facilitated the trading of shares on several stock exchanges. The migration of order flow to other markets has affected the local franchise and monopoly that the exchanges had in their respective countries. In addition, the emergence of alternative trading mechanisms, such as electronic communication networks (ECNs), has increased competition to an unprecedented level for the financial exchange industry. All these developments have eroded the true monopoly status of financial exchanges. Hence, what was once a captive market is no longer the sole 25

jurisdiction of the local exchange (Otchere, 2006). Nonetheless, in most countries apart from the US, there is usually only one national stock and/or derivative exchange which may create the appearance of a monopoly; therefore, in this section we provide a direct test of the monopoly and signaling hypotheses. Looking at the regular IPOs sample, we observe substantial variations in the long run performance, as 43% of the regular IPOs realized positive abnormal returns, whereas about 80% of the financial exchanges realized significantly positive abnormal returns. To examine whether financial exchanges performance is due to their quasi-monopoly status, we determine whether the stock exchange sample and the sub-sample of regular IPOs that also realized positive long run cumulative abnormal returns share some common characteristic. This categorization is influenced by Brav s and Gompers (1997) view that in analyzing the performance IPOs, we should look more broadly at the types of firms that underperform and not consider IPO firms as a distinct group. We conjecture that if the signaling hypothesis explains the long run performance of the stock exchange IPOs, then we would expect the regular IPOs that outperformed the market in the long run to share some common characteristics with the stock exchanges IPOs that may be related to proxies of the signaling hypothesis. Such evidence will be inconsistent with the monopoly hypothesis because the sub-sample of regular IPOs that realized significantly positive abnormal returns in the long run are not quasi monopolies. To test the conjecture, first, we estimate the level of underpricing for the stock exchange sample and the regular IPOs that realized positive long run abnormal returns and find that the regular IPOs that realized positive abnormal returns in the long run were underpriced by 4.9% while the financial exchange IPOs were underpriced by 34% (the difference is significant at 5%). We attribute the difference in underpricing to incentives resulting from different management 26