Initial Public Offerings on the Warsaw Stock Exchange – Do They Underperform?

(1)

Initial Public Offerings on the Warsaw

Stock Exchange – Do They Underperform?

MSc BA Finance Thesis

Maciej Bora

(1842188)

University of Groningen, Faculty of Economics and Business

First Supervisor: Nanne Brunia

(2)

2

Abstract

This master thesis examines the long-run performance of Initial Public Offerings on Warsaw Stock Exchange using the unique sample of 359 companies that go public in the period 1994 – 2007. I find that the cumulative average abnormal return over three and five years after listing are respectively -37.41% and -66.50% and are both statistically significant at the 1% confidence level. Polish IPOs are highly underpriced, the average initial return in the sample is 51.41%. I use the regression models to explain the long-run performance of Polish IPOs and find that large size companies and highly underpriced companies perform better in the long-run. Exploiting the uniqueness of the sample of Polish IPOs I find that bankruptcy has a negative and significant impact on the long-run performance of IPOs.

(3)

3

I. Introduction

The long-run underperformance of Initial Public Offerings (IPO) has been widely researched in recent 25 – 30 years. The most respected financial papers indicate that IPOs in the period of three or five years after the issue tend to give lower returns than specified control group. The long-run underperformance is well documented on the most important stock exchanges around the world. In the master thesis I would like to study the performance of IPOs on one of the Emerging Markets, Warsaw Stock Exchange. Emerging Markets gain on importance lately and it is worth checking if IPOs perform there similar as they do on a well established stock exchanges.

Although, the long-run underperformance of IPOs is a well-known issue among the financial specialists, there are still some important concerns and questions to be answered. Researchers still try to find the factors that are responsible for the long term underperformance of IPOs. Is it only the effect of dominance of small companies among IPOs, or is it rather something specific in IPOs themselves? Can the aftermarket performance of new offerings be predicted by the investors with the information available before the issue or on the first day of company’s listing on the exchange? Why are IPOs on average underpriced? Are IPOs different from the other companies already listed on the stock exchange? In the master thesis I try to answer some of these questions with the use of IPOs from Warsaw Stock Exchange in the period of 1994 – 2007.

The aim of the master thesis is to investigate the performance of Initial Public Offerings on one of the most popular and attractive Emerging Markets across the world, which is Warsaw Stock Exchange (WSE). The Polish Stock Exchange is one of the most important European Emerging Market as it is much bigger than any other Emerging Market in Europe. In 2008 WSE recorded the turnover of €45.7 billion, which was the highest turnover among exchanges of all emerging European countries. Concerning the market capitalization (data from the Federation of European Securities Exchanges) WSE is three times bigger than the second largest European Emerging Market in Prague.

(4)

4 The thesis is an attempt to examine if the long-run performance of Polish Initial Public Offerings is similar to the performance of the IPOs from other markets, especially from more mature ones like the US stock exchange or London Stock Exchange as there are a lot of researches giving evidence of IPOs long-run performance. The question will be answered by the mere act of comparing the results obtained in my research with those presented in previous papers dealing with the phenomenon of IPOs underperformance.

Another purpose of the thesis is to provide the readers with the information if the underperformance after the first day of the listing on a stock exchange can be predicted given such variables as market value (MV) and market to book ratio (MB) on the first day of company’s listing, initial return, hot or cold issue period and industry. I use these explanatory variables in the regression model, because the existing literature provides evidence that they are able to explain at least some part of IPOs underperformance. Having investigated the empirical results from the literature some IPOs (small companies, firms issued during the hot market period, etc.) may be more likely to underperform the market, than others and the research might enable us to find which are those.

Concerning the academic relevance of my thesis it shows the performance of IPOs on one of the largest IPO markets in Europe. Moreover, the sample of Polish IPOs used in my study is unique as there is a significant group of companies (16.16% of the total number of observations) that are delisted within the test period. I check the relation of bankruptcy of IPOs within the test period and their long-run performance. This has never been done before as it is always assumed that bankruptcy within a sample of IPOs has no significant influence on their average long term performance. In my model I find out that bankruptcy is significantly negatively related to the long-run market adjusted return of IPOs.

I also check the industry effect on the aftermarket performance of Polish IPOs. Many authors describe the cross sectional patterns concerning the industry sectors of IPOs, but it has not been researched if the sector in which IPO operates has an influence on the long run performance. In my model on cumulative average abnormal return I use industry dummies to check the impact of every sector.

(5)

5 foreign investors I use five different benchmarks to measure the aftermarket performance of Polish IPOs.

As other researchers who deal with the phenomenon of IPOs long-run performance, I find that IPOs from WSE strongly underperform when compared to the main index – WGI20. I check the performance in 1, 2, 3, 4 and 5 years after each company entered the market. Using cumulative average abnormal returns I compute the returns of the IPOs sample for the specified period of time. For a one-year period the underperformance reaches the level of -17%, for a two-year period it is -27.9%, for a 3 year period it reaches -37.4%. In contrast to researches on the US market, the underperformance during year 4 and 5 is still present and reaches the level of respectively -48.1% and -66.5%.

In addition to the long-run performance of Polish IPOs, I also study their initial returns. My research on Warsaw Stock Exchange confirms the presence of IPO underpricing on yet another market. The average initial return of a sample of Polish IPOs in 1994 – 2007 is 51.41% and is relatively high when compared to the levels of underpricing on other financial markets. In my model I check the influence of underpricing on the long-run return of Polish Initial Public Offerings.

The structure of the thesis is as follows. Section II is devoted to the literature review of the subject. Section III describes the data and methodology used to execute this research. Section IV presents the results of the long-run IPOs performance. Section V concludes the thesis with a summary and brief economic interpretation of the main findings.

II. Literature review

This section is devoted to the most important previous literature dealing with the long term performance of Initial Public Offerings. As it has been already stated, the underperformance of IPOs has been widely discussed, mainly on the US market, even though several researchers concentrate on other markets as well.

(6)

6 Table 1: Empirical results of long term performance and underpricing of IPOs from the

literature

Author(s) Country Long-run performance Methodology Underpricing

Ritter (1991) USA -29.13% CAR 14.32%

Loughran and Ritter (1995) USA

8.4% vs. 35.3% of matching companies; wealth relative1 of 0.8 (3 year period) 15.7% vs. 66.4% of matching companies; wealth relative of 0.7 (5 year period)

BAHR Not reported

Brav and Gompres (1997) USA

Venture-Backed IPOs: 44.6% vs. 65.3% of S&P 500 index (benchmark); wealth relative = 0.88 Nonventure-Backed IPOs: 22.5% vs. 71.8% of S&P 500 index; wealth relative = 0.71

BAHR Not reported

Ritter (1998) USA -5.2% per year in the 5

years after going public

Average Annual

Return

15.80%

Brav et al. (2000) USA -38.3% (5 year period) CAR Not reported Purnanandam and Swaminathan (2004) USA

-5.81% (3 year mean annual market adjusted

return)

BAHR

11.4% (abnormal initial return)

Sapusek (2000) Germany -21.04% (5 year period)

Buy and Hold Abnormal

Return

11.49%

Helwege and Liang (2004) USA

Underperformance; Helwege and Liang (2004) report only

wealth relatives

BAHR Not reported

Levis (1993) Great Britain

Underperformance varies from -8% to -23%

based on benchmark used (3 year period)

CAR 14.30%

Espenlaub et al. (2000) Great Britain -15.90% CAR Not reported

Goergen et al. (2006) Great Britain -22.00% CAR 9.74%

Kumar (2007) India -14.69% (3 year period)

-15.86% (5 year period)

Excess

BHAR 27.26%

Chi and Padgett (2005) China 10.30% CAR 127.31%

Chan et al. (2004) China A-shares: 25% B-shares:

31% CAR A-shares 178% B-shares 11.6%

1_{Wealth relatives are calculated as Σ(1+R}

i,T)/Σ(1+Rbench,T), where Ri,T is the buy and hold return on IPO i for

(7)

7

Bildik and Yilmaz (2007) Turkey -84.50% BAHR 5.94% (excess

initial return) Aussenegg (2000) Poland 225% vs. 213% of WGI (benchmark); wealth relative 1.037 BAHR 38.46%

The research conducted by Ritter (1991) constituted the starting point of the intense discussion about the IPOs long-run performance. Before, only two anomalies were documented in the academic literature: the underpricing phenomenon and the ‘hot market’ phenomenon. By anomalies I mean the violations of the market efficiency. Underpricing is a short term violation of market efficiency and occurs when market value of IPO is higher than its offer value. Underpricing is often purposely used by companies conducting IPO and their underwriters to attract the investors. The hot market phenomenon is defined by Ibbotson (1975) and is described later in the section.

Ritter (1991) uses the sample of 1,525 IPOs that went public in the period of 1975-84. He uses a 3 year test period in his research and finds that these companies significantly underperform a group of so called ‘matching firms’ selected by size and industry. The cumulative average abnormal return for the sample of IPOs is -29.13%. This number is statistically significant and imply that IPOs perform worse than similar companies listed on New York Stock Exchange.

Ritter (1991) points out that the companies with the highest initial return display a tendency to have the lowest returns in his sample of IPOs. This is economically rational since we can assume that high underpricing might be a result of the overoptimism of the investors. If the closing price of a company on the first trading day is way above the ‘fair value’ of this company, its price is pulled towards this ‘fair value’ in the long term. This is a straightforward reason for the aftermarket underperformance of many IPOs.

Loughran and Ritter (1995) in their research on US IPOs and seasoned equity offerings (SEO) find that based on the realized returns, an investor would have to invest on average 44% more money in the IPO company than in nonissuer of the same size to reach the same wealth 5 years after the offering date.

(8)

8 so called buy and hold returns. In addition, they calculate the wealth relatives to determine if the specified sample of IPOs underperform the benchmark. They find that the underperformance is stronger for the nonventure capital-backed companies than in case of venture capital-backed companies. Their paper suggests that IPOs should not be treated as one group, because they strongly differ from one another. Based on these differences Brav and Gompres (1997) point out that IPOs underperformance may not be an issuing firm effect, but a result of small size and low book to market ratio.

Purnanandam and Swaminathan (2004) use price multiples of non-IPO industry peers to compute fair value of IPOs. They compare calculated fair values of IPOs with their offer prices. Surprisingly, they come to the conclusion that IPOs are rather overvalued than underpriced. In the sample of more than 2000 IPOs from 1980 – 1997 the median IPO is overvalued by about 14% to 50% depending on the matching criteria. Purnanandam and Swaminathan (2004) state that IPO may be overvalued and underpriced in the same time. This is possible if issuers underprice the offer not based on the long-run fair value but with respect to the maximum offer price they can set with the given demand. Purnanandam and Swaminathan (2004) check the relationship of overvaluation and underpricing. Not in line with the asymmetric models they find out that the most undervalued IPOs earn the lowest first day returns. Undervalued IPOs underperform overvalued IPOs by 5% - 7% on the first day of trading. Purnanandam and Swaminathan (2004) also check the long-run performance of overvalued and undervalued IPOs. In the long term basis undervalued IPOs outperform the overvalued IPOs by 4% to 10% per year in the period of two to five years. As a whole group IPOs underperform broad market indices. The regression model used by Purnanandam and Swaminathan (2004) shows that overvaluation is significantly negatively related to the long-run risk adjusted return. In other words, the higher the overvaluation of IPO, the lower the long-run return of it.

Benchmark sensitivity

(9)

9 after entering the market. Sapusek (2000) divides his test period into two subperiods and the results are highly interesting. During the first subperiod (1983-1988) the performance is very sensitive to the benchmark used. For 5 indices he finds overperformance of IPOs, 4 other benchmarks result in underperformance. In the second test subperiod (1988-1993) the results are much more homogeneous. The sample of German IPOs underperform all 12 benchmarks. Sapusek (2000) explains the differences in the results between two subperiods with the fact that the second subperiod is characterized by the excellent performance of the German stock market in general. In particular the market indices performed very well. In comparison, the non-index companies perform poorly. IPOs are included mainly in the latter group of stocks in this subperiod.

Loughram and Ritter (1995) also present some evidence of benchmark sensitivity of IPOs on the US market. Their results are not as striking as Sapusek’s since their sample of IPOs always underperforms the benchmark used, but with different magnitude. Loughram and Ritter (1995) record the strongest underperfomance of IPOs against the sample of size-matched companies (average 5 year return of 15.7% vs. 66.4% and wealth relative of 0.7). The weakest underperfomance is a result of comparing the IPOs against Standard & Poor’s 500 (average 5 year return of 15.7% vs. 38.3% and wealth relative of 0.84).

Weighting scheme

(10)

10 In the thesis the standard approach of the majority of academic writers working on IPOs long-run performance is followed and presented results are obtained with both weighting schemes. I expect value weighted abnormal returns to be higher than those calculated with equal weighted methodology. This expectation is based on previous studies which prove that value weighted scheme eliminate much of the long-run underperfomance.

Loughran and Ritter (1995) report in their paper returns of US IPOs from the period of 1970 – 1990 calculated with both methods. The average five year return on portfolio of IPOs increases from 16% when calculated with equal weights to 34% when value weighting scheme is used. Interestingly, weighting scheme does not affect the return of the portfolio of matching firms. The difference is only 1 percentage point in favour of value weighting (67% vs. 66%). Visibly, the underperformance is limited in case of value weighted returns.

Brav et al (2000) obtain similar results. They investigate the performance of the US IPOs with both cumulative abnormal returns and buy and hold returns. Since the research presented in the thesis is focused on the first method, only CARs are taken into consideration. The sample of IPOs underperform all but one benchmark. The equal weighted five year CARs vary from 26.5% to 38.3% while value weighted five year CARs range from 20.8% up to -11.1%. The strongest and the weakest underperformance is recorded in both cases for the same indices, S&P 500 and CRSP Equal-Weighted respectively. Brav et al (2000) give evidence that value weighting returns cut the underperformance roughly in half. Moreover, they find no evidence of IPOs long-run underperformance against the portfolio of stocks matched by size and book to market value. Brav et al (2000) show that the returns of IPOs are similar to the returns on companies with similar size and book to market ratio that are already listed on the stock exchange. This raises a question if underperformance is driven by the issuance or by the small size and low book to market value of these companies.

(11)

11 Hot and cold issue periods

Another important feature of long-run performance of IPOs is a fact that this performance is not only sensitive to the benchmark used for a comparison, but also the period of issuance plays a crucial role. In the academic literature this issue is called ‘hot market’ phenomenon. There are several definitions of ‘hot and cold market’ periods. Ibbotson and Jaffe (1975) are among the first academics to examine the issue. They prove that the first month issue premium2 can be predicted by investors and issuers. Ibbotson and Jaffe (1975) show that the serial dependence is present in the series of first month average returns. This serial dependence allows one to predict the level of first month performance in the future months. To have high returns investors should accumulate their purchases in months when this issue premium is expected to be high. The ‘hot issue markets’ are defined by Ibbotson and Jaffe (1975) as specific periods in which IPO’s first month performance gives investors abnormal returns. It has to be borne in mind that in the master thesis different definition of hot period is used. Ibbotson and Jaffe (1975) come up with one important conclusion from their research. The investors, who decide to invest their money into the new offerings, can on average expect an abnormal return of 16.83% above the market index return by the end of the month. Furthermore, investors should avoid ‘cold issue’ markets when the average first month return is low or even negative.

The research of Ibbotson and Jaffe (1975) has one crucial downside. The one month abnormal return of the IPO is calculated including the initial return of the first day. Basing on the existing evidence from the academic literature it can be assumed that IPOs tend to be strongly underpriced which drives high initial returns. As a result the aftermarket performance is strongly influenced by the initial return. In majority of the financial papers on IPOs long-run performance the aftermarket period does not include the initial return. Moreover, investors’ overoptimism often affects the first month average abnormal return of new companies on the market. Later the value of these firms is pulled back towards their fair value which in majority of the cases is lower than their initial market value on the first day of listing. The second downside of measuring the performance of IPOs based on the offer price is a presence of reduction in share subscription (sometimes rate of reduction reaches even 90% or more). This problem is probably more significant in case of private investors since institutional investors are often ‘invited’ by the underwriters and can expect rational

2_{First month issue premium is defined as a difference between first month’s return on the particular security}

(12)

12 number of shares. The reasons stated above lead to the conclusion that it is more appropriate to measure the performance of IPOs based on the closing price on a first day of company’s listing.

In recent literature the concept of ‘hot’ and ‘cold’ issue market has a slightly different meaning than in the paper of Ibbotson and Jaffe (1975). In the present thesis I base the idea of ‘hot’ and ‘cold’ market periods on a research of Helwege and Liang (2004). For their research they assume that hot and cold markets are defined by the total number of IPOs completed per month. Basing on previous evidences, they expect IPOs from periods with the largest number of offerings to experience poor long-run performance. Moreover, they ignore in their research companies that go public in so called lukewarm markets. The method enables them to find hot and cold periods for the whole market. In contrast, setting hot and cold market based on underpricing is specific for individual companies.

According to Helwege and Liang (2004) it is also possible that several industries are in hot issue period, whereas other industries may be in cold period at the same time. Nevertheless, they state that IPOs in periods with relatively large number of IPOs are not more concentrated in specific industries than in cold markets. During their test period IPOs are mainly from the same range of industries. In conclusion, there are simply more issuing companies from the same industries during the hot market than during the cold periods.

Loughran and Ritter (1995) study the long-run performance of the US IPOs in the period of 1970-1990. Although, they do not comment on the hot and cold market effect, their results are very interesting for the issue. It occurs that companies which perform the initial public offering during the period of cold market (1974 – 1978) have the highest absolute and abnormal returns in the whole test period. In the literature on the long-run performance of IPOs it is believed that this superior returns of companies from cold periods can be explained by the fact that these companies are of higher quality (better established companies with higher operating earnings, higher value of assets, etc.). Lerner (1994) argues that hot market periods are a result of managers taking advantage of a ‘window opportunity’ to do an IPO and irrationality of investors, whose market views are often biased in periods of good prosperity.

(13)

13 differences, like the level of assets or operating income, disappear. However, it has to be emphasized that, similarly to other authors, Helwege and Liang (2004) find that when hot market is defined by the volume of IPO in the month of offering, hot market firms perform worse in the long-run than cold companies. Moreover, they perform 24 t-tests on the average abnormal performance of the IPOs. IPOs from cold periods six times significantly overperform the market (for all 24 observations the abnormal return is positive), whereas IPOs from hot period in 16 tests significantly underperform the benchmark used for comparison. The findings of Helwege and Liang (2004) about the performance of hot and cold IPOs should be treated with more attention as they concentrate their research only on this issue. Basing on the evidence from the literature I expect that IPOs from hot periods should have worse performance than IPOs from cold periods.

Ljungqvist et al. (2006) develop a model of IPO pricing in hot market which is trying to explain the relationship between underpricing (short-term issue) and long-run underperfomance (long-term phenomenon). The main assumption of the model comes from behavioral finance. In some periods investors tend to be ‘irrationally exuberant’ about new issues (i.e. about IPOs from particular industry). The existence of these investors links together three phenomena: IPOs underpricing, hot issue markets and long-run underperformance. Irrational investors and their beliefs about company’s value far beyond company’s fair value are considered one of the main reasons of underpricing. As in the long-run those investors tend to loose their optimism they cause the underperformance in the long-run. What Ljungvist et al. (2006) find puzzling is the fact that issuers do not take advantage of exuberance by raising the offer price during the hot market periods. This would lead to the elimination of underpricing and higher incomes from the IPO for the issuer.

International evidence of IPOs underperformance

(14)

14 a company and its long-run performance. With all other things being equal the larger the firm, the higher the market-adjusted return of a company.

I also examine some papers that deal with IPOs performance on Emerging Markets. Kumar (2007) investigates both short and long-run performance of bookbuilt3 IPOs in India. In case of the latter his results are quite confusing. One year after listing Indian IPOs have an excess return of -13.68%. Surprisingly the underperformance after two years is much weaker, the excess return reaches the level of -5.2%. Next in year four the excess return jumps to 10.4%, therefore we witness the overperformance of Indian IPO four years after going public. 60 months after listing we observe underperfomance again, with the magnitude of -15.86%. Kumar (2007) does not perform statistical tests if his results are significantly different from zero. The author gives no explanation of such performance of his sample of IPOs. It is highly probable that this research has insufficient number of observations as for a 5 year period there are only 10 companies in the sample.

There are also some studies made on probably the most important Emerging Market nowadays, China. Chi and Padgett (2005) examine the sample of Chinese IPOs issued from 1996 to 1997. First thing to be mentioned is a very high level of underpricing among Chinese IPOs. Chi and Padgett (2005) find that the average underpricing is 127.31%. The previous studies suggest that companies with such high underpricing should underperform on average the market in the long term perspective. According to Chi and Padgett (2005) this is not the case on the Chinese stock markets4. The average three year market adjusted cumulative return is positive and has a value of 10.3%. Chan et al. (2004) find that in the period of three years after listing both A-share and B-share IPOs outperform the benchmark portfolio5.

The European Emerging Markets are not particularly active as far as the initial public offerings are concerned. The only two where sufficient activity in this field can be observed are Turkey and Poland. Bildik and Yilmaz (2007) study the Turkish IPOs from the period of 1990-2000 and show that they are characterized with very strong long-run

3_{Book building is one of methods of issuing share. Book building involves asking the professional investors}

how many shares they want to buy and what price they are ready to pay. As a result the demand curve is constructed and based on it the company and its underwriter determine the IPO offer price.

4

There are two stock exchanges in China. IPO in the research of Chi and Padgett (2005) were taken from both of them.

5_{China B Shares were originally developed as stock shares for foreign investors, as the A share market is}

(15)

15 underperformance. For the three year holding period they find underperformance of -84.5% against the market portfolio represented by the ISE-100 index.

I conduct my research on one of the biggest IPO markets in Europe (by volume of offerings), Warsaw Stock Exchange. Even though this market seems to be ideal for a research of IPOs long-run performance (as an example of IPOs on Emerging Markets), there is a rather limited interest of academic scholars in it. Aussenegg (2000) study the issue of privatization versus private sector initial public offerings in Poland. Although his study and results may seem slightly out of date, I find them very interesting and worth mentioning. Using buy and hold returns and wealth relatives Aussenegg (2000) shows evidence of overperformance of the whole sample of IPOs. Privatization companies overperform the benchmark in a period of one, two and three years after going public. Private companies overperform in all cases with exception for the longest test period. The huge differences between the mean and median in the sample of all IPO seem to be especially interesting. While mean reaches the level of +11.5%, median is negative with value of -61.1% (median is significantly different from zero at the 1% confidence level). Out of all 83 companies in the sample 55 experience negative long-run performance. These numbers suggest that the results obtained by Aussenegg (2000) may be strongly affected by the positive outliers in the sample. Therefore, I believe there is a need for another study of Polish IPO long-run-performance.

III Data and Methodology

The aim of this section is to present the methodology and data used in this master thesis research to the reader. I present the methodology and main ideas that drive my research. What follows is a detailed data description. I reveal my sources of data and present the descriptive statistics of variables

Methodology

(16)

16 and hold return assumes that investor buys the stock on a first day of listing and holds it till the particular anniversary of stock listing. Having investigated a vast number of papers on IPOs aftermarket performance, it appears that both methods lead to similar results. Due to this fact it has been decided to use only cumulative abnormal returns in the thesis in order to study the IPOs performance. In the master thesis I use logarithmic returns instead of arithmetic ones. In my methodology I do not rebalance IPO portfolios every month, I only eliminate IPOs going bankrupt or delisted due to other reasons. As a result my cumulative average abnormal returns yield the same results as buy and hold returns presented by Ritter (1991). A group of academics such as Kothari and Warner (1997), Barber and Lyon (1997), Lyon et al. (1998) compare the various abnormal performance measuring methodologies and are left in confusion without choosing a clear winner as an optimal, best suited methodology. Fama (1998), who argues that CARs are less likely to reject market efficiency, is also supportive for the decision to use only CARs in my thesis. CARs posses lower tendency to magnify the underperformance than buy and hold returns. Last but not least, distributional patterns and statistical tests for CARs are better understood.

Brav and Gompres (1997) calculate returns in two ways, although in their paper they present only buy and hold results. Nevertheless, they state that both methods give very similar results in case of abnormal returns and magnitude of the underperformance.

The issue which is also given much concern in the IPOs long-run performance is the choice or formation of appropriate benchmark. What is often seen in the academic papers, especially in those dealing with the US IPOs, is the procedure of choosing a matching company for every IPO included in the database. However, Brav et al. (2000) criticize the process of picking single security similar in case of firm specific accounting or price based characteristics. In the thesis I do not create the matching group of companies for the sample of IPOs. The main reason behind my decision is a fact that WSE is incomparably smaller than New York Stock Exchange and it is impossible to find matching companies for each single observation.

(17)

17 Two kinds of returns are calculated for different time intervals. First of all, the initial return is calculated. The initial return (IR) is defined as the first closing price listed on the WSE to the offer price. It is presented in equation 1:

IR

i

= (P

1i

/OP

i

- 1) x 100

(1)

where IR i is the initial return of stock i, P1i is the closing price of stock i on a first day of its listing on WSE and OPi is the offer price of stock i.

The second kind of return used in this master thesis is the aftermarket return calculated as cumulative abnormal return. The initial return is not included. Five different periods are used to measure the long-run performance. This results in five different CARs, one for every test period. What needs an explanation is the definition of a month used in this study. I follow the methodology of Ritter (1991) here and months are defined as 21 successive trading days relative to the IPO date. Therefore, closing prices of days 1 and 22 compose month one, of days 22 and 43 compose month two, etc. For the IPOs which go bankrupt or are delisted from the WSE before each anniversary the test period is truncated and they are later not included in the tests for longer periods. As an illustration: if a company is delisted in the month 26, it is included only in the samples of one and two year period and is not taken into account for longer periods. Some researchers still include such companies in the test of longer periods. What is important, other authors use only one period in their researches of long-run performance. As such, they need to include delisted and bankrupt companies in the long test periods, because otherwise they would have to completely eliminate them from their sample. This would bias their results. Due to the fact that I study one, two, three, four and five year periods, I find it rational not to include companies with short stock listing period in the test for longer periods. If a stock is delisted from the market after 2 years, it is impossible for investors to hold this stock for longer period.

(18)

18 Warsaw General Index is an index of WSE companies calculated since the beginning of WSE in April 1991. The question may arise why as a main benchmark of my research I use WGI 20 instead of WGI. This is because the foreign investors, who invest on Warsaw Stock Exchange are mainly interested in companies listed within WGI 20.

Warsaw SWIG 80 is an index of the 80 small size companies. I use this index among alternative benchmarks since previous studies have proved that IPOs are mainly small size companies. Therefore, it is useful to compare them with similar sample of stocks.

Additionally, to make my research more interesting for all sorts of investors, I compare my sample of IPOs with two MSCI indices. The first one is created to track the performance of the Polish market, while the other tracks all Emerging Markets in the region.

The benchmark adjusted return for stock i in month t is calculated according to equation 2:

AR

it

= r

it

– r

mt (2)

where rit is the monthly return on stock i in month t and rmt is the market return in month t. The cumulative abnormal return for every stock i in the sample is calculated as the summation of ARit:

CAR

iz

=

1 z t=

∑

AR

it (3)

where z means length of the test period; z can have values of 12, 24, 36, 48 and 60 respectively for the one, two, three, four and five year period.

Having the CARs for every observation in the sample, it is easy to compute the cumulative average abnormal return of the whole portfolio of IPOs. The equally weighted CAR for every test period is calculated as follows:

1

CAR

CARxY =

n

n iz i=

∑

(4)

Significant positive value of CARxY is a signal that IPOs outperform the benchmark used, while significant negative value is a clear evidence of IPOs underperformance.

(19)

19

CARxY =

1 n i=

∑

w

i

× CAR

iz (5)

where wi is the weight of every particular stock i in the sample and is given by equation (6):

1

MV

w =

MV

i i n i i=

∑

(6) Regression Model

In the next part of my research I want to provide the investors with the information if the future performance of IPOs can be predicted with the data available on the first day of their listing. For this reason, I create a regression model:

CARxY

_i

=

α

5

COLD +

ε

i (7)

where lnMVi is a logarithm of market value, lnMTBVi is a logarithm of market to book value, IRi stands for the initial return of stock i, HOTi is a dummy variable, which equals 1 for companies issued during hot market period and 0 otherwise, COLDi is a dummy variable equal 1 for companies issued during cold market and 0 otherwise. The definitions of hot and cold market periods are given later in this section.

In table 2 I present the expected signs of the variable used in the regression model. I base my expectations mainly on the existing literature on IPO long-run performance.

Table 2: Expected signs of the variables used in the regression model Variable A priori sign expected

ln MTBV -

ln MV +

IR -

HOT -

COLD +

(20)

20 Since the long-run performance is measured on the basis of five different time intervals in this thesis, there are also five regression models. As the regression given in equation (7) measures the influence of company’s size (market value) and its market to book ratio, I do not use the Fama and French (1993) model.

where DF is a dummy variable that equals 1 for financial companies and 0 otherwise, DBM is a

(21)

21 Table 3: Classification of Initial Public Offerings by industry sector together with

cumulative average abnormal returns

Sector

No of

observations CAR1Y CAR2Y CAR3Y CAR4Y CAR5Y

Financials 34 -21.70% -26.26% -29.06% -25.56% -38.00% Basic Materials 34 -23.10% -30.35% -31.63% -46.63% -59.97% Technology 45 -11.00% -16.60% -19.08% -50.06% -67.22% Industrials 103 -20.96% -37.12% -61.43% -70.95% -96.30% Consumer Goods 80 -23.69% -35.16% -43.14% -44.56% -58.25% Consumer Services 37 7.90% 1.48% 12.43% 15.94% -31.34% Health Care 11 -25.33% -29.56% -22.38% -60.54% 6.54% Utilities 6 8.64% -6.04% -26.98% -49.02% -19.58% Oil&Gas 3 -30.41% -9.36% -10.86% 10.93% 16.80% Telecommunication 6 -0.50% -40.10% -30.67% -180.01% -159.76% TOTAL 359 Notes:

CAR1Y – Cumulative Average Abnormal Return in one year period CAR2Y – Cumulative Average Abnormal Return in two year period CAR3Y – Cumulative Average Abnormal Return in three year period CAR4Y – Cumulative Average Abnormal Return in four year period CAR5Y– Cumulative Average Abnormal Return in five year period

The industry effect is measured for all five test periods. It is possible that a significant industry effect for one sector may be in fact caused by i.e. initial return or market value from the previous regression model. To eliminate the risk of finding two significant variables, which in reality might be the same, I run another regression with significant dummy variables from equation (8). By doing so, I present all the significant factors in the aftermarket performance of IPOs on WSE, what should help investors to make a decision about their investments in particular IPOs.

(22)

22 Table 4: Final regression models on CARs with significant variables from models given

in equation (7) and equation (8)

11

D

E

+ ε

i

(14)

where DB

is a dummy variable equal 1 if company goes bankrupt before its sixth anniversary

and 0 otherwise, DMA is a dummy variable equal 1 if company is involved in merger or acquisition within six years after going public and 0 otherwise, DD is a dummy variable equal 1 if company is delisted by the decision of General Meeting of Shareholders and 0 otherwise, DE is a dummy variable equal 1 if company is excluded by the WSE and 0 otherwise.

(23)

23 influence on the long term performance of IPOs, while it is not decided what sign should be expected in case of delisting of a company based on the decision of its General Meeting of Shareholders.

Data and descriptive statistics

The majority of the data used for this thesis has been collected from Thomson Reuters DATASTREAM, the world’s largest financial database. Even though DATASTREAM offers its users the access to more than 140 million time series, over 10 000 different datatypes and over 3.5 million financial instruments and indicators, the data on Polish initial public offerings is in many cases incomplete.

First of all, many firms do not have complete stock price history in DATASTREAM. There are some companies that entered the market in 1997, but in DATASTREAM they are available only from 2001. For a research on IPOs, where the first day of listing is often crucial for the whole test period, it presents a serious problem. Eliminating companies with missing data would heavily limit my sample and strongly bias my results. Fortunately, I managed to overcome this problem by completing my database with use of WSE InfoSpace6 web database.

I encountered the same problem while collecting the data for market value and market to book value variables. Again, DATASTREAM occurred to be a low quality source of data for this kind of information for companies listed on Warsaw Stock Exchange. I managed to complete the market value variable by multiplying the initial number of stocks by the closing price at the first day of listing of particular firm. Having the market value of every IPO in my sample, all I needed was book value in order to calculate the market to book ratio. I completed the book values of companies in my sample using the annual reports for the companies still listed on WSE, for the companies which were delisted or went bankrupt I used ‘The Almanac of Polish Capital Market’ as a data source. Appendix A presents organized raw and transformed data used in this master thesis.

I conduct my research on a period of time between 1994 and 2007 on a total number of 359 initial public offerings. I also use the stock returns from 2008 to measure the long-run performance of IPOs. During the analyzed period a total number of 431 of IPOs has been

6_{WSE Infospace is an economic newswire covering Polish and foreign firms listed on the Warsaw Stock}

(24)

24 executed on the Warsaw Stock Exchange. The number of IPOs included in the research and actually done on WSE vary because of missing data for 33 domestic companies and 25 foreign firms listed in Poland. Moreover, I excluded from my sample 14 National Investment Funds that entered the stock exchange in 1997. In my research I study the long-run performance of Polish IPOs with five test periods. I compute one, two, three, four and five year cumulative average abnormal returns. The fact that a significant number of companies go bankrupt or are delisted during five years after issue is the second factor which affects the number of observations for particular CARs.

Figure 1 presents the number of IPOs on Warsaw Stock Exchange during the research period. As can be seen, the number of IPOs is not stable throughout the test period and there is no clear trend in it. The lowest number of IPOs can be seen for a period of 2000 – 2003, when financial markets were in recession (especially Emerging Markets including the financial crisis in Russia, Argentina, etc.).

Figure 1: Number of IPO on WSE in period 1994-2007

In the sample of Polish IPOs in 1994 – 2007 I specify two special states of the market concerning the quantity of initial public offerings conducted in particular year. One of them is characterized by relatively high number of IPOs (hot issue market), while the other one has relatively low number of IPOs (cold issue market). In the master thesis I follow the methodology of Helwege and Liang (2004). I simplified their procedure as they are using monthly quantity of IPOs and I use yearly numbers to define the hot and cold periods. Hot market period is the top quartile of observations, while cold period is defined as the bottom quartile of observations (every year is treated as a single observation). The

(25)

25 observations with more than 32.25 IPOs (value of quartile three) are classified as hot market periods, whereas years with less than 13 observations (value of quartile one) constitute the cold market periods. As a result years 1997, 1998, 2005 and 2007 are hot market periods, while years 2000, 2001, 2002, 2003 are cold market periods.

An important feature of my sample of IPOs is the presence of significant group of companies that are delisted from the stock exchange within six years after going public. I decided to check the influence on the long-run-performance of IPOs of bankruptcies, mergers & acquisitions, delisting based on the decision of General Meeting of Shareholders and exclusions forced by WSE. In other researches the group of companies delisted during the test period is said to be too small to have a significant influence on the average long-run performance of IPOs. In the sample of Polish IPOs from the period 1994 -2007 there are 58 companies (16.16% of the whole sample used in the master thesis) delisted before their sixth anniversary. Table 5 presents the descriptive statistics of IPOs delisted within 6 year period.

Table 5: Descriptive statistics of Polish IPOs in 1994 -2007 delisted within 6 years after going public

0 Bankruptcies Mergers and

Acquisitions Delistings by GMoS Exclusions by WSE No of observations 17 21 15 5 IR 21.81% 241.73% 298.61% 17.91% CAR1Y -26.66% -14.95% -6.75% -37.45% CAR2Y -56.37% -4.16% -16.36% -31.78% CAR3Y -104.16% -9.67% -37.36% -91.75% CAR4Y -123.18% -3.49% -20.56% -125.10% CAR5Y -193.86% -10.59% -22.41% -216.14% Notes:

(26)

26 acquisitions or are delisted due to their own decision. Underperformance within first two groups is much stronger than in the two others.

Among 58 companies delisted for different reasons from Warsaw Stock Exchange 22 are from Consumer Goods sector, 12 from Industrials, 7 from Technology and 6 are Basic Materials producers. Within the Consumer Goods sector 27.5% of companies are delisted before their sixth anniversary. 5 companies issued during the cold market periods face delisting (17.2% of all cold market companies), while in case of hot market period number of such companies is 36 (18.6% of companies issued in hot market).

The number of observations used to calculate every single CAR is presented in the table 6, along with other descriptive statistics of dependent variables. The number of observations varies among test periods. This is a result of a fact that for example IPOs from 2006 are included only in the one and two year test periods. Moreover, as already mentioned, there are companies, which go bankrupt or are simply delisted within first five years of their listing on the stock exchange.

Table 6: Descriptive statistics of dependent variable – Cumulative Average Abnormal Return

CAR 1Y CAR 2Y CAR 3Y CAR 4Y CAR 5Y

Mean -0.17 -0.28 -0.37 -0.48 -0.67 Median -0.18 -0.31 -0.39 -0.52 -0.65 Maximum 1.81 3.32 2.18 2.19 2.76 Minimum -1.62 -3.20 -3.17 -3.21 -3.53 Std. Dev. 0.52 0.81 0.86 0.94 1.03 Skewness 0.03 0.26 0.05 -0.15 -0.10 Kurtosis 3.10 4.61 3.31 3.49 3.58 Jarque-Bera 0.18 34.89 1.15 3.04 2.81 Probability 0.91 0.00 0.56 0.22 0.25 Observations 359 295 257 220 177

(27)

27 distributed. All other CARs are normally distributed based on Jarque-Bera test; their kurtosis is close to 3 and they are characterized with only minor skewness. According to Brooks (2008) the violation of the normality assumption in case of CAR2Y is virtually inconsequential, as a result of large number of observation.

Table 7 provides the descriptive statistics of the variables included in the basic regression model with CAR of 1 year test period as a dependent variable. As shown in the table, any of the variables follow normal distribution.

In case of the initial return variable the maximum value is extremely high. This is an IPO of company named EMAX which is done in 2002. The offer price for this company is 1 PLN and at the end of a day this stock reaches the value of 46 PLN. What is interesting, the maximum initial return is recorded for IPO executed during the cold issue period, while this return is lower on average than during the hot issue period. The worst Polish IPO, with a negative initial return of -52.5% belongs to company named Murawski, which entered WSE in 1998. In contrast, this minimum level of initial return is reached during the hot issue period. Table 7: Descriptive statistics of independent variables used in the regression model

(equation 7) with Cumulative Average Abnormal Returns of 1 year period as dependent variable LN_MV LN_MTBV IR HOT COLD Mean 4.699 0.632 0.514 0.540 0.081 Median 4.486 0.603 0.048 1.000 0.000 Maximum 10.106 2.965 45.000 1.000 1.000 Minimum 1.430 -1.430 -0.525 0.000 0.000 Std. Dev. 1.409 0.695 3.431 0.499 0.273 Skewness 0.760 0.396 11.983 -0.162 3.077 Kurtosis 4.039 3.389 152.205 1.026 10.467 Jarque-Bera 50.74 11.60 341596.60 - - Probability 0.000 0.003 0.000 - - Observations 359 358 359 359 359

The descriptive statistics of dependent variables from four other regression models can be found in the appendix B to this master thesis. The descriptive statistics from models defined by equations (9) – (13) are presented in appendix C.

(28)

28 Table 8: The long-run performance of IPOs in 1994 – 2007 classified by cold and hot

issue periods Cold issue period Hot issue period CAR1Y -27.42% -19.88% CAR2Y -48.75% -27.90% CAR3Y -49.79% -37.08% CAR4Y -38.82% -51.91% CAR5Y -35.59% -80.76% Notes:

Unfortunately, hot issue period includes two very recent observations. This fact affects the results as i.e. IPOs from 2007 are included only in CAR1Y. Warsaw Stock Exchange IPOs issued during cold period underperform the market stronger than those from hot period for the first three years after going public. The situation changes for longer periods. CARs of cold issue period raise, while those of hot period drop dramatically. CAR5Y for IPOs issued during cold market is -35.59%, way above the average of the whole sample. In contrast, CAR5Ycalculated for hot market IPOs has a value of -80.76%. Such a strong long-run underperformance of hot market IPOs is a result of poorer performance of IPOs from the nineties.

(29)

29 1997 and 1998). Two other hot periods have relatively low levels of underpricing. In case of 2005 it is only 10.73% and for 2007 even less, 4.87%, which is very close to the average initial return of cold periods when measured without the outlier. The recent IPOs are probably better priced by the underwriters, companies are also willing to get as much money as possible from the issue and it is in their interest to price their stocks high, but close to the price that the investors are keen to pay. On the other hand, it can be the case that investors are not so optimistic about the IPOs in the recent years as they were in the past. Having in mind poor long-run performances of IPOs from the 90s’ they are more careful with picking new issuances for their portfolios.

The long-run performance of IPOs from WSE in 1994 – 2007 is also checked on the basis of first day market value and market to book ratio. The results are presented only in appendix D as they are not crucial for this master thesis. Not surprisingly, the top quartile of IPOs by the market value based on first day closing price has the highest average CARs for every five test periods. More interesting is the fact that bottom quartile of IPOs is not the one with the strongest underperformance in any of the test periods.

Table 9: The long-run performance of IPOs in 1994 – 2007 by the industry sector

IR CAR1Y CAR2Y CAR3Y CAR4Y CAR5Y

Financials _21.27% _-21.70% _-26.26% _-29.06% _-25.56% _-38.00% Basic Materials _75.99% _-23.10% _-30.35% _-31.63% _-46.63% _-59.97% Technology _122.63% _-11.00% _-16.60% _-19.08% _-50.06% _-67.22% Industrials _25.10% _-20.96% _-37.12% _-61.43% _-70.95% _-96.30% Consumer Goods _71.70% _-23.69% _-35.16% _-43.14% _-44.56% _-58.25% Consumer Services _21.36% _7.90% _1.48% _12.43% _15.94% _-31.34% Health Care _33.75% _-25.33% _-29.56% _-22.38% _-60.54% _6.54% Utilities _17.12% _8.64% _-6.04% _-26.98% _-49.02% _-19.58% Oil&Gas _5.13% _-30.41% _-9.36% _-10.86% _10.93% _16.80% Telecommunication _5.12% _-0.50% _-40.10% _{-30.67% -180.01% -159.76%} Notes:

(30)

30 Table 9 presents the long-run performance of IPOs classified by the industry sector. The underpricing is also reported in the table to give a better picture of the performance of every sector. What can be striking in the results reported in table 9 are the CARs for the two longest periods for the Telecommunication sector. It has to be borne in mind that for CAR4Y and CAR5Y only 3 companies are included. Two of them perform very poor against the market and this is the reason of such a strongly negative CARs in this sector.

In case of CAR5Y there are two sectors with positive CAR value. Health Care sector with a market adjusted return of 6.54% and Oil&Gas with 16.8%. Again, these are sectors with low number of observations in the sample. Sectors with high level of underpricing, like Technology, Basic Materials or Consumer Goods in the long-run finish with relatively strong underperformance.

With the use of ANOVA test I check if the industry classification has a significant effect on the initial return and the long-run performance measured with cumulative average abnormal return. The results of ANOVA F-test are presented in the appendix E. I use the confidence level of 10% to reject the null hypothesis that all industry sectors yield the same results in case of underpricing and CARs. In case of initial return and CAR2Y p-value is way above 10% confidence level and the industry classification has no value in explaining those two variables.

IV. Results

It occurs that IPOs on Warsaw Stock Exchange from the period 1994 - 2007 strongly underperform the market in every test period. Moreover, the longer the aftermarket period, the higher is the level of negative market-adjusted return. The negative return of 66.5% against the market index is also a result of using logarithmic returns. The logarithmic returns magnify the underperformance when compared to arithmetic returns. For example, an arithmetic return of 50% is equivalent to a logarithmic return of 40.55%, while an arithmetic return of -50% is equivalent to a logarithmic return of -69.31%.

(31)

31 not the case for this sample of IPOs. The underperformance grows by about 10 percentage points for every added year in the test period with the exception of 5 year period when CAR drops by 18.4 percentage points. The strong underperformance of IPOs proves that the Warsaw Stock Exchange is an inefficient market. The market efficiency implies that the performance of IPOs should be neutral and this is visibly not the case here. It has to be emphasized that the underperformance of the IPOs on WSE is statistically significant even on 1% confidence level for every test period.

The long-run performance of Polish IPOs is presented in the table 6. This table contains CARs measured with alternative benchmarks as well. They are discussed later in the section. It is not clear why Polish IPOs perform so badly in the long term, especially as investing in initial public offerings is a very popular strategy among domestic private investors. One of the most probable explanations, also presented in the section devoted to the literature review in this master thesis, is the investors’ overoptimism about companies conducting public offer. This behavior results in high initial returns of many companies. The average initial return for an IPO in the period of 1994-2007 is 51.4%, which is not common on more mature European stock exchanges, not to mention New York Stock Exchange. Since one of the main principles of Warsaw Stock Exchange (stated in Agenda Warsaw City 2010) is increasing the capitalization of the market, the large number of initial public offerings is executed every year. This strategy brings success as in July 2009 Warsaw Stock Exchange overtook the Wiener Boerse in case of capitalization7. As a result, it is easy and often cheap to proceed the initial public offering in Poland and there is a kind of a ‘window opportunity’ to get cheaper capital (than for example bank loan) for the companies. Among Polish IPOs small companies taking advantage of the overoptimism of investors are the biggest group. They get high initial returns, but in the long-run the majority of them do not fulfill the expectations of investors, thus they perform worse than the market indices. In many cases they end even in the state of bankruptcy just few years after going public as shown in the data section of the master thesis. As presented in the literature review part of this master thesis it is expected that the value weighting scheme of calculating CARs changes the magnitude of the underperformance. I present the value weighted CARs of Warsaw Stock Exchange IPOs from the period of 1994 – 2007 in third column of table 6.

7_{The capitalization of WSE was €81 billion at the end of July 2009, while the capitalization of Wien Stock}

(32)

32 Table 10: The long-run performance of IPOs in 1994 – 2007 against WGI 20 and alternative benchmarks

CAR – Cumulative Average Abnormal Return Index

WGI 20 WGI SWIG80 MSCI Poland MSCI EM Europe

Equally Value Equally Value Equally Value Equally Value Equally Value weighted weighted weighted weighted weighted weighted weighted weighted weighted weighted

CAR1Y -17.00%* -9.12%*** -18.40%* -8.60%*** -14.89%* -1.98% -19.64%* -13.51%** -28.71%* -26.88%* CAR2Y -27.87%* -11.93% -32.54%* -15.49% -35.74%* -25.63%** -32.37%* -16.85%*** -48.36%* -32.27%* CAR3Y -37.41%* -8.84% -45.42%* -19.42%*** -51.19%* -33.38%*** -44.19%* -16.38% -70.11%* -38.40%** CAR4Y -48.06%* -5.80% -59.39%* -18.69% -51.05%* -18.63% -57.51%* -14.68% -93.33%* -51.36%** CAR5Y -66.50%* -23.20% -82.35%* -43.02%*** -66.52%* -28.44% -76.65%* -36.46%** -127.06%* -85.26%** Notes:

* significant at 1% confidence level

** significant at 5% confidence level

(33)

33 The results are very interesting. The value weighting procedure eliminate much of the underperformance of investigated IPOs. Moreover, in contrast to the equally weighted CARs, these returns do not follow any trend. The weakest underperformance can be seen in case of four year CAR. It reaches the value of only -5.80% against WGI 20. This can be interpreted as a minor underperformance, especially if compared with the equally weighted underperformance of -48.06% for the same test period. Unfortunately, value weighting scheme eliminates the significance of all but one CAR.

Only for the longest time period underperformance is relatively high and reaches the value of 23.2%. CAR5Y is the only case when the value weighted CAR is lower than -20% (when only 1 year equally weighted CAR is below 20%). It is difficult to give trustful explanation why Polish IPOs present such a weak performance against the market during the fifth year of their listing which is visible in case of both equally and value weighted CARs.

The value weighted cumulative average abnormal returns imply that the underperformance is mainly common among the smallest stocks in the sample. The differences between the results obtained with two calculation schemes are broad. Based on the results of value weighted CARs it can be stated the underperformance of IPOs on Warsaw Stock Exchange is driven by the group of small companies. In a long-run period small companies often face financial distress, are unable to fulfill their development plans and frequently end up with a stock price lower than the initial one. Simultaneously, there is a group of IPOs with high market value. These IPOs have high weights when the value weighted CARs are calculated. The straightforward conclusion is that the biggest IPOs in the sample do not underperform the market as much as the small companies do.

A number of alternative benchmarks are used to eliminate the threat that certain results are only the result of index utilized as a benchmark. In the table 6 the equally and value weighted CARs obtained with different indices are presented.

(34)

34 of a fact that WGI 20 includes stocks with the highest market capitalization and liquidity. This kind of stocks does not yield such high expected returns as so called growth stocks. The growth stocks are companies whose earnings are expected to grow at an higher than average rate relative to the market. These stocks often yield high returns and I believe they can be the reason of a higher long-run return of WGI in comparison to WGI 20.

As already mentioned, IPOs are generally small size, low book to market value companies. This also holds for the sample of Polish IPOs used in the master thesis. The median market value of 88,74mln PLN is much lower than the average market value of 461,35mln PLN. These numbers indicate that the majority of companies in the sample are small. Because of that, I compare my sample of IPOs with SWIG80 index, which is an index of small companies listed on Warsaw Stock Exchange. Cumulative average abnormal returns measured against SWIG 80 index are the only ones that give evidence of no underperformance of IPOs during a single period in case of equally weighted scheme. Between year three and four CAR increases slightly from -51.19% to -51.05%. This means that in this period IPOs performed minimally better than the index. This stands in contrast with the results obtained against all the other benchmarks used in this master thesis. CAR against WGI 20 drops between year three and four by 10.65 percentage points, against WGI by 13.97 percentage points, against MSCI Poland by 13.32 percentage points. The value weighted methodology eliminate much of the underperformance as in case of WGI 20. In case of SWIG80 CAR1Y reaches the value of only -1.98%, so it can be even argued that the performance of IPOs against this benchmark is almost neutral. In contrast to other alternatives, CAR5Y of SWIG80 is relatively high, only 5.24 percentage points lower than CAR5Y for WGI 20, which is the highest one obtained in this research. The volatility of CARs in case of SWIG80 can be simply explained with the composition of this particular index. It is built up with small companies, whose stock prices and returns have the highest standard deviation among the listed companies.