Long-run performance of Initial Public Offerings in The Netherlands and Belgium

(1)

Long-run performance of Initial Public

Offerings in The Netherlands and

Belgium

University of Groningen

Faculty of Economics and Business

MSc Business Administration – Specialization Finance

Final Version, April 2009

Jos Meijer

De Laan 2B9

9712AV Groningen

S1335650@student.rug.nl

Student number: 1335650

(2)

I

Abstract

This study examines the abnormal three-year return of 104 Dutch and Belgian initial public offerings (IPOs) between 1997 and 2005. Based on the book-to-market and market capitalization share portfolios of Fama and French (1993), IPOs are compared to matching benchmark portfolios. Although literature suggests that a long-run difference in returns does not exist, the IPOs outperform their benchmarks by a 15.55% and a cumulative 20.11% after three years. However, these returns are not statistically significant. The first year abnormal returns are higher than the abnormal returns in the second and third year, with 33.36%, -3.45% and -9.79%, respectively. Belgian IPOs perform better than their Dutch counterparts, especially in the second and third year. The Dutch IPO sample has an abnormal return of -0.01% over 3 years. Although the difference in return is not significant, the variance of the returns of Belgian IPOs is significantly higher than the variance of Dutch IPO returns (at the 1% level). IPOs in the telecommunications, media and technology (TMT) sector yield a higher three-year abnormal buy-and-hold return than non-technology companies. The difference is 76.52%. Furthermore, during ‘hot-issue’ years, IPOs return more than their benchmarks in comparison to issues in ‘cold-issue’ periods.

JEL Classification: G14, G15

(3)

II

III Introduction

The share return of an Initial Public Offering (IPO) is something that has been able to capture the imagination of researchers and investors alike. During the tech boom in the late 1990s many growth shares made their way to the exchanges and share prices soared. In the heydays of this boom, some of the shares even saw their values doubling or tripling in a few months time. Investors were mesmerized by new tech introductions, assigning incredible value potential to the ‘dotcom’ business model. Despite the first optimism and believe in new valuation methods1, it quickly became clear that many technology-related shares were priced excessively. As a consequence, share owners lost a large part of their capital almost overnight.

This study addresses the abnormal long-run returns of IPOs. The IPOs of Amsterdam and Brussels are included in this research, in the period from 1997 through 2005. During this period we have seen the rise and fall of many technology-related IPOs, for instance World Online and KPNQwest in The Netherlands. But what about the survivors, what is their aftermarket return compared to their peers companies? By using Fama and French-style (1993) portfolios, this thesis will touch upon these issues.

In the first chapter, the literature will be explored, and a review of past empirical studies will be presented. The second chapter addresses the research methodology. Then, the third chapter covers the data. In the fourth chapter, there will be a brief comparison to earlier studies and there will be touched upon the expectations of the results. The fifth chapter presents the results, followed by a chapter with the conclusions, summary and discussion.

I would like to express my thanks to dr. von Eije for both supervising the research and providing me with useful insights.

1_{C. Shapiro and H. Varian describe a new way of valuation through valuing ‘network effects’ in their Information}

(5)

1. Background literature on long-run returns of IPOs

1.1 Introduction

In the first section the reasons for underperformance and the results of earlier studies will be presented. As Jay Ritter (1991) suggests, historically, research on IPOs mainly focuses on the ‘underpricing effect’. This is the performance during the first day of a share’s listing and it is also regarded in the literature as the short-run performance.2

The research on long-run performance of IPOs emerged later than studies on the underpricing effect, but it has intensified over the past decade. In the 1970s, many studies focusing on the underpricing effect take the IPO returns after 1 month and 12 months into account. Research in this era focusing on longer periods of for instance 3 to 5 years is scarce. Post-IPO research mainly focused on the (middle term) survival of newly introduced companies (Dawson, 1987).3

Jay Ritter was among the first to extensively research a second anomaly of IPOs, namely the long-run returns of IPOs. Ritter (1991) identifies that the topic is interesting for a number of reasons. For instance, investors seeking positive abnormal returns may benefit from the results. In addition, when non-zero long-run returns are found, the information efficiency of the IPO markets can be disputed. Furthermore, the ‘timing-effect’ could be important. When IPOs yield negative long-run share return during periods of high share trading volumes (or booms), companies and investors can benefit from that knowledge. Finally, Ritter states that the costs of equity do not only depend on the transaction costs of going public, but also on the share returns that are earned in the aftermarket.

2_{Various authors, most notably Ibbotson (1975) and Ritter (1984a, 1991), find that “the underpricing of IPOs is a}

generalized phenomenon” (Álvarez and González, 2005: p.326).

3_{As becomes clear from research from the U.S. Securities and Exchanges Commission (SEC, 1963), 29% of the}

(6)

1.2 Reasons for underperformance

Research shows multiple reasons for lower returns on IPOs. Shiller (1990) conducts a study on IPOs during the hot-issue market of 1987.4 He compares underwriters with impresarios of the music industry, through the following analogy: “it is often better to create an excess demand for the tickets, creating scenes of people standing in long lines for tickets, or trading among themselves at higher prices. This impression will tend to produce greater demand for subsequent events” (Shiller, 1990: p.62). Shiller claims that hot-issue markets emerge when the salesmen discover that they can create ‘fads’ for IPOs, just as impresarios tend to do in the entertainment industry. This theory of fads, or overestimation theory, is further examined by Ritter (1991). He observes that high volume public offerings during a hot-issue market or boom are especially prone to long-run underperformance. The hot-issue market theory has been researched extensively. For instance, Aggarwal and Rivoli (1990) find that the underperformance of IPOs is probably a consequence of overvaluation or fad in the early trading days of securities. This initial rise in share price results in a long-run underperformance, this implies that initial underpricing influences the long-run returns of shares. Another reason of underperformance is the systematic underpricing of shares by underwriters. This phenomenon is found by various researchers, and shows resemblance with Shiller’s (1990) theory. According to Álvarez and González (2005), many companies go public in the near-peak of industry fads. Lee, Shleifer and Thaler (1991: p.106) find that “the investor sentiment hypothesis suggests that these IPO's should be more prevalent in times when individual investors are optimistic, so the stocks will fetch high prices relative to their fundamental values.” Ritter (1991) gives two other possible explanations for the long-run underperformance, namely bad luck and mismeasurement of performance. The latter could become apparent when alternative benchmark portfolios are used5 Teoh, Welch and Wong (1998) show that underperformance can be caused through the practice of window dressing in the period prior to

4_{‘Hot-issue’ markets are characterized by high volumes of new IPOs and high initial returns. Hot markets are}

periods with economic growth and favorable conditions on the stock exchange. In contrast, during cold-issue periods the volume and the initial returns for IPOs are low. For instance, 1980 and the 1996-2000 period are hot-issue market periods and 2000-2001 is a cold-hot-issue period (Ibbotson and Jaffe (1975), Ritter (1984) and Buttimer, Hyland and Sanders (2005)).

5

(7)

the IPO. In a study on the relation between the return of IPO-shares and the reputation of underwriters, Carter, Dark and Singh (1998) conclude that when IPOs are lead by prestigious underwriters the share performance is better than when less prestigious underwriters lead the IPO process.

Kunz and Aggarwal (1994) conclude in a study on Swiss IPOs, that firms should receive a higher issue price “by taking advantage of the competition among the investment banks” (p.722), and reducing risks by providing more transparency about their financial statements, owner structure and the organization. In this way, the lower long-run abnormal returns could be a less common phenomenon.

Von Eije, de Witte and van der Zwaan (2004) show that an IPO can bring changes to organizations, which manifest themselves mainly within the financial management of the company. Long-run performance is higher for companies that show signs of such IPO-related change, than for companies that do not show any signs of IPO-related change.

1.3 Results from the literature in the U.S.

Ibbotson (1975) was one of the first to report on IPO aftermarket performance. He analyses abnormal returns of IPOs between 1960 and 1969. Ibbotson concludes that “results generally confirm that there are no departures from market efficiency in the aftermarket” (p.265). Ibbotson finds positive performance in the first year, negative returns in the second through fourth year of about 1% month, and positive performance in the last, fifth year. Ibbotson measures the long-run performance over the 5 years on a month-to-month basis.

A study by Reilly (1977) on the one-year returns of 486 U.S. IPOs between 1972 and 1975 finds that the average abnormal return of the sample equals -11.60%. The NASDAQ-index is taken as a benchmark for this study and the results are statistically significant at the 5% level.

(8)

for underperformance is that many firms go public during the peak of industry fads, and investors are considered to be overoptimistic.

Loughran (1993) takes into account the data of a prior study by Reinganum (1990)6 on differences in return between NASDAQ and NYSE securities. Loughran finds that the six-year average abnormal return of 3,656 NASDAQ IPOs is as much as -33.30% (not significant). In this research, the results are corrected for size and companies are compared with the NASDAQ.

Loughran and Ritter (1995) conduct a study of 4,753 IPOs in the 1970-90 period. While the three-year average abnormal return is -26.90%, after 5 years this underperformance increases to as much as -50.70%, both returns being statistically insignificant. In this research, IPO shares are benchmarked by same-size companies and not by industry-peers. The authors made this choice because there are little same industry companies with matching size.

Carter, Dark and Singh (1998) study the three-year returns of 2,292 U.S. IPOs between 1979 and 1991. As a benchmark, the weighted market returns of the NYSE, AMEX and NASDAQ are used. Carter et al. (1998) find a long-run abnormal return of -19.92%, being statistically significant at the 5% level.

Yi (2001) studies the abnormal returns of 1,031 IPOs on the U.S. NASDAQ between 1987 and 1991. Yi makes a distinction between companies that have losses and those who make a profit. The profitable companies have a three-year abnormal return of -0.18% and unprofitable companies have a three-year abnormal return of -32.49% (both results significant at the 1% level). This study includes 184 profitable companies and 29 companies with losses. Yi recognizes that the results differ significantly with that of Ritters’ study on NASDAQ IPOs.

Then, there are some studies covering a shorter event window. Aggarwal and Rivoli (1990) find in their study of 1,598 NASDAQ IPOs in the 1977-87 period, that the average underperformance of new issues is as much as -13.73% (significant at the 5% level). They

6_{Reinganum (1990) finds that small securities listed on the New York Stock Exchange outperform similar}

(9)

measure the performance from the end of the first day until the 250th trading day. In their conclusions they indicate that this significant underperformance opens the doors to various short-selling tactics. The results of Peavy III (1990) on 41 IPOs in 1986 and 1987 are in line with the outcomes of the prior studies, although the aftermarket period is 100 days. This study shows a -13.49% abnormal return, and is statistically not significant.

Two studies conclude a positive abnormal performance. The first, by Buser and Chan (1987), shows an abnormal aftermarket two-year return of 11.20% when NASDAQ IPOs are compared to the NASDAQ (significant at the 5% level). The study comprises 1,078 IPOs between 1981 and 1985. Ritter’s (1991) comment on these positive results is that the NASDAQ underperforms other indexes in this particular period.

The second study, by Chalk and Peavy III (1987) concludes that IPOs have a significant positive abnormal aftermarket return of 17.99% over the first 190 days. This study includes 649 companies that went public in the 1975-82 period, and the result is significant at the 5% level. The benchmark of the IPOs is the NASDAQ-index, which could prove to be not representative as a benchmark. Also, the authors conclude that IPO shares with a starting quote of less than USD 1.00 yield a large abnormal aftermarket return.7

1.4 Results from the literature outside the U.S.

Most studies in the U.S. indicate a negative abnormal long-run return of IPOs. More evidence comes from Europe. Levis (1993) reviews 632 IPOs in the U.K. between 1980 and 1988. He finds that after the initial underpricing in the first day, the underperformance starts beyond 3 years of listing. The negative abnormal return compared to the HGSC smallcap index is -8.10%, which is larger than the difference with the large capitalization FTA-index with -4.20%. According to Levis, during 1984-1987 there were signs of the small-cap effect in the U.K.

7_{Chalk and Peavy III (1987) address several reasons for the anomalous aftermarket returns among low priced}

(10)

Van Hoeijen and van der Sar (1999) examine the long-run performance of IPOs after 3 and 5 years. They found a 17.10% outperformance of IPOs after 3 years, but a 17.90% underperformance after 5 years. The results are not statistically significant.

Doeswijk, Hemmes and Venekamp (2006) study the three-year performance of 154 new issues in the Netherlands between 1977 and 2001. They find that the IPOs on average underperform their benchmark with -10.00%, but this result is statistically not significant. 4 main sectors were identified as a benchmark, namely: cyclical shares, defensive shares, high growth shares and interest rate sensitive shares. Furthermore, the authors take a closer look at the hot-issue periods of 1986/87 and 1997 through mid-2000. Especially the latter period is of importance for this research. Doeswijk et al. (2006) recognize that most IPOs in this period were so-called ‘high growth shares’, which include “information- related industries (content and processing), such as telecom, media and information technology” (p.412). The average three-year abnormal performance for high growth IPOs between 1997 and 2000 is -38.40% (significant at the 1% level). Doeswijk et al. (2006) qualify 36 out of 55 companies as high growth shares in this sample. For the total sample of 55 IPOs, the three-year performance is -18.20%, although this result is statistically insignificant.

Kunz and Aggarwal (1994) study the performance of Swiss IPOs. They find that the 34 shares issued in the 1983 to 1989 period yield an average abnormal return of -6.10%. Another study of Swiss IPOs is conducted by Drobetz, Kammermann and Wälchli (2005), and it contains 87 IPOs in the period between 1983 and 2000. They compare the three-year returns of IPOs to both a broad index and an index of comparable size companies. The abnormal return compared to the broad index is -7.45% and to the similar size index is -1.02% (not significant). Drobetz et al. (2005) contribute this to the ‘small-cap’ effect, as most of the IPOs are small capitalizations, their benchmark should contain small companies as well.

(11)

poor performance is mainly apparent in the 1988-1990 period. The results are statistically significant at the 1% level.

Another study on German IPOs by Stehle, Ehrhardt and Przyborowsky (2000) reviews data used by Stehle and Ehrhardt (1999). In this earlier study they indicate that a sample of 187 German IPOs in the 1960-92 period yields a three-year abnormal return of -5.04%. In the study by Stehle et al. (2000) different benchmarks are used, namely a value-weighted portfolio of shares in contrary to the equally weighted portfolio, as was used in the 1999 research.8 It becomes clear that when similar size companies in an equally-weighted benchmark are used, the -5.04% abnormal return becomes a 1.54% abnormal return, although both results are not statistically significant.

The fourth study by Sapusek (2000) measures the three-year returns of 246 German IPOs between 1983 and 1993. The IPOs yield a -20.03% abnormal return compared to a benchmark of several matching firms (significant at the 1% level). When the German DAX-index is applied as a benchmark, it yields a mean-excess return of 1.66% higher than the index (not statistically significant). Finally, when the DAFOX-Small-Cap-Performance-Index is used as benchmark, the IPOs show a -25.81% abnormal return (significant at the 1% level). This result emphasizes that it is crucial to seek for an appropriate benchmark, keeping in mind the small-cap effect.

Álvarez and González (2005) consider 37 Spanish IPOs in the 1987-97 period. The three-year average abnormal return, compared to a set of comparable companies, is -28.24% (statistically significant at the 10% level). Álvarez and González (2005) also find that initial one-day underpricing is positively related to the five-year underperformance of IPOs in the Spanish market. According to their study; “firms choose to undervalue with the aim of selling more

8

In equal-weighting (EW), all the new issues receive the same weight, or importance. In value-weighting, the issuer’s common stock market value at the beginning of the holding period is taken into account for the benchmark. This, as a consequence, rules out the small-cap effect to a large extent

(12)

stocks in seasoned offerings, at a higher price than they could obtain in the absence of the signal” (p.346).9

A study of Portuguese IPOs (Borges, 2007) finds that between 1988 and 2004 the average abnormal three-year return of 41 new issues yields 8.74%. It should be noticed that in the period on average only 2.5 companies were introduced per annum. Further, in the 1990s there was the phenomenon of a speculative bubble, which manifested especially under ICT-shares. This effect has not been taken into account by Borges.

For markets outside the U.S. and Europe Dawson (1987) studies the one-year returns of issues in Hong Kong, Malaysia and Singapore between 1978 and 1984. In Hong Kong, Dawson finds a negative abnormal return of -9.30% (21 observations), for Singapore -2.70% (39 observations) and Malaysia 18.20% (21 observations). None of the results are significantly different from zero.

McGuinness (1993) finds in Hong Kong that 72 IPOs in the 1980-90 period give a 500-day negative abnormal return of -18.30%. This result is significant at the 5% level.

Aggarwal, Leal and Hernandez (1993) study the long-run returns for Brazil, Chile and Mexico. In Brazil, the study comprises 48 new issues between 1980 and 1990. The average abnormal return of the IPOs is -47.00% after 3 years. For Chile, a sample of 18 new issues between 1982 and 1990 yields an average abnormal return of -23.70%. Finally in Mexico, there are only one-year returns in the 1987-90 period. The 38 new issues give a market-adjusted return of -19.60%. It must be understood that this research is held in a period with great financial turmoil, e.g. the October ’87 crash. This could disturb the results. In addition, Aggarwal, Leal and Hernandez (1993) indicate that the research is influenced by the small samples and the heavy inflation. In

9_{According to the authors, the performance of IPOs is positively related to the initial underpricing of the IPO.}

(13)

Brazil for instance, the average annual inflation during the sample period varied between 83% in 1980 and a staggering 2,938% in 1990.

The first of 2 studies of the Australian stock exchange is conducted by Finn and Higham (1988). They analyze the one-year return of 93 issues between 1966 and 1978, and find aftermarket abnormal performance of average -6.52% (not significant). The second study focusing on Australian issues, by Lee, Taylor and Walter (1996), examines the three-year aftermarket return of 169 IPOs between 1976 and 1989. They find that the long-run abnormal return of an equally weighted portfolio of these IPOs is -51.26%.

Studies of IPOs in Tunisia (Naceur, 2000) and Egypt (Omran, 2005) contain relatively little observations. The study in Tunisia (1992-97), covering 12 IPOs, shows that after 1 year, there is a positive abnormal return of 11.04% (significant at the 5% level). For Egypt, Omran (2005) reviews 53 IPOs of privatized companies in the 1994-98 period, concluding that the average performance of IPOs was -25.00% and -27.00% lower than the market, after 3 to 5 years respectively. These results are statistically significant at the 1% level.

1.5 Alternative benchmarks

(14)

sentiment, it is these firms” (p.1819). Thirdly, the authors state that small firms are more likely to be prone to information asymmetry than large firms. Investors in small companies mainly constitute of individuals and not large institutional investors, and in contradiction to large institutional investors, individual investors spend less time tracking the returns of companies and analyzing its performance. Fourthly, individuals “might derive utility from buying the shares of small, low book-to-market firms because they value them like a lottery ticket” (p.1819).

Another study by Brav, Geczy and Gompers (2000), including 4,622 U.S. IPOs between 1975 and 1992, reviews the five-year abnormal returns. The abnormal returns of the IPOs are tested against different benchmarks like NASDAQ, S&P-500 and size and book-to-market portfolios. Similarly to the Brav and Gompers (1997) study they find that correcting for size eliminates the abnormal negative return of IPOs. The IPOs are distributed among 25 portfolios that represent different 5 book-to-market quintiles and 5 size quintiles. It turns out that 51.80% of all IPOs fit in the smallest portfolio (1,1), and just 0.6% fit in the portfolio with the highest book-to-market and size (5,5). This approach is based on the on the work of Fama and French (1993). After correcting for size, the abnormal performance turns from 25.70% (S&P 500) or -15.60% (NASDAQ) to 1.40% (size and book-to-market benchmark).

Eckbo and Norli (2005) address the five-year return of 6,139 NASDAQ IPOs (1973-2002), in a research on risk-return characteristics of issuers. Similar to Brav et al. (2000) they take into account the risk factors liquidity10 and leverage11 in their benchmarks. Their main hypothesis is: “(…) that IPO stocks have lower expected return due to lower exposures to these and other risk factors” (p.2). It turns out that on average most IPO shares are significantly less leveraged but more liquid (turnover) than their counterparts. This may lower the systematic risk of IPOs. In addition, Eckbo and Norli (2005) recognize that low aftermarket “(…) returns may be a manifestation of the more general finding of Fama and French (1992) that small growth stocks

10_{Eckbo and Norli (2000) notice that a number of studies suggest that greater stock liquidity reduces risk. See for}

instance: Brennan and Subrahmanyam (1996), Brennan, Chordia and Subrahmanyam (1998), Datar, Naik and Radcliffe. (1998), Chordia, Subrahmanyam and Anshuman (2001) and Eckbo and Norli (2002).

11

(15)

tend to exhibit low returns during the post-1963 period” (p.2). The abnormal long-run return for IPOs after size matching is 28.90% when an equally weighted benchmark is applied, and -18.80% when a value-weighted benchmark is used. When the control firms are also matched on book-to-market, the performance of IPOs is -2.40% (equally weighted) and 0.30%, being statistically insignificant.

1.6 Conclusions from the literature

(16)

Table 1

Long-Run Abnormal Returns of Initial Public Offerings (IPOs)

This table shows the empirical results of the studies on long-run abnormal returns of IPOs. The abnormal return is measured by subtracting the return of a benchmark from the raw returns of an IPO. The results from developed countries are first summarized in alphabetical order. This is followed by a summary of lesser developed countries. Long-run returns are exclusive of initial (first day) returns. ***,**,* connote statistically significant at the 1%, 5% and 10% level, respectively.

Sample Abnormal Return Country Author(s), year Period Return (%) Size Period Australia Finn and Higham (1988) 1966-78 -6.52 93 1 year

Lee et al. (1996) 1976-89 -51.26*** 169 3 years Germany Ljungqvist (1997) 1970-93 -12.1*** 177 3 years Sapusek (2000) 1983-93 -20.3*** 246 3 years Stehle and Ehrhardt (1999) 1960-92 -5.04 187 3 years Stehle et al. (2000) 1960-92 1.54 187 3 years Great Britain Levis (1993) 1980-88 -30.59* 632 3 years Hong Kong Dawson (1987) 1978-83 -9.30 21 1 year

McGuinness (1993) 1980-90 -18.3** 72 500 days The Netherlands Hoeijen and Van der Sar (1999) 1980-1996 17.1 81 3 years

Doeswijk et al. (2006) 1977-2001 -10.0 154 3 years Portugal Borges (2007) 1988-2004 8.74 41 3 years Singapore Dawson (1987) 1978-83 -2.70 39 1 year Spain Álvarez and González(2005) 1987-97 -28.24* 37 3 years Switzerland Drobetz et al. (2005) 1983-2000 -1.02 87 3 years Kunz and Aggarwal (1994) 1983-89 -6.1 34 3 years United States Aggarwal and Rivoli (1990) 1977-87 -13.73** 1,598 250 days

(17)

2. Analysis framework and methodology

2.1 Hypothesis

Based on the literature study and its empirical results, this research will address the following hypothesis:

Hypothesis 1: Initial Public Offerings between 1997 and 2005 indicate no significant underperformance to a benchmark consisting of similar size and market-to-book companies.

The hypothesis tests whether the IPOs yield long-run abnormal return in comparison to the benchmark of listed shares on the Amsterdam and Brussels exchanges. The abnormal returns of the IPOs are measured by collecting the returns of the IPOs from the end of the first month of listing until exactly 3 years later. This return will be compared with the benchmark, by using Fama and French-style (1993) portfolios. More on this is covered in the methodology section. 2.2 Methodology

In this study, the returns of the IPOs will be compared to the returns of the size and market-to-book corrected benchmark portfolios. To create a proper benchmark, I have constructed Fama and French-style (1993) portfolios, which take the size and market-to-book characteristics of individual shares into account. The benchmark consists of shares listed on the Amsterdam and Brussels exchanges. Size is expressed as the market value of common equity, measured by multiplying the number of shares outstanding by its closing price. The market-to-book ratio is measured as the common equity that is reported on the balance sheet divided by the market value of the common equity.

(18)

The benchmark consists of 16 quartiles. Originally, Fama and French (1993), Brav and Gompers (1997), Brav, Gezcy and Gompers (2000) and Eckbo and Norli (2002) design their portfolios by using 25 quintiles (5 x 5). However, the IPO sample sizes used in these studies range from 934 to 4,622 companies. In the current study, comprising 2 smaller indexes, a total of 104 IPOs are examined (appendix table 1 and 2), and compared to 179 benchmark companies. This relative small population is too little for 25 portfolios. Other studies that differ from Fama and French’s quartiles includeÁlvarez and González (2005), using terciles in stead of quintiles in their research containing only 52 IPOs.

To give shape to the benchmark portfolios, size quartiles are formed by sorting all Dutch and Belgian shares in 4 even groups. Using the information of January 1st, size is measured as the total market value of the outstanding shares. Within each of the 4 different size groups, the market-to-book ratios of all Dutch and Belgian benchmark shares are sorted in 4 sub-groups, hence arriving at 16 unique portfolios. From 1997 up to and including 2007, the portfolios are annually rebalanced. This rebalancing is necessary because the ratios and market capitalizations do not remain stable for 3 consecutive years. Companies that were taken over or went bankrupt are removed from the benchmark sample. With 16 unique portfolios per year, stretching over 11 years, 176 unique portfolios are formed. This information is summarized in table 1 of the appendix. Each and every IPO can now be assigned to a portfolio with corresponding size and market-to-book characteristics. By doing this, smaller IPOs with a low market-to-book ratio will be compared to companies with similar size and market-to-book characteristics.

(19)

For instance, for the April 2000 IPO of the company SNT, benchmark portfolios 2000, 2001 and 2002 are included in measuring the returns of the benchmark. Thus in every consecutive year, this IPO is matched to a new portfolio: 2000: 3,4; 2001: 3,2; 2002: 3,1. Figure 1 in the appendix illustrates this process. By annually reconstructing the portfolios and reassigning the IPO to its corresponding portfolio in that respective year, the relative return is always measured to a proper benchmark. To make sure that IPOs are not compared with themselves, all IPOs are excluded from the benchmarks.

Takeovers or liquidations result in strong price movements and a delisting of the share. When a delisting occurs, the benchmark company is removed from the portfolio when the portfolio is rebalanced at the 1st of January. Eckbo and Norli (2002) indicate that the frequency of liquidations under IPOs and benchmarks are very similar. In addition, the event of a take-over or merger was pretty similar for both categories. Henceforth, Eckbo and Norli conclude that a delisting does not have a significant influence on the abnormal returns of IPOs.

2.3 Three-year buy-and-hold abnormal returns

This study examines the long-run return of the sample of Dutch and Belgian IPOs. Following the methodology of Barber and Lyon (1997) and others, the long-run return of the IPOs is expressed as the and-hold abnormal return (BHAR). To come to the BHAR, first the buy-and-hold return of the IPO is calculated as follows:

.

1 )

1 (

1

−













+

=

∏

= T t it i T

R

BHR

_. ₍₁₎ it

R stands for the return of IPO i in year t. T denotes the period during which the BHR is calculated. The same approach is used for the benchmark portfolio:

(20)

mt

R stands for the return of the benchmark portfolio m in year t. Then, following the methodology of Eckbo and Norli (2005), the buy-and-hold-abnormal return of each IPO is simply calculated as follows:

m T i T

BHR

BHARipo

=

−

(3)

The same calculation counts for the entire sample:

m T i T

BHR

ipos

BHAR

=

−

(4)

2.4 Cumulative abnormal returns

Among others, Ritter (1991) uses the alternative cumulative abnormal return (CAR) to measure the return of an IPO relative to its benchmark. To make the results of this study comparable to earlier studies, the cumulative abnormal return (CAR) of share i from year q to event year s is:

∑

=

s q t

BHARipo

s

CARq,

₍₅₎

3. Data

3.1 Sources of data

(21)

For the IPO-sample, daily market-to-book, size and return data have been acquired from DATASTREAM. For the benchmark-sample, only size and market-to-book information at January 1st is necessary.

3.2 Data description

The sample includes every Dutch and Belgian IPO between January 1997 and July 2005. In this period, 79 Dutch companies, and 92 companies from Belgium went public.12 For 28 Dutch companies and 17 Belgian companies, market-to-book information is not accessible. Furthermore, the Dutch IPO-sample contains 8 real estate companies, who are removed as well, and 2 firms (Vopak and Univar) that were actually a product of the split-up of a listed company (Vopak). For Belgium, 13 real estate companies are omitted from the population, leaving 41 suitable Dutch and 63 Belgian companies. To avoid survivorship bias13, companies that went bankrupt, taken over, or merged are included in the sample, but removed the year thereafter. Tables 2 and 3 of the appendix give an overview of the total IPO sample.

Of the 41 Dutch IPOs, the TNT Post Group shows the highest market capitalization with EUR 10,565 bln. after 1 month of trading. The smallest company is the Koninklijke Brill, with a market capitalization of EUR 16 mln. after 1 month of trading. In Belgium, the largest company after 1 month of the IPO is brewer Inbev (or Interbrew back then) with a market value of EUR 16 bln. The smallest company in Belgium is Sucraf with EUR 2 mln. in market capitalization. The average market value of the Dutch IPOs after 1 month is EUR 622 mln. and EUR 695 mln. for Belgium.

Figure 1 gives a view on the stagnating IPO-market. As from 2001 the unfavorable economic conditions virtually put the IPO-market to a halt. This effect was stronger in The Netherlands than in Belgium.

12

Information from DATASTREAM.

13_{According to Drobetz, Kammermann and Wälchli (2005) one must include delisted companies. When removing}

(22)

Figure 1

Number of Initial Public Offerings per Year

Figure 1 includes annual initial public offerings (IPOs) in The Netherlands and Belgium. The sample is collected from DATASTREAM for the 1997-2005 period. For inclusion in the sample, the IPO must have published a market-to-book ratio. Real estate IPOs are left out of the sample. The total number of remaining IPOs is 63 for Belgium and 41 for The Netherlands.

0 5 10 15 20 25 30 1997 1998 1999 2000 2001 2002 2003 2004 2005 Year IPOs Belgium The Netherlands

4. Comparison to earlier studies and expectations

What has to be expected from the results? In this section I present 4 characteristics of this study on IPOs that might influence the extent in which the study may be comparable to earlier works. This includes the industry characteristics, the market-to-book value, the company size and the number of IPOs included in the research.

(23)

of these companies went bankrupt.14 If the rules of Drobetz et al. (2005), Eckbo and Norli (2005) and Stehle et al. (2000) apply, which is that takeovers positively compensate the loss caused by companies going into default, the benchmark should be able to rule out large problems with the benchmark as a comparison measure. However, as can be seen in table 1, only 4 studies cover the entire period of the internet bubble (1997-2000). Only 1 of them, the study of Borges (2007), includes the returns of the IPOs in 2000 in the research and problems with generalizing the outcomes of this study may loom.

The second point of evaluation is the market-to-book value of the typical IPO in contrast to the benchmark. In the current study, 62.50% of all IPOs are located in the highest market-to-book quartile. This percentage is comparable to the numbers found in the studies of Brav and Gompers (1997), Brav, Gezcy and Gompers (2000), Álvarez and González (2005) and Eckbo and Norli (2005). According to Brav et al. (2000), the reason that many IPOs are clustered in the high market-tot-book quartiles is that IPOs on average represent better future growth options than benchmark shares. With the 62.50% of all IPOs located in the highest market-to-book quartile, it seems that the sample is comparable to other studies.

A third point is the market size of the typical IPO in contrast to the average size of the benchmarks. In their study on NASDAQ IPOs, Eckbo and Norli (2005) recognize that the typical IPO is not small relative to the benchmark. On the other hand, the typical size of the IPOs in the studies of Brav and Gompers (1997), Brav, Gezcy and Gompers (2000), Stehle et al. (2000) and Álvarez and González (2005) is rather small. In these studies, small IPOs represent the majority of the sample. Table 1 of the appendix shows the average capitalizations of the IPOs between 1997 and 2007. It can be found that the typical IPO is much smaller than the typical benchmark company. The smallest difference can be found in 2004, where the benchmark companies are on average EUR 1,555 mln. larger than the average IPO. The largest difference is recorded in 1998, where the gap is EUR 3,391 mln. It can be concluded that the size of the IPOs in this study is rather typical.

14_{Van Mulligen, Freddy, 2006, Nieuwkomers worden achterblijvers,}

(24)

Finally, the number of IPOs included differs considerably from U.S. studies on IPO returns. As can be seen in table 1, the average study on U.S. IPOs holds 2,295 companies. For non-U.S. research the average is 112 companies, compared to 104 IPOs for The Netherlands and Belgium combined. Some of the studies comprising non-US IPOs also stretch their period of research over more years than the current study.

(25)

Figure 2

The Distribution of Initial Public Offerings per Industry

Figure 2 depicts the distribution of IPOs per industry for the 2 countries. The industries are based on the ICB-classifications15 retrieved from EURONEXT.COM, and contain 9 major industries. N = 104.

0 5 10 15 20 25 30 35 Bas ic M ater ials Indu stria ls Con sum er G oods Hea lthC are Con sum er S ervi ces Tele com mun icat ions Util ities Fina ncia ls Tech nolo gy IPOs

5. Results

5.1 Three-year abnormal IPO returns

Table 2 presents the annual buy-and-hold abnormal returns (BHAR) for the sample. The sample includes 104 initial public offerings in The Netherlands and Belgium between 1997 and 2005. The returns are measured after the first, second and third year after going public, and the cumulative abnormal returns (CAR) over the total 3 years are displayed. The three-year BHAR is 15.55%, with an insignificant t-test. The Kruskal-Wallis test of medians however, gives significance at the 1% level. The annual BHARs are 33.36% (1 year), 3.45% (2 year) and -9.79 (3 years). The mean BHAR of the first year is significant at the 5% level, the second year mean BHAR is not significantly different from zero. The mean BHAR of the third year is

15_{The Industry Classification Benchmark (ICB) is a Dow Jones and FTSE classification system which offers a}

(26)

significant at the 10% level. The median first year return is not significant. The second- and the third-year results are significant at the 1% level. The CAR after three years is 20.11%, but this result is not significant. The cumulative performance after two years is 29.91%. This CAR is significant at the 10% level.

Table 2

Annual Average Buy-and-hold Abnormal Returns of IPOs

This table holds the annual average buy-and-hold abnormal returns (BHARs) and the cumulative abnormal returns (CARs). The column “year” denotes the year in which the IPO return is measured. Column “N” contains the number of IPOs included in the sample. The column marked “IPOs” holds the average 1-year buy-and-hold return (BHR) over each one of the 3 years, and the three-year buy-and-hold return. The next row “Benchmark” marks the annual and total return of the corresponding benchmark portfolio. The annual BHR is an average of the benchmark portfolio (1/N). The column marked “difference” computes the difference in return between the IPOs and the benchmark portfolios. Thus, “difference” stands for the buy-and-hold abnormal returns (BHARs) of the IPOs. In the next column the results of the student’s T-test of means, with the p-values in parentheses, are summarized. The results from the Kruskal-Wallis median test of one-way analysis of variance by ranks follow in the last column, with the p-values in parentheses. In the bottom part of this table, the cumulative abnormal returns (CARs) are presented. “1 CAR” stands for the cumulative abnormal return after 1 year. “1,2 CAR” is the cumulative abnormal return after 2 years, and “1,3 CAR” is the cumulative abnormal return for three years.

Tests for comparison of results Year N IPOs Benchmark Difference Mean t-test Median

Kruskal-Wallis 1 BHAR 104 48.43 15.07 33.36 2.1910 (0.0296) 0.1179 (0.7314) 2 BHAR 104 0.95 4.40 -3.45 -0.4646 (0.6427) 10.390 (0.0013) 3 BHAR 104 -8.05 1.74 -9.79 -1.6619 (0.0980) 9.3901 (0.0022) 1,3 BHARs 104 37.77 22.22 15.55 1.5174 (0.1307) 8.7791 (0.0030) 1 CAR 104 48.43 15.07 33.36 2.1910 (0.0296) 0.1179 (0.7314) 1,2 CAR 104 49.38 19.47 29.91 1.7013 (0.0904) 0.6355 (0.4253) 1,3 CAR 104 41.32 21.21 20.11 1.0453 (0.2971) 2.2844 (0.1307)

(27)

3 years of 17.10% between 1980 and 1996 for a Dutch sample. However, it is important to emphasize that the periods included in these earlier studies are different than in this study.16

The hypothesis that ‘Initial Public Offerings between 1997 and 2005 do not show long-run underperformance to the benchmark, consisting of similar size and same market-to-book companies’ can not be rejected. Although the sample shows a 15.55% BHAR after 3 years, the results appear to be statistically insignificant.

The CAR after 3 years is 20.11%. This is a large difference with for instance the -29.13% found by Ritter after 3 years. Again, on the basis of the CAR the hypothesis can not be rejected, because the results are not significant. The CAR after 2 years and 1 month after the public offering is 29.91% and significant at the 10% level. Ritter (1991) finds a -16.89% CAR after 24 months for his sample. This again is a large difference with the findings in this paper.

5.2 Industry characteristics

Doeswijk, Hemmes and Venekamp (2006) explicitly test the influence of the growth-shares within their sample. Of the total sample of 154 IPOs, 64 are classified as growth-shares, of which 36 are issued between 1997 and 2000. This category includes information-related industries as telecom, media and information technologies (TMT). The abnormal three-year return of the growth-shares between 1997 and 2000 is -38.40% (significant at the 5% level). Non-growth companies yield 20.10% in the same period (not statistically significant). Thus, growth IPOs have a negative return, according to the Doeswijk et al. study. What can the effect of such shares be on this research? In the current study I have segregated the three-year returns of the growth shares following ICB-classifications.17 The growth shares of the Dutch and Belgian sample can be found in the telecommunication and the technology sector. The results are visible in table 3. This table reveals that the telecommunication sector has a -85.11% three-year buy-and-hold abnormal return (BHAR). This result is significant at the 5% level for the

16_{As mentioned earlier, 1997-2000 is widely recognized as a ‘hot-issue’ period. After 2000, there is a ‘cold-issue’}

period. See footnote 4 for more information on this matter.

(28)

mean test, and the 10% level for the median test. The technology sector, including 29 companies, gives a BHAR of 117.09% after 3 years (not significant).

The three-year BHARs of the 2 industries combined (marked as TMT) is 92.58%. For non-technology companies, the three-year BHAR is 16.06%. The results of the T-tests are not significant, but the median test gives significance at the 5% level for both samples.

The difference in BHARs of the TMT companies between the Doeswijk et al. study (-38.40%) and this study (92.58%) is considerably large. This return is heavily influenced by a substantial return of Option N.V. of 1,493.53% in 3 years. To make the comparison more exact, the TMT-sample is stripped from Belgian companies. The result is visible in the bottom of table 3. The BHAR of the Dutch TMT IPOs is -13.19%, which seems somewhat more comparable to Doeswijk et al.’s -38.40%.

The BHARs for the non-technology IPOs are rather similar for the total Dutch and Belgian sample, namely 16.06% in this study versus 20.10% of Doeswijk et al. The results of this paper are not significant for the T-test, but the median test shows significance at the 5% level. For the Dutch only sample, the BHARs are substantially different, dropping to -83.73%. This result is significant at the 5% level using the T-test, and at the 1% level using the median test.

(29)

Table 3

The Three-year Buy-and-hold Returns per Industry for the Total Sample

The table shows the total buy-and-hold returns (BHRs) for both the IPOs and the benchmarks. The three-year industry BHRs are compared to portfolio benchmarks consisting of multi-industry companies. The sectors in the “industry” table are classified according to the Industry Classification Benchmark (ICB) by Dow Jones and FTSE. In addition, a total return is given for the non-technology IPOs and the TMT, for technology, media and telecommunications sector. “N” denotes the number of companies included in every industry. “IPOs” gives the total 3 year average BHR of every industry sector. “Benchmark” gives the average three-year return of the corresponding benchmark portfolio. The column marked “difference” computes the difference in return between the IPOs and the benchmark portfolios. Thus, “difference” stands for the buy-and-hold abnormal returns (BHARs) of the IPOs. In the next column the results of the student’s T-test of means, with the p-values in parentheses, are summarized. The results from the Kruskal-Wallis median test of one-way analysis of variance by ranks follow in the last column, with the p-values in parentheses.

Tests for comparison of results Industry N IPOs Benchmark Difference Mean t-test Median

Kruskal-Wallis Basic materials 3 16.41 68.12 -51.71 -0.5666 (0.6013) 0.4286 (0.5127) Industrials 24 34.73 22.43 12.30 0.3404 (0.7351) 2.0283 (0.1548) Consumer goods 17 11.03 59.14 -48.11 -1.6184 (0.1154) 2.4561 (0.1171) Healthcare 9 215.96 19.89 196.07 1.5485 (0.1411) 1.2183 (0.2697) Consumer services 8 -21.26 24.85 -46.11 -0.9696 (0.3487) 5.8345 (0.0157) Telecommunication 4 -58.42 26.69 -85.11 -2.5469 (0.0437) 3.0000 (0.0833) Utilities 2 311.08 21.02 290.06 0.9092 (0.4592) 0.6000 (0.4386) Financials 8 9.56 29.29 -19.73 -0.5277 (0.6060) 0.8934 (0.3446) Technology 29 114.93 -2.16 117.09 1.1606 (0.1138) 1.9372 (0.1640) N = 104 Non-technology 71 49.89 33.83 16.06 0.6464 (0.5191) 6.1234 (0.0133) TMT 33 93.92 1.34 92.58 1.4284 (0.1580) 3.8752 (0.0490) N = 104 Non-technology NL 21 18.05 101.79 -83.73 -2.1850 (0.0348) 16.099 (0.0001) TMT NL 20 -17.95 -4.76 -13.19 -0.4392 (0.6630) 2.4751 (0.0133) N = 41 5.3 Individual years

(30)

of the IPOs between 1997 and 2005. It becomes clear that the IPOs outperform their benchmarks in 1997, 1998 and 1999. However, from 2000 on there is a three-year period of underperformance. In 2003, 2004 and 2005, the IPOs outperform their benchmarks once again. The returns of the years 1997, 1998, and 2002 up to and including 2008 are not significant. The results of the T-test of 1999 and 2001 are not significant, although the median test is significant at the 1% level in 1999 and 2000. The median test is significant at the 5% level in 2001. The student’s T-test is significant at the 1% level for the results of 2000.

Table 4

The Buy-and-hold Abnormal Returns of Initial Public Offerings per Year

This table depicts the average buy-and-hold abnormal returns per calendar year. The column “year” denotes the year of which the return is measured. Under “N”, the number of IPOs measured in that particular year is given. These numbers include 1st_, 2nd and 3rd year listings of IPOs, totaling 312 observations. The column marked “IPOs” holds the total buy-and-hold return (BHR) over that particular year per IPO. The next row marks the total three-year return of the corresponding benchmark in the same year. The column “difference” computes the difference between the average IPO and the average benchmark returns in the particular year. Thus, “difference” stands for the buy-and-hold abnormal returns (BHARs) of the IPOs. In the next column the results of the student’s T-test of means, with the p-values in parentheses, are summarized. The results from the Kruskal-Wallis median test of one-way analysis of variance by ranks follow in the last column, with the p-values in parentheses.

Kruskal-Wallis 1997 22 84.25 62.01 22.24 1.0907 (0.2816) 2.0167 (0.1556) 1998 43 4.61 -7.34 11.95 1.2162 (0.2273) 0.0042 (0.9484) 1999 70 15.36 10.79 4.57 0.4747 (0.6357) 7.8668 (0.0050) 2000 58 -28.55 -4.66 -23.89 -4.6732 (0.0000) 18.932 (0.0000) 2001 41 -25.90 -12.61 -13.29 -1.5756 (0.1191) 5.8814 (0.0153) 2002 17 -16.94 -4.75 -12.19 -0.8793 (0.3858) 0.6006 (0.4384) 2003 15 44.19 18.90 25.29 0.9405 (0.3550) 0.2275 (0.6634) 2004 13 48.37 46.33 2.04 0.0693 (0.9453) 0.4109 (0.5215) 2005 17 48.29 33.21 15.08 1.1435 (0.2613) 0.6006 (0.4384) 2006 9 15.20 28.20 -13.00 -0.9292 (0.3666) 0.5634 (0.4529) 2007 7 -19.70 -10.60 -9.10 -0.7486 (0.4685) 1.1796 (0.2774) N = 312

(31)

periods, the abnormal returns of IPOs are substantially outperforming the benchmark in hot-issue periods and underperforming in the cold-hot-issue periods

As Ritter (1991) argues, companies do take benefits from going public during ‘windows of opportunity’ in the hot-issue market years. Evidence for this theory seems to be backed by the data on annual volumes of new IPOs. Figure 1 shares this conception.

5.4 Differences among the countries

In this section the difference in return among the 2 countries is addressed. Table 5 shows the BHARs of The Netherlands and Belgium. The three-year average BHAR of The Netherlands is -0.01%, and the average Belgian BHAR is 34.84%. The differences in average annual returns are not significant; neither are the differences in median returns.

Comparing these results of the Dutch IPOs with Doeswijk et al. (2006) and van Hoeijen et al. (1999), gives some notable differences. The one-year BHAR of Doeswijk et al. is 12.40% (not significant) for all the IPOs issued between 1997 and 2000. For this research, the return is 31.52% (not significant). The three-year BHAR of Doeswijk et al. is -18.20% (not significant), while this study gives a -0.01% insignificant BHAR of the IPO-sample.

(32)

Table 5

Annual Buy-and-hold Abnormal Returns for the Two Countries

This table presents the annual average buy-and-hold abnormal returns (BHARs). The column “year” denotes the year of which the IPO return is measured. “N” contains the number of IPOs in the sample. The column marked “BHAR” holds the annual average buy-and-hold return (BHAR), and the three-year return. The column marked “difference” computes the difference in BHARs between the Dutch and the Belgian IPOs. In the next column the results of the student’s T-test of means, with the p-values in parentheses, are summarized. The results from the Kruskal-Wallis median test of one-way analysis of variance by ranks follow in the last column, with the p-values in parentheses.

Netherlands Belgium Tests for comparison of results Year N BHAR N BHAR Difference Mean t-test Median

Kruskal-Wallis

1 41 31.52 63 34.57 -3.05 -0.1034 (0.9178) 0.1084 (0.7420) 2 41 -10.92 63 1.40 -12.32 0.8367 (0.4047) 0.9626 (0.3265) 3 41 -19.16 63 -0.04 -19.12 1.5568 (0.1226) 1.3628 (0.2431) 1,3 41 -0.01 63 34.84 -34.85 1.5126 (0.1335) 1.9980 (0.1575)

5.5 Three-year buy-and-hold abnormal returns with without rebalancing benchmarks Finally, in stead of using annually rebalanced benchmark portfolios, the first-year starting benchmark portfolios are used during the entire three-year measuring period. This method will serve as a robustness test for this study. Table 6 shows the abnormal returns when the initial benchmark portfolio is compared to the same IPO for 3 years. This implies a three-year BHAR of an IPO, in stead of 3 times an annual BHAR. The IPOs yield a three-year outperformance of 28.96%, in contradiction to the 15.55% with the annual portfolio rebalancing method. This result is not significant when applying the T-test, but the median test is significant at the 1% level. Although not being significant, the outperformance of the IPOs over the benchmark seems to be robust.

(33)

become exaggerated when the benchmarks are not rebalanced. Unfortunately, these results do not correspond to the results of table 2. The Belgian results seem robust, because in both methods substantial outperformance of the IPOs has been found. The Dutch returns are not very robust, because the differences between the rebalancing method and the method without rebalancing are pretty large.

The findings could support the methods of Fama and French (1993). When we take a closer look at the research of Eckbo and Norli (2005), the authors present their findings after matching for size and using equal weighting of the companies. This results in a -28.80% underperformance. When value-weighted, the difference shrinks to -19.10%. Further, when the benchmark is also matched to market-to-book ratio, the difference shrinks further to -2.40% with equal weighting and even 0.30% with value weighting. Similar to the Eckbo and Norli findings, it seems that the rebalancing-method gives more exact results.

(34)

Table 6

The Three-year Unbalanced Buy-and-hold Returns for the Sample and the Benchmark This table gives the three-year buy-and-hold return for the total initial public offering (IPO) sample. The benchmark of the IPO-sample is the first-year benchmark portfolio. Instead of annually rebalancing this benchmark as in the Fama and French (1993) methodology, it is continuously compared with the same IPO. The “N” denotes the number of IPOs included. “IPOs” stands for the buy-and-hold average return of the IPOs. “Benchmark” gives the average buy-and-hold return of the corresponding benchmark, and “difference” is the difference between the average returns of the IPOs and the benchmarks. Thus, “difference” stands for the buy-and-hold abnormal returns (BHARs) of the IPOs. The sixth column holds the T-tests of means, with the p-values in parentheses. The results from the Kruskal-Wallis median test of one-way analysis of variance by ranks follow in the last column, with the p-values in parentheses.

Kruskal-Wallis Total sample 1,3 104 64.33 35.37 28.96 1.0421 (0.2986) 9.2354 (0.0024) Dutch sample 1,3 41 0.49 18.92 -18.43 -0.6529 (0.5157) 9.6811 (0.0019) Belgian sample 1,3 63 105.87 46.07 59.80 1.4468 (0.1505) 1.9134 (0.1666)

5.6 Testing the variance of the returns between The Netherlands and Belgium

(35)

IPO can be regarded as buying a ‘lottery ticket’, as to speak in terms of Shiller (1990). It has a higher chance on a large gain, but also on a significant loss.

Table 7

The Test of Variances between Dutch and Belgian IPO-returns

This table displays the results of the variance ratio F-test for the Dutch and Belgian populations. Under “year”, the relevant period can be found, which is BHAR1,3, or the buy-and-hold returns for 3 years, CAR1,2 or the cumulative abnormal returns after two years or CAR1,3, and the cumulative abnormal returns after 3 years. “N” holds the number of companies included in the Dutch or Belgian samples. The next column holds the standard deviations of the samples, and the final column marked “F-test”, which holds the results of the variance ratio test, with the p-values in parentheses. The F-statistic is given by:F =s_L2/s_S2. The F-statistic has an F-distribution with (n_L −1,n_L −1)degrees of freedom.

Year N Standard deviation F-test

BHAR1,3 DUTCH SAMPLE 41 1.5641 3.9722 (0.0000) BELGIAN SAMPLE 63 3.1174 CAR1,2 DUTCH SAMPLE 41 1.4653 3.9302 (0.0000) BELGIAN SAMPLE 63 2.9049 CAR1,3 DUTCH SAMPLE 41 2.9993 4.0056 (0.0000) BELGIAN SAMPLE 63 6.0028

5.7 Multiple regression analysis

(36)

More results of 2000-2002 are available in tables 5 and 8 of the appendix. Finally, the regression results of the IPOs of the period 2003-2005 can be found in table 11, with full information added in tables 6 and 9 of the appendix. Figures 2, 3, 4 and 5 of the appendix display scattergrams with the relation between the market-to-book ratios and the BHARs. Scattergrams plotting the relation between the logsize and BHARs can be found in figures 6, 7, 8 and 9 of the appendix.

Table 8

Definition of explanatory variables Variable Definition

LOGSIZE natural logarithm of market capitalization (million EUR) M/B the value of the market-to-book ratio

INDUSTRY holds a dummy which takes on a 1 for technology and

telecommunications-related IPOs, and a 0 for other industry IPOs COUNTRY holds a dummy which takes on a 1 for Dutch IPOs and a 0 for

Belgian IPOs

The results of the first hot-issue period, which was between 1997 and 1999, are displayed in table 9. The results of the regression analysis are mainly insignificant, except for the intercept coefficients and the natural logarithm of the logsize of the BHAR1,3, CAR1,2 and CAR1,3. The results point at a negative relation between size and the BHARs and CARs after 3 years. The CAR1,2 however, is positively associated with logsize. As expected, the country dummy for The Netherlands gives a positive relation in the first year, and negative in the second and third years, but these results are not significant. The low values of theR and adjusted2 R , and 2 the insignificant F-test values point at a poor fit of an invalid model.

The second period, which was a cold-issue period, shows better annual BHAR results. The regressions are summarized in table 10. For the first BHAR, the 2

R is 51.90%, with a significant F-statistic at the 10% level. For the second and third BHAR, the 2

(37)

the logsizes are significant at the 10% and 5% level, respectively. In this period, the relation between the BHAR2 and BHAR3 and the logsize is slightly positive.

The third and final period covers a “relative”18 hot-issue period between 2003 and 2005, and is displayed in table 11. Again, for the first year BHAR, the R is high with 65.30% and the F-2 statistic gives significance at the 1% level, pointing at a good fit and a valid model. Furthermore, the results of the country dummy (The Netherlands) are substantially negative, as was expected. This result is significant at the 5% level. The TMT variable, which can be found under “industry”, is large and significant at the 1% level. For the other years, and the BHAR1,3, CAR1,2 and CAR1,3, the F-values are not significant. TheR are higher than in other periods, 2 averaging around 40%. The results of the explanatory variables are not significant except for the industry dummy in BHAR1,3, CAR1,2 and CAR1,3, with all results significant at the 5% level. Interestingly, these results are substantial, but they differ much, reaching from -4.6 in CAR1,2 to +9.6 in CAR1,3. The result in CAR1,3 was expected, because the TMT-IPOs show good results, but this is result is mainly reached in the last year. This, on the other hand is not supported by the significant BHAR1 results.

Overall, the multiple regressions do not produce very impressive results. Most F-values point at invalid models with a poor fit, displayed through low 2

R and negative adjusted- 2

R .

18_{Although the economic climate during 2003 and 2005 can not be compared with the late 1990s, there was a}

(38)

Returns (CARs) 1997-1999

This table shows the results of the regression analysis using the firm characteristics as independent variables, and the BHARs and CARs as dependent variables. The table includes the IPOs that went public during the hot-issue period of 1997-1999. The annual BHARs are corrected for the returns of the benchmark portfolio using size (expressed in market value of equity) and the market-to-book ratio. The two- (1,2) and three-year (1,3) CARs are also corrected for the returns of the benchmark and the market-to-book ratio. The difference in measuring the return is explained in the methodology section. For BHAR1, BHAR2 and BHAR3 the corresponding size and M/B values are used. Using the first-year data does not alter the model. For BHAR1,3, CAR1,2 and CAR1,3 the zero-year logsize and the first-year M/B are applied. “Logsize” is expressed as the firm market capitalization measured as the natural logarithm of the share price times the number of outstanding shares. “M/B” is for the market-to-book variable. “Country” includes a dummy which takes on a 1 if the IPO is Dutch and a 0 if the IPO is Belgian. “Industry” includes a dummy, which gives a technology or telecommunications IPO a value of 1 and all other companies a 0. ***,**,* connote statistically significant at the 1%, 5% and 10% level, respectively. T-values are shown in parentheses. Table 4 and 7 show the full regression results of this period.

1997-1999

Dependent variables Explanatory

variables BHAR1 BHAR2 BHAR3 BHAR1,3 CAR1,2 CAR1,3

(39)

This table shows the results of the regression analysis using the firm characteristics as independent variables, and the BHARs and CARs as dependent variables. The table includes the IPOs that went public during the cold-issue period of 2000-2002. The annual BHARs are corrected for the returns of the benchmark portfolio using size (expressed in market value of equity) and the market-to-book ratio. The two- (1,2) and three-year (1,3) CARs are also corrected for the returns of the benchmark and the market-to-book ratio. The difference in measuring the return is explained in the methodology section. For BHAR1, BHAR2 and BHAR3 the corresponding size and M/B values are used. Using the first-year data does not alter the model. For BHAR1,3, CAR1,2 and CAR1,3 the zero-year logsize and the first-year M/B are applied. “Logsize” is expressed as the firm market capitalization measured as the natural logarithm of the share price times the number of outstanding shares. “M/B” is for the market-to-book variable. “Country” includes a dummy which takes on a 1 if the IPO is Dutch, and a 0 if the IPO is Belgian. “Industry” includes a dummy, which gives a technology or telecommunications IPO a value of 1 and all other companies a 0. ***,**,* connote statistically significant at the 1%, 5% and 10% level, respectively. T-values are shown in parentheses. Table 5 and 8 show the full regression results of this period.

2000-2002

(40)

This table shows the results of the regression analysis using the firm characteristics as independent variables, and the BHARs and CARs as dependent variables. The table includes the IPOs that went public during the hot-issue period of 2003-2005. The annual BHARs are corrected for the returns of the benchmark portfolio using size (expressed in market value of equity) and the market-to-book ratio. The two- (1,2) and three-year (1,3) CARs are also corrected for the returns of the benchmark and the market-to-book ratio. The difference in measuring the return is explained in the methodology section. For BHAR1, BHAR2 and BHAR3 the corresponding size and M/B values are used. Using the first-year data does not alter the model. For BHAR1,3, CAR1,2 and CAR1,3 the zero-year logsize and the first-year M/B are applied. “Logsize” is expressed as the firm market capitalization measured as the natural logarithm of the share price times the number of outstanding shares. “M/B” is for the market-to-book variable. “Country” includes a dummy which takes on a 1 if the IPO is Dutch, and a 0 if the IPO is Belgian. “Industry” includes a dummy, which gives a technology or telecommunications IPO a value of 1 and all other companies a 0. ***,**,* connote statistically significant at the 1%, 5% and 10% level, respectively. T-values are shown in parentheses. Table 6 and 9 show the full regression results of this period.

2003-2005

(41)

This study focuses on the existence of long-run underperformance of 104 initial public offerings (IPOs) in the Dutch and Belgian market between 1997 and 2005. Following the theory of Fama and French (1993) I have formed benchmark portfolios on the basis of the market capitalization and the market-to-book ratio of shares. Subsequently, the one-, two- and three-year returns of the IPOs are compared with the average returns of the benchmark portfolios.

Conventional research on long-run performance of IPOs suggests that IPOs perform worse than their benchmarks. However, the market-to-book and size portfolios approach of Fama and French (1993) points at a marginal difference in long-run (3 or 5 year) returns between IPOs and their benchmarks.

The hypothesis of this study is: ‘Initial Public Offerings between 1997 and 2005 do not show long-run underperformance to the benchmark, consisting of same size and same market-to-book companies’. The three-year buy-and-hold abnormal return (BHAR) of the total IPO sample is 15.55%. Because this average return is not significantly different from zero, the hypothesis cannot be rejected. The BHARs for this sample are not very similar to earlier research on long-run performance of IPOs. For the Western countries, only the 17.10% return (not significant) of the Dutch IPO-sample of Hoeijen and van der Sar (1999) is somewhat comparable with the results of this study. However, the time frame of the Hoeijen and van der Sar is not comparable to this study, as it covers 1987-1996.

The three-year cumulative abnormal return (CAR) of the IPO sample is 20.11% (not significant). The two-year CAR is 29.91%, which is significant at the 10% level. These CARs differ from earlier results (for instance Ritter, 1991).

Long-run performance of Initial Public Offerings in The Netherlands and Belgium