Evidence for the European Union
MSc BA Finance Thesis
Clemens C. van der Heijden
1Student at the University of Groningen, Faculty of Economics and Business
Abstract
This paper examines the valuation of IPOs using multiples. The empirical evidence shows that the median IPO is significantly overvalued relative to industry matching firms for a sample examining the European Union between 2000 and 2008. This overvaluation ranges from 41% to 93%, depending on the sample of IPO firms and valuation procedure. Besides, for a sample including firms with negative EBITDA, cross-sectional regression tests show a positive relationship between overvaluation and the first-day return. This result is robust for the book-to-market ratio, size and profitability, but does not hold for a sample excluding firms with negative EBITDA. Besides, cross-sectional regression tests show that the most overvalued IPOs earn the lowest return in the long-run, this indicates that valuation using multiples does a reasonable job in distinguishing between under- and overvalued IPOs. However, this relationship is not robust for the book-to-market ratio, size and profitability.
Keywords: IPO, Initial Public Offering, Valuation, Multiples, Underpricing, First-day return, Long-run
return
JEL Classification: G12, G14
First Supervisor: Nanne Brunia Second Supervisor: Lammertjan Dam
April, 2009
1. Introduction
Empirical research documents on the on average positive first-day returns of Initial Public Offerings (IPOs). In scientific literature this phenomenon is called: underpricing. Theoretical literature that attempts to justify underpricing, assumes that the first-day closing price determines the fair value of an IPO. Since IPOs earn an on average positive first-day return, these theoretical models conclude that IPOs are on average undervalued. Purnanandam and Swaminathan (2004) challenge the assumption that the first-day closing price is in line with the fair value of an IPO. They assume that the value of a matching firm is in line with fair value, hence they value IPOs using multiples, and find that the median IPO in the United States between 1980 and 1997 is significantly overvalued by 50%. This overvaluation result is consistent with the empirical research that documents on the long-run underperformance of IPOs, in comparison to matching firms. Purnanandam and Swaminathan’s (2004) overvaluation result and the long-run underperformance of IPOs, are together in line with an efficient market in the long-run, since stock prices ultimately turn to their fair values. In this study, I use the methods of Purnanandam and Swaminathan (2004) to value IPOs using multiples. Besides, I adjust their methods, and apply them to a different geographic region and time period. This study examines the European Union between 2000 and 2008 and is, as far as I know, the first to examine IPO valuation using multiples in the European Union. Furthermore in this study, like Purnanandam and Swaminathan (2004), I relate the valuation of an IPO to subsequent stock returns. From this point comes the professional relevance of this paper. If IPO valuation is related to subsequent stock returns, investors may base trading strategies on this result. Besides, as Purnanandam and Swaminathan (2004) state, if a valuation procedure does a reasonable job in distinguishing between under- and overvalued IPOs, the most overvalued IPOs should earn the lowest returns in the long-run. In this paper I study three research questions.
Second, I estimate the relationship between IPO valuation and market adjusted first-day return using cross-sectional regression tests. In these regression tests I control for the book-to-market ratio, size and profitability. Market adjusted first-day return is defined as the percentage change between the offer price and the first-day closing price adjusted for market return, which I calculate similar to Purnanandam and Swaminathan (2004). The results show no relation between overvaluation and first-day returns for the sample excluding firms with negative EBITDA. On the other hand, the results show a significant positive relation between overvaluation and first-day returns for the sample including firms with negative EBITDA. The last result suggests that stock market investors should pick the most overvalued IPO to earn the largest first-day return.
Third, I estimate the relationship between IPO valuation and market adjusted long-run returns, using cross-sectional regression tests. I estimate these regression tests with and without controlling for the book-to-market ratio, size and profitability. The book-to-market adjusted long-run returns are monthly returns minus the market return, which I cumulate for an 1,2,3,4 and 5 year period after the IPO date. Loughran and Ritter (1995) also cumulate returns for these periods. Since Purnanandam and Swaminathan (2004) only measure long-run returns for a period of 5 years after the IPO date, I expand on them on this point. Purnanandam and Swaminathan (2004) suggest that if their valuation procedure does a reasonable job of distinguishing between under- and overvalued IPOs, the most overvalued IPOs should earn the lowest long-run return. Cross-sectional regression tests without the control variables show that the most overvalued IPOs significantly earn the lowest return in the long-run, which indicates that my valuation procedure does a reasonable job in distinguishing between under- and overvalued IPOs. My cross-sectional regression results show no relation between overvaluation and long-run return with the control variables.
Section 2 presents a literature review, Section 3 presents data and methodology and Section 4 presents the results. Finally, Section 5 concludes.
2. Literature Review
In this section I give an overview of the relevant literature. First, I discuss the research on first-day and long-run returns of IPOs. Thereafter, I discuss the overvaluation result of Purnanandam and Swaminathan (2004) and the relation of this result to the other literature. Ultimately, I consider the relation between IPO valuation and subsequent stock returns.
2.1 Research on first-day and long-run returns of IPOs
between 1990 and 1998. Then, underpricing jumps to 65% during the internet bubble years of 1999 and
2000 and reverts to 12% between 2001 and 2007. 2 Ljungqvist (2004) finds roughly the same result for the
European Union. Namely, IPOs in the European Union are underpriced by 20% between 1990 and 2003. In this context, underpricing is defined as the percentage change between the offer price and the first-day closing price.
Asymmetric information models are theoretical models that try to rationalize why IPOs are underpriced. These models assume that one of the key parties to an IPO transaction knows more than the others. The key parties to an IPO transaction are: the issuing firm, the bank underwriting and marketing the deal, and investors. Rock (1986) presents a model in which investors have superior information, Baron (1982) presents a model in which the bank has superior information and Welch (1989) presents a model in which the issuer is better informed. According to asymmetric information models, this information asymmetry between key parties to an IPO transaction leads to an offer price below fair value. Asymmetric information models assume that the closing price on the first day of trading equals the fair value of an IPO firm, since these models assume an efficient market in the short-run. So, the first-day return is a proxy for the valuation of the IPO. Since IPO firms earn historically an on average positive first-day return, asymmetric information models conclude that IPOs are undervalued.
Regarding the long-run returns of IPO firms, Loughran and Ritter (1995) document on the long-run underperformance of IPO firms. For a sample examining the United States between 1970 and 1990, they find that 44% more money would need to be invested in IPO firms than in comparables to achieve the same wealth after 5 years. Therefore, they conclude that IPOs are misvalued at the time of the offer. Ritter (1991) studies a sample of IPOs in the United States between 1975 and 1982. He finds that investments in IPO firms significantly underperform investments in comparables in the 3 years after going public. This underperformance is 17%. As explanation for the long-run underperformance of IPOs, Ritter (1991) concludes that investors are overoptimistic about the future prospects of an IPO firm at the time of the offer.
2.2 IPO valuation using multiples
Barberis et al. (1998) state that if security prices are not efficient in the short-run, an alternative view of the observed underpricing is that issuers and the bank underwriting the deal underprice the IPO relative to some maximum price they could charge given the demand at the time of the offer. In this framework, as Purnanandam and Swaminathan (2004) state, value is not identical to the first-day market price, but presumably to some notion of long-run fair value. Keeping this in mind, Purnanandam and Swaminathan (2004) value IPOs relative to their fair values. They assume that the value of the matching firm is the fair
value of an IPO and value IPOs using the market price multiples of matching firms. Purnanandam and
Swaminathan (2004) find for a sample of IPOs in the United States that the median IPO is significantly
2
overvalued between 1980 and 1997. In a study in line with Purnanandam and Swaminathan (2004), Pukthuanthong-Le and Varaiya (2007) include firms with negative EBITDA in their sample, and find for a sample of IPOs in the United States that the median IPO between 1990 and 2000 is also significantly overvalued. Both studies use different financial metrics in their valuation procedure. Table 1 shows the IPO overvaluation results of Purnanandam and Swaminathan (2004) and Pukthuanthong-Le and Varaiya (2007) by financial metric.
Financial metric Overvaluation Significant #
Sales 54% yes 2288
EBITDA 49% yes 2288
Earnings 54% yes 1843
Financial metric Overvaluation Significant #
Sales 43% yes 3087
Gross cost margin 36% yes 3087
Forecasted current year earnings 62% yes 1608
Forecasted next year earnings 47% yes 1510
Table 1
Purna nandam and Swa minatha n (2004) and Pukthuanthong-Le and Varaiya (2007) value IPOs based on multiples of industry matching firms. This table the reports overvaluation results of IPO firms by the financial metric used in the va luation procedure. Significant indicates whether the results are significant or not and # indicates the sample size.
Panel A: Purnanandam and Swaminathan (2004)
Panel B: Pukthuanthong-Le and Varaiya (2007) IPO valuation using multiples
Both studies choose for each IPO firm one matching firm in the same industry with comparable operating characteristics. When Purnanandam and Swaminathan (2004) add analyst earnings growth forecasts as an additional matching criterion to the industry and operating characteristics of the firms, the overvaluation drops to 33%, with a sample size of 1520. When they match only on this growth forecasts, the overvaluation drops to 14%, with a sample size of 1652. Unfortunately, Purnanandam and Swaminathan (2004) do not disclose information about the financial metric and significance on this point.
2.3 IPO valuation and the relation to subsequent stock returns
Purnanandam and Swaminathan (2004) find a significant positive relation between overvaluation and first-day return. This relationship is robust for variables that capture the cross-sectional variation in stock returns, these variables are: the book-to-market ratio, growth, accruals, sales and EBITDA. Table 2 shows the results of Purnanandam and Swaminathan’s (2004) cross-sectional regression model concerning IPO valuation and first-day returns. In this context, asymmetric information models predict a negative relation between overvaluation and first-day returns. Because, in a market which is efficient in the short run, the most undervalued IPOs should earn the largest first-day return.
Furthermore, Purnanandam and Swaminathan (2004) find a significant negative relationship between overvaluation and long-run returns. They control this relationship for the same variables as the relationship between valuation and first-day return. Table 2 shows the results of Purnanandam and Swaminathan’s (2004) cross-sectional regression model concerning IPO valuation and long-run returns.
Ln (Price-to-Value ratio) 2.36 2.35 -0.38 -0.50 5.49 4.98 -2.69 -3.53 Ln (Book-to-Market ratio) -5.05 -6.43 -0.35 -0.68 -2.96 -4.13 -1.27 -1.71 Ln (Growth) --- 5.53 --- -1.60 --- 1.89 --- -2.43 Accruals -0.60 -1.31 -0.24 -0.36 -0.22 -0.48 -0.33 -0.47 Ln (Sales) -1.38 -1.59 0.19 -0.07 -4.28 -4.17 1.96 -0.62 EBITDA Margin 0.06 0.07 -0.01 0.00 2.06 2.27 -1.16 0.37 # 2057 1686 2057 1686
Purna nandam and Swa minatha n (2004) relate IPO valuation to subsequent stock returns. This table reports their cross-sectional regression results for IPO valua tion related to first-day return as well as long-run returns. Specifically, this table reports coefficients and t-statistics for every independent variable in their regression model. Coefficients a re in percent and t-statistics are White (1980) heteroskedasticity consistent t-statistics and are in italic. Ln indicates the natural log of the varia ble. Price-to-Value ratio refers to price of the IPO relative to fair value, computed using multiples. Growth refers to the analyst growth over the next five years. Ac cruals is the ratio of accruals to total assets. EBITDA margin is the ratio of EBITDA to Sales. # indicates the sample size. The regression model is examined with and without applying growth as independent variable.
Table 2
Cross-Sectional Regression IPO valuation and subsequent stock returns by Purnanandam and Swaminathan (2004) Independent Variable
Coefficients and t-statistics
First-day return Long-run return
adjusted for the market return. Since the most overvalued IPOs earn the lowest return in the long-run, Purnanandam and Swaminathan (2004) conclude that their valuation procedure does a reasonable job in distinguishing between under- and overvalued IPOs.
3. Data & Methodology
I divide this section into four parts. First, I discuss the documentation of the data. In which I, among other, discuss my IPO valuation procedure using multiples as well as Purnanandam and Swaminathan’s (2004) procedure. Second, I give the descriptive statistics. Third, I consider the methods on selecting matching firms and on relating IPO valuation to subsequent stock returns. In which, I also discuss the methods by Purnanandam and Swaminathan (2004). Fourth, I consider the tests of the three hypotheses in this paper.
3.1 Data documentation
In this chapter, I first discuss the selection of the sample, thereafter I consider the data transformation. Then, I consider IPO valuation using multiples. Ultimately, I provide general sample information.
3.1.1 Sample selection
Symbol Definition Currency Database IPO Matching
Pop IPO offer price home Zephyr x
Sales Sales in the fiscal year prior to the IPO € Amadeus x x
EBITDA EBITDA in the fiscal year prior to the IPO € Amadeus x x
Pcp Closing share price at the IPO offer date € Datastream x x
SHno Shares outstanding at the IPO offer date Datastream x x
Rt Monthly Total Return Index after the IPO, till august 2008 Datastream x
BVE Book value of the equity at the end of the fiscal year of the IPO € Amadeus x
Debt Debt at the end of the fiscal year of the IPO € Amadeus x x
Rm Return on MSCI European Index at the offer date Datastream x
Rm,t
Monthly Total Return Index after the IPO on the MSCI European
Index, till august 2008 Datastream x
ERya
Yearly average exchange rate in the year prior to the IPO, with Euro as base currency and the country in which the company is
listed as term currency Datastream x x
BvDEP ID number Zephyr x
ISIN number Zephyr x x
NACE Rev. 1.1 code Amadeus x x
Table 3
Data requirements for IPO and matching firms
This table reports the symbol and description of the raw da ta. It also reports the currency, a nd the database from which the data is obta ined. Furthermore, it reports if the data is obtained for IPO and/or matching firms. H ome currency refers to the currency in which the company is listed.
You can find the sample selection criteria in appendix B and details about the methods on selecting matching firms in paragraph 3.3.1.
3.1.2 Transformed Data
First, I convert all variables given in Euro to the currency in which the company is listed. Which I compute
as the exchange rate (ERya) times the specific variable. Concerning the accounting data of the IPO and
matching firms, I compute the market capitalization (MC), EBITDA margin (EM) and book-to-market ratio (BM) similar to Purnanandam and Swaminathan (2004). I compute the gross cost margin (GCM) similar to Pukthuanthong-Le and Varaiya (2007). You can find the calculation of MC, EM, BM and GCM in Table 4.
Concerning the stock returns of the IPO firm, I compute the first-day return (ARfd) as market adjusted
first-day return, which is similar to Purnanandam and Swaminathan (2004). I use the MSCI European Index as market index, since my sample is subjected to the European Union. Furthermore, for every IPO, I compute monthly abnormal returns and cumulative abnormal returns similar to Barber and Lyon (1997). I compute
the monthly abnormal return (ARt) as the market adjusted return in month t. Cumulating across τ periods,
yields a cumulative abnormal return (CARt). Again, I use the MSCI European Index as market index. I
cumulative abnormal returns for an 1,2,3,4 and 5 year period after the IPO. You can find the calculation of
ARfd , ARt and CARt in table 4.
Formula IPO Matching
x x x x x x x x x
This table reports the computation of the transformed da ta. Furthermore, it reports if the data is computed for IPO and/or matching firms.
Table 4 Transformed Data t t t AR CAR
∑
= = τ 1 m of of cp f d R P P P AR − − = t m t t R R AR = − , no op ipo P SH MC = * no cp mat ching P SH MC = * Sales EBITDA EM = EBITDA Sales Sales GCM − = no cp SH P BVE BM * =3.1.3 IPO valuation using multiples
In this paragraph, I discuss the IPO valuation procedure using multiples. Since my valuation procedure is based on Purnanandam and Swaminathan (2004), I first discuss the procedure that they use, followed by relevant alternatives. Ultimately, I discuss my own methods.
Purnanandam and Swaminathan (2004) study the price of IPOs relative to their fair values. They estimate this fair value using the market price of one matching firm. In which, the price of the firm is the market capitalization. Dividing this price by a financial metric gives the price multiple. Dividing the IPO firms offer price multiple by the matching firms market price multiple gives the valuation of the IPO firm. Purnanandam and Swaminathan (2004) use the closing share price of the matching firm at the IPO date as market price and compute the price multiples for IPO and matching firms by using prior year sales, prior year EBITDA and prior year earnings as financial metrics.
Pukthuanthong-Le and Varaiya (2007) include IPO firms and matching firms with negative EBITDA in their sample. Since price multiples with a negative financial metric make no economic sense, Pukthuanthong-Le and Varaiya (2007) use prior year gross cost margin instead of prior year EBITDA as financial metric in their valuation procedure. Gross cost margin is defined as sales minus EBITDA, divided by sales.
In Purnanandam and Swaminathan’s (2004) and Pukthuanthong-Le and Varaiya’s (2007) method to compute the value of an IPO, market capitalization is used as price of the firm, which is equity value. However, according to Koller et al. (2005) enterprise value should be used instead of equity value in valuation using multiples, since differences in leverage between the IPO and matching firm can affect valuation results. The problem with leverage is that the financial metrics sales, EBITDA and the gross cost margin cover the value of the total enterprise, rather than the equity part alone. For instance, if an IPO firm is financed with less equity and more debt in comparison to its matching firm, valuation is understated when equity value is used. So, I agree with Koller et al. (2005) that enterprise value should be used as price of the firm, which is market capitalization plus debt. Zheng (2006) also suggest, in reaction on Purnanandam and Swaminathan (2004), to use enterprise multiples with valuation using multiples.
My own valuation procedure is as follows. For each IPO firm, I compute price-to-value (P/V) ratios, which give the valuation of the IPO. To compute the P/V ratios, I divide the IPO firms offer price multiple by the matching firms market price multiple. I compute both price multiples as the price of the specific firm divided by a financial metric. I compute six different P/V ratios, by using two different methods to compute the price of the firm in combination with one of three financial metrics.
I use two methods to compute the price of the firm. In the first method I use equity multiples, which is similar to Purnanandam and Swaminathan (2004). In this case the price of the IPO and matching firm is the market capitalization. In the second method I adjust Purnanandam and Swaminathan’s (2004) valuation method by taking leverage into account. So, I use enterprise multiples in this method. In this case the price of the IPO as well as matching firm is the market capitalization plus debt. Like Purnanandam and Swaminathan (2004), I compute the market capitalization for IPO firms as the offer price times the number of shares outstanding at the offer date. And for matching firms as the closing share price at the IPO date times the number of shares outstanding at this date. Furthermore, I use debt at the end of the fiscal year of the IPO, which is similar to Zheng (2006).
measures operating cash flow and is only little subject to accounting distortions. I compute the price multiples for IPO and matching firms as follows:
Metric Financial Debt MC FM P ipo ipo ) ( ) / ( = + and, Metric Financial Debt MC FM P matching matching ) ( ) / ( = +
In which, (P/FM)ipo is the price multiple of the IPO firm and (P/FM)matching is the price multiple of the
matching firm. Furthermore, the financial metric is Sales, EBITDA or GCM. I compute the P/V ratio in the following way: matching ipo FM P FM P V P ) / ( ) / ( / =
The valuation procedure of IPOs is challenging, since limited information about the IPO firm is publicly available at the IPO date. Generally, limited historical accounting data and information about the prospects of the IPO firm are available. Taking this into account, due to unavailability of data, I am not able to use earnings and forecasted earnings as financial metric in the valuation procedure, like Purnanandam and Swaminathan (2004) and Pukthuanthong-Le and Varaiya (2007), respectively. In line with the latter, Koller et al. (2005) recommend to use forward-looking estimates in multiple valuation, since the value of a company equals the present value of future cash flow.
3.1.4 Sample information
Appendix B shows the number of firms by year, country and industry. Table B.1 and B.3 show the same firm numbers between IPO and matching firms for every specific year and industry. Table B.2 shows approximately the same firm numbers between IPO and matching firms for every specific country.
3.2 Descriptive statistics
variable Mean Median Mean Median Mean Median Mean Median Market Capitalization 720,002 95,200 795,835 44,452 576,110 76,342 624,435 30,690 Sales 513,839 39,431 443,897 42,763 375,390 19,542 323,818 21,253 EBITDA 84,553 5,747 73,110 4,653 59,894 2,533 51,837 2,088 Gross Cost 429,286 32,582 370,787 34,191 315,651 17,163 271,981 19,352 Debt 936,464 31,305 610,441 30,000 682,239 16,103 443,145 15,335 Bookequity 279,059 34,056 --- --- 225,100 27,845 ---
---This table reports firm characteristics for the samples excluding a nd including firms with negative EBITDA. It compares charac teristic s between IPO and mat ching firms. Numbers represent Euro thousands.
IPO firms Matching firms IPO firms Matching firms
Table 5 Firm characteristics
Sample excluding firms with negative EBITDA Sample including firms with negative EBITDA
Table 6 shows descriptive statistics on P/V ratios. Given the values on Jarque Bera, the P/V ratios does not have a normal probability distribution.
Multiple Metric Mean Median Max. Min. Std.Dev. Skewness Kurtosis Jarque-Bera
Equity Sales 39.00 1.70 14,165.02 0.00 612.17 22.46 516.80 6,107.14
Enterprise Sales 3.78 1.41 292.78 0.00 15.55 14.33 240.67 1,315.67
Equity EBITDA 63.17 1.58 15,347.66 0.00 831.40 16.50 281.01 1,799.45
Enterprise EBITDA 9.28 1.42 1,120.34 0.00 66.73 14.23 220.55 1,105.17
Multiple Metric Mean Median Max. Min. Std.Dev. Skewness Kurtosis Jarque-Bera
Equity Sales 8,029.53 1.92 6,148,941.00 0.00 221,304.10 27.73 769.98 19,021.45
Enterprise Sales 42.06 1.57 23,239.34 0.00 843.35 27.05 743.53 17,733.90
Equity GCM 9,400.33 1.93 7,198,517.00 0.00 259,079.10 27.73 769.98 19,021.41
Enterprise GCM 55.00 1.46 27,206.11 0.00 1,010.11 25.54 679.90 14,822.65
Thi s table reports descripti ve statist ics on P/V ratios for the samples including a nd exc luding fi rms with negative EBITD A. M ultiple indica te whether the P/V ratio involves equi ty or enterprise multiples. Metric indicates if the P/V ratio involves sales, EBITDA or GCM as financial metric. # indicates the sample size and Jarqua-Bera numbers are in thousands.
Sample excluding firms with negative EBITDA (# = 551)
Sample including firms with negative EBITDA (# = 772) Table 6
Descriptive statistics on Price-to-Value (P/V) ratios
P/V ratio # Mean Median p-value Max. Min. Std.Dev. Skewness Kurtosis Jarque-Bera ARfd 551 11.70% 5.13% 0.0000 257.08% -30.83% 0.26 4.77 35.02 25626.40 CARt t=12 533 -5.13% -3.28% 0.0196 243.11% -176.72% 0.53 0.20 4.86 80.46 CARt t=24 411 -13.10% -4.28% 0.0034 249.08% -238.30% 0.74 -0.25 3.48 8.16 CARt t=36 266 -9.23% 2.93% 0.2813 219.05% -315.19% 0.90 -0.42 3.51 10.71 CARt t=48 202 3.69% 6.40% 0.3535 291.14% -262.84% 0.97 -0.17 3.43 2.53 CARt t=60 171 6.83% 3.30% 0.5627 475.09% -283.12% 1.13 0.37 4.29 15.73
# Mean Median p-value Max. Min. Std.Dev. Skewness Kurtosis Jarque-Bera
ARfd 772 12.55% 5.41% 0.0000 257.08% -40.00% 0.29 4.50 30.13 26283.56 CARt t=12 741 -11.33% -8.88% 0.0000 376.78% -279.11% 0.59 0.22 6.99 497.18 CARt t=24 579 -24.27% -16.31% 0.0000 249.08% -286.49% 0.80 -0.29 3.40 12.00 CARt t=36 385 -22.62% -14.59% 0.0003 219.05% -331.27% 1.00 -0.42 3.28 12.37 CARt t=48 290 -1.94% -0.08% 0.7077 602.25% -399.68% 1.11 0.54 6.73 181.79 CARt t=60 233 3.19% 4.53% 0.7450 475.09% -452.81% 1.21 0.09 4.58 24.51
Thi s table reports descriptive st atist ics on stock returns for the samples excluding and including firms wit h negative EBITDA. ARfd indicates the market adjusted first -day return. CAR indi cates the market a dj usted l ong-run return and t indicates the month a fter the IPO dat e. The value i s the p-va lue from the Wil coxson signed-rank test, performed in Eviews, testing the hypothesis if the median is equal to 0. # indicates the sa mple size
Sample excluding firms with negative EBITDA
Sample including firms with negative EBITDA Table 7
Descriptive statistics on stock returns of IPO firms
Further descriptive statistics are in appendix C. Table C.1 shows that the Spearman rank-order correlation coefficients between P/V ratios are all above 0.7. I use 0.7 as reference point and standard rule of thumb whether two variables are correlated or not. So, the P/V ratios are highly correlated between the different financial metrics that I use in the analysis. As stated by Purnanandam and Swaminathan (2004), this is encouraging, since the valuation based on multiples is not to far apart. Tables C.2 and C.3 show correlation between the independent variables for the regression in this paper. Given the Spearman rank-order correlation coefficients, there is no multicollinearity between the variables, since the coefficients are all below 0.7. Tables C.4 and C.5 present Jarque-Bera statistics for the variable in regression models the regression models in this paper, which are not already mentioned in the descriptive statistics. Given the values on Jarque-Bera, the variables do not have a normal probability distribution.
3.3 Methods
In this chapter, I discuss the process of selecting matching firms as well as the methods on examining the relationship between IPO valuation and subsequent stock returns. The methods of this study are based on those of Purnanandam and Swaminathan (2004). Therefore, I first consider their methods in each paragraph, eventually followed by relevant alternatives. Ultimately, I consider my own methods.
3.3.1 Selecting matching firms
EBITDA divided by sales. In the process of selecting matching firms, Purnanandam and Swaminathan (2004) first group possible matching firms in each industry into three portfolios based on past sales and then each sales portfolio into three portfolios based on past EBITDA profit margin. Second, each IPO is then matched to an appropriate industry-sales-EBITDA margin portfolio. Third, from the industry-sales-EBITDA margin portfolio the firm closest in sales to the IPO firm is selected as matching firm.
In the process of selecting matching firms, Pukthuanthong-Le and Varaiya (2007) do not form portfolios like Purnanandam and Swaminathan (2004). They select a matching firm in the same industry with sales between 70% and 130% of the IPO firms revenue. The firm closest in gross cost margin to the IPO firm is selected as matching firm.
My own process of selecting matching firms is as follows. Similar to Purnanandam and Swaminathan (2004), I select for every IPO firm one matching firm in the same industry with comparable sales and EBITDA profit margin. Like the IPO firms in the sample, these matching firms should be located in the European Union. I match on industry, since firms in the same industry have, according to Purnanandam and Swaminathan (2004), similar operating risks, profitability and growth. To classify firms by industry, I use NACE Rev. 1.1 codes. NACE is a standard European Union system to classify industry groups. It is the European Union equivalent of the United States SIC system, which Purnanandam and Swaminathan (2004) use. I classify the IPO and matching firms into 62 industry groups, based on the first two digits of the four digit NACE Rev. 1.1 code. Next, I match on sales, since sales is an measure of size. Similar to Pukthuanthong-Le and Varaiya (2007), I consider firms with sales in the range of 70% to 130% of the IPO firms sales as matching firms. In this step Purnanandam and Swaminathan (2004) construct portfolios. I do not follow Purnanandam and Swaminathan (2004), since data limitations in the sample of matching firms deter me from constructing portfolios. If no firms are within the 70% to 130% range, I choose the firm closest in sales to the IPO firm as matching firm. Finally, I match on EBITDA profit margin to control for differences in profitability across firms. Furthermore, for IPO firms with positive EBITDA, I select a matching firm with positive EBITDA. This prevents that IPO firms with positive EBITDA have different matching firms between the two samples in this paper. Similar to Purnanandam and Swaminathan (2004), I choose the firm which has the closest EBITDA profit margin to the IPO firm as appropriate matching firm. The data requirements for possible matching firms are in Table 3. Similar to Purnanandam and Swaminathan (2004), the matching firm shouldn’t have gone public in the 2 fiscal years prior to the IPO.
3.3.2 Relationship between IPO valuation and market adjusted first-day return
Purnanandam and Swaminathan (2004) study the relationship between IPO valuation and market adjusted first-day return of the IPO, using cross-sectional regression tests. In the regression tests they control for the book-to-market ratio, accruals, size, growth and profitability. Furthermore, they compute the first-day return adjusted for the market return.
My own methods are as follows. In the cross-sectional regression model between IPO valuation and market adjusted first-day return, I control for the to-market ratio, size and profitability. I control for the book-to-market ratio and size, since Fama and French (1992) find that these variables capture cross-sectional variation in stock-returns in average stock returns. I compute these variables similar to Purnanandam and Swaminathan (2004). Like, Purnanandam and Swaminathan (2004), I use sales as a proxy for size. I control for profitability, since Purnanandam and Swaminathan (2004) find that profitability is related to the cross-section of stock returns. Moreover, I use the EBITDA margin as control for profitability when I exclude firms with negative EBITDA and the gross cost margin when I include firms with negative EBITDA in the sample. Purnanandam and Swaminathan (2004) and Pukthuanthong-Le and Varaiya (2007) indicate that the EBITDA margin and the gross cost margin control for profitability across firms, respectively. Furthermore, I calculate both variables similar to Purnanandam and Swaminathan (2004) and Pukthuanthong-Le and Varaiya (2007), respectively. Besides, I compute the first-day return similar to Purnanandam and Swaminathan (2004).
Jarque-Bera statistics from the variables in the regression model indicate that the data does not have a normal probability distribution. However, for the regression analysis and the residuals from this analysis, I assume that the data has an approximately normal probability distribution. Namely, if the sample is larger than 30, the central limit theorem assumes, as stated by Anderson et al. (2002), that the data has an approximately normal probability distribution. The sample sizes in this paper are 551 and 772.
Addressing the problem of limited information about IPO firm at the time of the offer, due to unavailability of data and unlike Purnanandam and Swaminathan (2004) I do not control for accruals and growth. Purnanandam and Swaminathan (2004) and Teoh et al. (1998) find that growth and accruals are related to the cross-section of IPO firms stock returns, respectively.
3.3.3 Relationship between IPO valuation and market adjusted long-run returns
regression tests they control for the same variables as in their regression model regarding IPO valuation and first-day returns.
Purnanandam and Swaminathan (2004) use the Fama and French (1993) three-factor model to compute long-run returns. The Fama and French (1993) three-factor model assumes that the expected return can be explained by excess market return, a book-to-market factor and a size factor. I criticize Purnanandam and Swaminathan (2004) on correcting for these factors in computing long-run returns. Since, they already correct for size and them in the cross-sectional regression model concerning the relation between overvaluation and long-run returns.
My own methods are as follows. Unlike Purnanandam and Swaminathan (2004), I do not control for the book-to-market ratio and size when computing long-term returns, since I already control for them in the regression model. I compute cumulative abnormal returns adjusted for the market return, which is similar to Barber and Lyon (1997). I compute the returns for an 1,2,3,4 and 5 year period after the IPO, excluding the return on the first-day. By computing returns for 1 till 4 years after the IPO, I expand on Purnanandam and Swaminathan (2004), since they only compute returns for a 5 year period. The independent variables for my regression model are defined exactly as in my regression model regarding IPO valuation and first-day return. Besides, I examine the relation between IPO valuation and long-run returns with and without control variables. Furthermore, like the regression model regarding IPO valuation and first-day return, I assume that the data can be approximated by a normal probability distribution.
If a valuation procedure does a reasonable job in distinguishing between under- and overvalued IPOs, as Purnanandam and Swaminathan state, the most overvalued IPOs should earn the lowest return. Since in an efficient market stock prices return to their fair values. To test this, I examine the relationship between overvaluation and long-run returns without applying the control variables that explain the cross-section in stock-returns, which is similar to Liu et al. (2002). They test if valuation using multiples can predict subsequent stock returns.
3.4 Hypothesis testing
3.4.1 Testing whether IPOs are overvalued
The P/V ratio indicates the valuation of the IPO. I test the following hypothesis for both samples and every P/V ratio:
H10: IPOs are valued at their fair value
I do not reject H10 if the median P/V ratio is equal to one. I test this hypothesis with the Wilcoxson signed-rank test, performed in Eviews. According to Sheskin (1997), this test is a non-parametric procedure employed in hypothesis testing, involving a single sample, in order to determine whether a sample is derived for a population with a certain median.
3.4.2 Testing the relationship between IPO valuation and first-day return
I use cross-sectional regression tests between P/V ratios and first-day returns to test the following hypothesis for both samples and every P/V ratio:
H20: IPO valuation is unrelated to first-day return
H2a: IPO valuation is related to first-day return
I test H2 for the sample excluding firms with negative EBITDA with the following regression model:
i i i i i i i fd LnPV LnBM LnSales LnEM AR , =α +γ1* +γ2* +γ3* +γ4* +ε (1)
I test H2 for the sample including firms with negative EBITDA with the following regression model:
i i i i i i i fd LnPV LnBM LnSales LnGCM AR , =α +γ1* +γ2* +γ3* +γ4* +ε (2)
In which, the index i denotes the IPO firm. The dependent variable ARfd, is the market adjusted first-day
return. Furthermore, I take the natural log (Ln) for all the independent variables. In which, LnPV refers to the six P/V ratios, LnBM to the book-to-market ratio, LnSales to sales, LnEM to the EBITDA margin and
LnGCM to the gross cost margin.
I estimate the regression models (1) and (2) using cross-sectional approach, which is similar to
Purnanandam and Swaminathan (2004). I reject H20 if the coefficient for LnPV has a White (1980)
heteroskedasticity consistent t-statistic that is significant at 5%. I use White (1980) heteroskedasticity consistent t-statistics, since my residuals are heteroskedastic.
3.4.3 Testing the relationship between IPO valuation and market adjusted long-run returns
H30: IPO valuation is unrelated to long-run returns
H3a: IPO valuation is related to long-run returns
I test H3 for the sample excluding firms with negative EBITDA with the following regression model:
i i i i i i t LnPV LnBM LnSales LnEM CAR, =α+γ1* +γ2* +γ3* +γ4* +ε (3)
I test H3 for the sample including firms with negative EBITDA with the following regression model:
i i i i i i t LnPV LnBM LnSales LnGCM CAR, =α+γ1* +γ2* +γ3* +γ4* +ε (4)
In regression models (3) and (4) index i denotes the IPO firm and t is the period in months after the IPO date
plus one day. CARt is the market adjusted long-run return for t months, in which t = 12, 24, 36, 48 or 60
months. The independent variables in the regression models (3) and (4) are defined exactly as in regression models (1) and (2).
I estimate the regression models (3) and (4) using cross-sectional approach, which is similar to
Purnanandam and Swaminathan (2004). I reject H30 if the coefficient for LnPV has a White (1980)
heteroskedasticity consistent t-statistic that is significant at 5%. I use White (1980) heteroskedasticity consistent t-statistics, since my residuals are heteroskedastic. Furthermore, I compute the regression model also without the control variables.
4. Results
In this section, I discuss the results of the three research questions. Furthermore, I relate these results to the literature discussed in this paper. Besides, I also consider the results when excluding outliers from the sample.
4.1 IPO valuation
Tables 8 and 9 present the median, 25th percentile and 75th percentile of the P/V ratios. These tables also
present the p-values from the Wilcoxson signed-rank test, for testing the hypothesis that the median is equal to one. Table 8 presents results for the sample excluding firm with negative EBITDA and Table 9 for the sample including these firms. The empirical evidence shows that the median IPO is significantly overvalued
regardless of the sample and P/V ratio, therefore I do not reject H1a. The overvaluation ranges from 41% to
Multiple Metric med. 25th 75th p-val.
Equity Sales 1.70 0.75 4.57 0.0000
Enterprise Sales 1.41 0.75 2.86 0.0000
Equity EBITDA 1.58 0.68 3.95 0.0000
Enterprise EBITDA 1.42 0.72 3.21 0.0000
IPO valuation using multiples for the sample excluding firms with negative EBITDA (# = 551) Table 8
This table reports the distribution of the pric e-to-value (P/V) ratios for the sample exc luding firms with negative EBITDA. Multiple indicates whether the P/V ratio is computed using equity or enterprise multiples. Metric indicates which financial metric is used to compute the P/V ratio. The table presents median, 25th percentile and 50th percentile for each valuation procedure. It a lso provides Wilcoxson signed-rank test p-values if the median is equal to 0. # indicates the sample size.
Multiple Metric med. 25th 75th p-val.
Equity Sales 1.92 0.73 6.33 0.0000
Enterprise Sales 1.57 0.72 3.90 0.0000
Equity GCM 1.93 0.66 5.56 0.0000
Enterprise GCM 1.46 0.67 3.57 0.0000
This table reports the distribution of the pric e-to-value ratios for the sample including firms with negative EBITDA. GCM refers to the gross cost margin. The further description of this table equal to the description of table 8.
IPO valuation using multiples for the sample including firms with negative EBITDA (# = 772) Table 9
Appendix D presents the median, 25th percentile and 75th percentile of the P/V ratios for both samples by
year, country and industry. Looking at different countries and industries in the sample, the significant overvaluation result is, for my sample, robust for the country and industry. This holds for both samples and every P/V ratio. However, looking at the different years in the sample, the significant overvaluation result does not hold for all separate years in the sample. Because in certain years, the median of not every P/V ratio is significantly different from one.
The significant overvaluation result is in line with Purnanandam and Swaminathan (2004) and Pukthuanthong-Le and Varaiya (2007), since they find that the median IPO is significantly overvalued by around 50% and 40%, respectively. Besides, my significant overvaluation result is inconsistent with the predictions of the asymmetric information models, since these models predict that the median IPO is undervalued by the value of the first-day return. The overvaluation result is together with the long-run underperformance of IPOs in line with an efficient market in the long-run. The overvaluation result is also in line with the investor overoptimism about future prospects of IPO firms at the time of the offer, as documented by Ritter (1991).
4.2 Relationship between IPO valuation and market adjusted first-day return
profitability. Consequently, I reject H2a for the sample excluding firms with negative EBITDA. Table 12 presents coefficients and White (1980) heteroskedasticity consistent t-statistics for the independent variables from regression model (2). This regression model is subjected to the sample including firms with negative EBITDA. According to White (1980) heteroskedasticity consistent t-statistics on LnPV, overvaluation is significantly positive related to the first-day return, which is robust for including the control variables. So, I
do not reject H2a for the sample including firms with negative EBITDA. Looking at the control variables
from regression model (1) and (2), the first-day return is significantly negative related to LnBM and significantly positive to LnSales. LnEM from regression model (1) is insignificant in the regression model, and LnGCM from regression model (2) is significantly positive related to the first-day return.
Equity multiple Enterprise multiple Equity multiple Enterprise multiple
LnPV 0.06 0.67 -0.23 0.33 0.11 0.82 -0.45 0.49 LnBM -4.70 -4.52 -4.85 -4.64 -2.86 * -2.83 * -2.94 * -2.90 * LnSales -1.89 -1.86 -1.90 -1.89 -4.03 * -3.98 * -4.06 * -4.04 * LnEM 1.69 1.57 1.62 1.80 1.09 1.02 1.08 1.19
Cross-Sectional Regression IPO valuation and first-day return for sample excluding firms with negative EBITDA (# = 551) Table 10
Independent Variable
Coefficients & t-statistics
Metric: Sales Metric: EBITDA
This table reports results from the following regressi on model:
For each independent variable the table report s coeffici ents and t-statistics. Coefficients are in percent, t-stati stics are White (1980) heteroskedasticity consistent t-statistics and are in itali c. ARfd refers to the market adjusted first-day return. Ln indicates the natural log of the variable. PV refers to the price-to-value (P/V ) ratio and BM to the book-to-ma rket ratio. Sales and EBITDA refers to the financial metric used to compute the P/V ratio. Furthermore, the table report s whether the P/V ratio is computed using equity or enterprise multiples. ** indicat es significance at 1% level and * indicates signifi cance a t 5% level. # indicates the sample size.
i i i i i i i fd LnPV LnBM LnSales LnEM AR , =α +γ1* +γ2* +γ3* +γ4* +ε
Equity multiple Enterprise multiple Equity multiple Enterprise multiple
LnPV 1.20 2.36 1.10 2.06 2.54 ** 3.25 * 2.45 ** 3.12 * LnBM -4.76 -4.75 -4.81 -4.84 -3.83 * -3.94 * -3.88 * -4.00 * LnSales -1.26 -1.17 -1.43 -1.50 -3.09 * -2.95 * -3.40 * -3.53 * LnGCM 3.37 3.69 3.75 4.37 2.83 * 3.02 * 3.13 * 3.49 * Table 11
Cross-Sectioanl Regression IPO valuation and first-day return for sample including firms with negative EBITDA (# = 772)
This table reports results from the following regressi on model:
Independent Variable
Metric: Sales Metric: GCM
GCM refers to the Gross Cost Margin. The further description of this table is equal to descri ption of t able 10
Coefficients & t-statistics
Purnanandam and Swaminathan (2004) find a significantly positive relation between overvaluation and first-day return. This is inconsistent with my results, since I find no relation between overvaluation and first-first-day return for the sample excluding firms with negative EBITDA. However, for the sample including firms with negative EBITDA, I find a positive relation between overvaluation and the first-day return. The asymmetric information models predict that the most overvalued IPOs should earn the lowest first-day return, since these models assume an efficient market in the short-run. This is inconsistent with my findings for both samples. My empirical results on the sample including firms with negative EBITDA, imply for stock market investors that they can base active trading strategies on this result to produce superior returns. Namely, the most overvalued IPOs earn the largest first-day return. The positive relation between overvaluation and first-day returns, suggest that the overoptimism by investors about future prospects of the IPO firm at the time of the offer, as documented by Ritter (1991), is still present at the first trading day. So, markets show no efficiency in the short-run.
4.3 Relationship between IPO valuation and market adjusted long-run returns
Tables 12 and 13 present coefficients and White (1980) heteroskedasticity consistent t-statistics for LnPV from regression models (3) and (4) without using the control variables. Regression model (3) is subjected to the sample excluding firms with negative EBITDA and regression model (4) is subjected to the sample including firms with negative EBITDA. According to the White (1980) heteroskedasticity consistent t-statistics on LnPV, the relationship between overvaluation and long-run returns is significantly negative in 29 out of 40 cases. This significantly negative relationship is spread among the samples, P/V ratios and time periods. However, this relationship is not robust for the book-to-market ratio, size and profitability. Tables D.7 and D.8 present coefficients and White (1980) heteroskedasticity consistent t-statistics from regression model (3) and (4) with the control variables. Given the insignificant relation between overvaluation and
long-run returns with using the control variables, I reject H3a. Looking at the control variables of regression
Multiple Metric t = 12 t = 24 t = 36 t = 48 t = 60 Equity Sales -2.71 -5.32 -7.73 -7.46 -8.92 -2.05 * -2.43 * -2.84 ** -2.15 * -2.00 * Enterprise Sales -2.25 -5.35 -10.92 -10.18 -9.07 -1.24 -1.58 -2.71 ** -2.03 * -1.35 Equity EBITDA -3.71 -7.59 -7.14 -6.45 -7.78 -2.87 ** -3.88 ** -2.72 ** -1.86 -1.77 Enterprise EBITDA -3.20 -8.27 -7.10 -6.17 -5.80 -1.97 * -3.34 ** -2.24 ** -1.42 -1.06 # 533 411 266 202 171
Coefficients & t-statistics for LnPV
This table reports results from the following regressi on model:
For the independent variable, this table reports coefficients a nd t-statist ics. The independent variable LnPV refers to t he natural log of the price-t o-va lue (P/V) ratio. CAR indi cates the market a djusted l ong-run return and t indicates the month a fter the IP O date. Coefficients are in percent , t -statistics are Whit e (1980) heteroskedast icity consistent t-statistics and are in italic. Multiple indicate whether the P/V ratio is computed equity or enterprise multiples. Metric indicates which financial metric is used t o compute the P/ V rat io. # indicates the sample size. ** indic ates significance at 1% l evel and * indicates significance a t 5% level.
Table 12
Cross-Sectional Regression IPO valuation and long-run returns for sample excluding firms with negative EBITDA
i i i t LnPV CAR, =α+γ1* +ε Multiple Metric t = 12 t = 24 t = 36 t = 48 t = 60 Equity Sales -3.33 -4.56 -5.38 -6.46 -6.05 -3.06 ** -2.79 ** -2.38 * -2.44 * -1.52 Enterprise Sales -4.11 -5.89 -8.19 -8.89 -6.68 -2.76 ** -2.77 ** -2.69 ** -2.29 * -1.27 Equity GCM -2.93 -4.13 -5.15 -5.95 -4.73 -2.83 ** -2.77 ** -2.44 * -2.51 * -1.35 Enterprise GCM -3.37 -4.99 -7.45 -7.85 -4.76 -2.49 * -2.77 ** -2.69 ** -2.32 * -1.13 # 741 579 385 290 233
GCM refers to the Gross Cost Margin. The further descript ion of this table is equal to the desc ripti on of table 12
Coefficients & t-statistics for LnPV Table 13
Cross-Sectional Regression IPO valuation and long-run returns for sample including firms with negative EBITDA
cross-section in stock returns, which is inconsistent with my findings. Opposing to Purnanandam and Swaminathan (2004), the book-to-market ratio is significant in my regression model, I consider this due to the fact that I do not control for this ratio in the long-run returns already.
4.4 Outliers
The descriptive statistics reveal that the sample of IPO firms may contain outliers. Anderson et al. (2002) recommend treating data with a z-score less than – 3 or greater than 3 as an outlier. Excluding data following this rule, does not yield different results regarding significance. In appendix D, I present these results for the sample including firms with negative EBITDA and for the valuation procedure involving enterprise multiples in combination with sales as financial metric. Table D.9 shows that the median IPO is still significantly overvalued. Also, overvaluation is still significantly positive related to the first-day return and unrelated to long-run returns, with controlling both relationships for the book-to-market ratio, size and profitability. I present both results in Table D.10. Furthermore, Table D.11 shows that overvaluation is still significantly negative related to long-run returns, without controlling for variables that explain the cross-section in stock returns.
5. Conclusion
Are IPOs really overvalued? Yes, they are. My results for the European Union between 2000 and 2008 show that the median IPO is significantly overvalued relative to industry matching firms. This overvaluation ranges from 41% to 93% depending on the sample of IPO firms and valuation procedure. Together with my result that IPOs earn a positive market adjusted first-day return, IPOs can be overvalued and underpriced at the same time. This is not in line with a market which is efficient in the short-run. Furthermore, my results show that IPOs earn below market returns till 3 years after the IPO. This is together with my overvaluation result in line with an efficient market in the long run, since stock prices ultimately turn to their fair values. Consequently, my results show that stock prices are not efficient in the short-run, but show efficiency in the long-run.
Suggesting that investors are overoptimistic about an IPO firms prospects at the time of the offer, as documented by Ritter (1991). The overoptimism by investors at the time of the offer is consistent with the findings of Purnanandam and Swaminathan’s (2004) that the overvaluation of IPOs declines when they use growth forecasts as matching criterion. Furthermore, my results are in line with Daniel et al. (1998). In their behavioral model, investors’ overconfidence about private information causes initial overreaction, continuing overvaluation and ultimately long-run reversals. This pattern is consistent with my results of overvaluation, followed by underpricing and ultimately long-run underperformance.
Concerning the relation between overvaluation and subsequent stock returns, I find significant results in two cases. First, overvaluation is significantly positive related to the first-day return for the sample including firms with negative EBITDA, which is robust for the book-to-market ratio, size and profitability. Which suggest that investor overoptimism about the future prospect of an IPO firm at the time of the offer, as documented by Ritter (1991), is still present at the first day of trading. Besides, it shows that the market is not efficient in the short-run, therefore this result is not in line with asymmetric information models of IPO pricing. Second, IPO valuation is significantly negative related to long-run returns, which is not robust for the book-to-market ratio, size and profitability. The fact that the most overvalued IPOs earn significantly the lowest return in the long-run, suggest that my valuation procedure does a reasonable job in distinguishing between under- and overvalued IPOs and is in line with an efficient market in the long-run. In general, the empirical results on relating IPO valuation to subsequent stock returns, do not have implications for stock market investors. My results on relating IPO valuation to subsequent stock returns are roughly in line with Purnanandam and Swaminathan (2004). Since they find that overvaluation is related significantly positive to the first-day return and significantly negative to long-run returns, whith controlling for variables that explain the cross-section in stock-returns.
I adjust the study of Purnanandam and Swaminathan (2004) by including firms with negative EBITDA in the sample of IPO firms and using enterprise multiples in addition to equity multiples in the valuation procedure. Both adjustments, however, do not yield different results, since IPO overvaluation is still significant with this adjustments. Consequently, both adjustments generalize the findings of Purnanandam and Swaminathan (2004).
firm-of possible matching firms for an IPO firm, one should classify firms in less than 62 industry groups. I recommend to make these two adjustments in further research. However, the latter only when the group of possible matching firms is small.
References
Anderson, D. R., Sweeney, D. J., and Williams, T. A., 2002. Statistics for Business and Economic, 8th
edition, West Publishing Company.
Barberis, N., Shleifer, A., and Vishny, R., 1998. A Model of Investor Sentiment, Journal of Financial Economics, 49, 307-343.
Barber, B.M., and Lyon, J.D., 1997. Detecting Long-Run Abnormal Stock Returns: The Empirical Power and Specification of Test-Statistics, Journal of Financial Economics, 43, 341-372.
Baron, D.P., 1982. A Model of Demand for Investment Banking Advising and Distribution Services for New Issues, Journal of Finance, 37, 955-976.
Daniel, K., Hirshleifer, D., and Subrahmanyam A., 1998. A Theory of Overconfidence, Self-Attribution, and Security Market Under- and Overreactions, Journal of Finance, 53, 1893-1886.
Fama, E.F., and French, K.R., 1992. The Cross-Section of Expected Stock Returns, Journal of Finance, 47, 427-465.
Fama, E.F., and French, K.R., 1993. Common Risk Factors in the Returns on Stocks and Bonds, Journal of
Financial Economics, 33, 3-56.
Liu, J., Nissim, D., and Thomas, J., 2002. Equity Valuation Using Multiples, Journal of Accounting
Research, 40, 135-172.
Ljungqvist, A., 2004. IPO Underpricing: A Survey. Handbook In Corporate Finance: Empirical Corporate
Finance, Espen Eckbo.
Loughran, T., and Ritter, J., 2004. Why Has IPO Underpricing Changed Over Time?, Financial
Management, 37, 5-38.
Koller, T., Goedhardt, M., and Wessels, D., 2005. Valuation: Measuring and Managing the Value of
Companies, 4th edition, John Wiley and Sons.
Pukthuanthong-Le, K., and Varaiya, N., 2007. IPO Pricing, Block Sales, and Long-Term Performance,
Financial Review, 42, 319-348.
Purnanandam, A., and Swaminathan, B., 2004. Are IPOs really underpriced?, Review of Financial Studies, 17, 811-848.
Ritter, J., 1991. The Long-Run Performance of Initial Public Offerings, Journal of Finance, 46, 3-27.
Rock, K., 1986. Why New Issues are Undervalued?, Journal of Financial Economics, 15, 187-212.
Sheskin, J., 1997. Parametric and @onparametric Statistical Procedures, CRC Press.
Teoh, S.H., Welch, I., and Wong, T.J., Earnings Management and the Long-Run Market Performance of Initial Public Offerings, Journal of Finance, 53, 1935-1974.
Welch, I., 1989. Seasoned Offerings, Imitation Costs, and the Underpricing of Initial Public Offerings,
Journal of Finance, 44, 421-449.
White, H., 1980. A Heteroscedasticity Consistent Covariance Matrix Estimator and a Direct Test for Heterescedasticity, Econometrica, 48, 817-838.
Appendices
Appendix A: Sample Information
Year IPO firms Matching firms IPO firms Matching firms
2000 110 110 148 148 2001 27 27 42 42 2002 22 22 32 32 2003 19 19 22 22 2004 38 38 64 64 2005 64 64 95 95 2006 158 158 209 209 2007 106 106 140 140 2008 7 7 20 20
# indicates the total sample size
Table A.1
Bumber of firms by year for sample excluding and including firms with negative EBITDA
Sample excluding firms with negative EBITDA Sample including firms with negative EBITDA (# = 772)
(# = 551)
Country IPO firms Matching frims IPO firms Matching firms
Austria 6 1 9 6 Belgium 13 12 22 12 Switzerland 5 16 7 20 Czech Republic 0 4 0 4 Germany 118 75 162 95 Denmark 2 8 5 9 Estonia 1 2 2 3 Spain 20 13 21 13 Finland 8 24 9 25 France 147 133 193 177 Great Britain 86 113 162 221 Greece 45 82 48 90 Hungary 0 1 0 1 Ireland 5 1 6 5 Italy 52 10 59 11 Luxemburg 1 0 1 0 Latvia 0 1 0 1 the Netherlands 5 4 9 8 Norway 0 5 1 7 Poland 12 5 13 5 Portugal 6 4 7 6 Romania 1 2 1 2 Sweden 18 28 35 44 Slovenia 0 5 0 5 Slowakia 0 2 0 2
# indicates the total sample size
Table A.2
Bumber of firms by country for sample excluding and including firms with negative EBITDA
Sample excluding firms with negative EBITDA Sample including firms with negative EBITDA (# = 772)
Industry IPO firms Matching firms IPO firms Matching frims
1 A Agriculture, hunting and related service activities 1 1 1 1
5 B
Fisthing, operation of fish hatcheries and fish farms; service activities
incidental to fishing 4 4 4 4
11 C
Extraction of crude petroleum and natural gas; service activities incidental
to oil and gas extraction, excluding surveying 2 2 6 6
13 C Mining of metal ores 2 2 5 5
14 C Other mining and quarrying 0 0 4 4
15 D Manufacture of food products and beverages 13 13 15 15
17 D Manufacture of textiles 1 1 1 1
18 D Manufacture of wearing apparel; dressing and dyeing of fur 1 1 1 1
19 D
Tanning and dressing of leather; manufacture of luggage, handbags,
saddlery, harness and footwear 3 3 3 3
20 D
Manufacture of wood and of products of wood and cork, except furniture;
manufacture of articles of straw and plaiting materials 2 2 2 2
22 D Publishing, printing and reproduction of recorded media 11 11 15 15
23 D Manufacture of coke, refined petroleum products and nuclear fuel 2 2 2 2
24 D Manufacture of chemicals and chemical products 11 11 20 20
25 D Manufacture of rubber and plastic products 7 7 8 8
26 D Manufacture of other non-metallic mineral products 9 9 9 9
27 D Manufacture of basic metals 3 3 3 3
28 D Manufacture of fabricated metal products, except machinery and equipment 5 5 8 8
29 D Manufacture of machinery and equipment n.e.c. 12 12 14 14
30 D Manufacture of office machinery and computers 3 3 5 5
31 D Manufacture of electrical machinery and apparatus n.e.c. 6 6 8 8
32 D
Manufacture of radio, television and communication equipment and
apparatus 10 10 16 16
33 D
Manufacture of medical, precision and optical instruments, watches and
clocks 9 9 20 20
34 D Manufacture of motor vehicles, trailers and semi-trailers 3 3 5 5
35 D Manufacture of other transport equipment 2 2 3 3
36 D Manufacture of furniture; manufacturing n.e.c. 5 5 7 7
40 E Electricity, gas, steam and hot water supply 6 6 6 6
41 E Collection, purification and distribution of water 2 2 3 3
45 F Construction 15 15 16 16
50 G
Sale, maintenance and repair of motor vehicles and motorcycles; retail sale
of automotive fuel 2 2 2 2
51 G
Wholesale trade and commission trade, except of motor vehicles and
motorcycles 46 46 63 63
52 G
Retail trade, except of motor vehicles and motorcycles; repair of personal
and household goods 15 15 21 21
55 H Hotels and restaurants 8 8 8 8
60 I Land transport; transport via pipelines 3 3 4 4
61 I Water transport 2 2 2 2
62 I Air transport 2 2 3 3
63 I Supporting and auxiliary transport activities; activities of travel agencies 11 11 14 14
64 I Post and telecommunications 14 14 16 16
70 K Real estate activities 27 27 30 30
71 K
Renting of machinery and equipment without operator and of personal and
households goods 2 2 2 2
72 K Computer and related activities 60 60 91 91
73 K Research and development 2 2 40 40
74 K Other business activities 173 173 223 223
80 M Education 1 1 1 1
85 @ Health and social work 6 6 7 7
90 O Sewage and refuse disposal, sanitation and similar activities 2 2 2 2
92 O Recreational, cultural and sporting activities 21 21 26 26
93 O Other service activities 4 4 7 7
Table A.3
Bumber of firms by industry for samples excluding and sample including firms with negative EBTIDA
NACE Rev. 1.1 code (Section )
Sample excluding firms with negative EBITDA (# = 551)
Sample including firms with negative EBITDA (# = 772)
Appendix B: Sample selection criteria
For inclusion in the final sample the IPO firms has to meet the following criteria:
- The IPO firml should have the offer price announced in press releases about the IPO deal in Zephyr.
(3183)
- The IPO firm should have a BvDEP ID number available in Zephyr, in order to find data on sales
and EBITDA in Amadeus. (2835)
- The IPO firm should have an ISIN number available in Zephyr, in order to find stock price
information in Datastream. (2592)
- The IPO firm should be listed in Amadeus, which is the only database I consider for accounting data.
(2116)
- The IPO firm should have data available on sales and EBITDA in the fiscal year prior to the IPO in
Amadeus. (943)
- The IPO firm should have a NACE Rev. 1.1 code available in Amadeus. (932)
- The IPO firm should have data on daily share price and daily shares outstanding at the IPO date in
Datastream. Also, the IPO firm should have a monthly total return index (TRI) available after the IPO date plus one day in Datastream. (835)
- Similar to Purnanandam (2004), the IPO firm should not be a financial firm, a closed-end funds or a
real estate investment trust. (772)
- The IPO firm should have positive EBITDA in the prior fiscal year. (551)
