Underpricing and overvaluation of IPO’s during the crisis : behavioral finance vs efficient market theory

(1)

University of Amsterdam

Bachelor thesis Economics and Finance

“Underpricing and overvaluation of IPO’s during

the crisis”

Behavioral finance vs efficient market theory

Peter van der Laan

Student number: 9972072

Specialization: Economics and Finance Field: Corporate Finance

(2)

1. Introduction

2. Literature review

3. Data

4. Methodology

5. Results

6. Discussion

7. Conclusion

8. Literature

(3)

1. Introduction

An initial public offering (IPO) of the stock of a company is a big event for any company. Firms go public to raise capital and to convert the shareholder value of the original owners into cash. The issuer (the company going public) of the shares usually takes its offer to an investment bank which acts as underwriter of the IPO. Underwriting the IPO means that the investment bank estimates the value of the company and promises to find primary investors to buy the IPO stock. The

underwriter establishes the offer price of the IPO together with these primary investors, usually large institutional investors. After the IPO has been completed, the shares are brought to the secondary market where interested investors can start trading the stock.

There have been numerous studies that show two particular recurring effects in IPO stock prices, namely underpricing the primary value of the stock which results in a first day price jump and a subsequent underperformance of the stock

compared to Seasoned Equity Offerings (SEO). In these types of occurrences where consistent abnormal returns are found, one will find a lively discussion between believers of the efficient market hypothesis and those who support the behavioral approach to finance. In this thesis I will use a self‐compiled dataset of firms that went public during the financial crisis that started in 2008 and a set of firms that went public in the years before the crisis which will be used as

benchmark period, to test the validity of behavioral finance arguments and efficient market arguments.

As we will see the arguments made for the efficient market hypothesis are mainly mechanical, like for instance management boosting earnings in the period

preceding the IPO, underwriter reputation and rent‐seeking or underpricing as an insurance against litigation. The behavioral finance arguments consist of

psychological arguments like investor sentiment, irrationally high growth

expectations and divergence in investor beliefs. According to for instance Miller (1977), the divergence of investor beliefs gives a special upwards thrust to IPO

(4)

prices since short‐selling is not allowed and the price of the stock is determined by the most optimistic investor.

Since the proceeds of the IPO are the new capital for the issuer, the underpricing would indicate that there is money left on the table by the issuer. This indicates that on average there is a consistent abnormal return to be found in IPO stocks, which the primary investors capitalize on. The subsequent underperformance of the IPO stocks compared to equivalent non‐IPO stocks indicate that time and again, investor expectations are too high. These effects are very peculiar, since the Capital Asset Pricing Model (CAPM), developed mainly by Sharpe, Markowitz and Miller, states that all fluctuation in prices should be found in the stocks’ market beta and no abnormal return should occur. Furthermore, the efficient market hypothesis states that future payoffs and expectations should be priced into a stock price and structural abnormal returns will result in arbitrage possibilities, which will then cause these abnormal returns to disappear.

The overvaluations were higher than normal during the internet bubble period of 1999‐2001 with subsequently lower performances on stocks resulting in a burst of the bubble (Ljungqvist, Nanda and Singh, 2006). During this period mean returns jumped from around 20% to around 40% on the first day of selling (Dong and Michel, 2009). Some researchers have even found percentages of 65% for this period (Ritter and Loughran, 2004). Theory suggests that these higher initial returns are followed by lower subsequent performance of stocks compared to similar non‐IPO stocks (Ritter, 1991; Agarwal, Liu and Rhee, 2008). Most IPO’s are overvalued at between 14% and 50% when compared to their intrinsic value during the period of 1980‐1997 (Purnanandam and Swaminathan, 2004).

We see that market sentiments and/or other factors during the bull market of the beginning of the twenty‐first century additionally enlarges the effects of

underpricing. The resulting lower performance of IPO stock then suggests that lower quality firms go to market during bull markets or that the factors

responsible for the effects are enlarged by the bull market circumstances. The boosting of the effects during bubble periods would be evidence for both sides of

(5)

the discussion. The behavioral finance argument is that expectations, sentiments or mind set are at a peak during this period, causing the effects to be greater than in a normal period. For this thesis a normal period is considered to be a period during which the business cycle is not disrupted by any significant disturbance due to a social or economic phenomenon. From an efficient market point of view, for issuers and underwriters, bull markets, i.e. bubble periods, create opportunities to bring more and lower quality companies to market. The larger underpricing is a sort of compensation made by all companies, for the average subsequently lower performance.

The hypothesis I want to test in this thesis is whether we see the opposite effect during the recent financial crisis. Less firms will go to the capital market and one would expect that the ones that do go during a crisis period, will be either forced by financial distress or be of a higher average quality. According to behavioral finance, since expectations are low all‐round, the underpricing effect is expected to be smaller. One would also expect that the subsequent underperformance of the IPO stock compared to SEO stock will be lower. A crisis period should not be a reason to stop rent‐seeking or reputation‐boosting behavior for underwriters so we will not see a significant difference compared to a normal period if the reasons for the effects are underwriter manipulation of the prices or other types of market mechanics.

In the next section I will review the relevant literature on the subject, divided up in efficient market hypothesis arguments and behavioral finance arguments.

Subsequently, the data set will be described. As follows from the different studies on the subject, the quality of the data set is very important for the results and significant time went into compiling the data set. A separate database was set up with data from different sources. I will describe the methodology in part 4, which will be the comparison of averages of the underpricing and a regression analysis of the cumulative returns of 1 week, 1 month, 1 year, 2 year and 3 year periods of the cumulative daily return of the IPO stock compared to the cumulative daily return of the S&P 500 over a matched period for each IPO stock in a normal period (Jan 2003‐June 2008) and the recent financial crisis (July 2008‐Dec 2011). In

(6)

section 5 I will discuss the results of the research and the implications of these results and in section 6 I will give my conclusion.

2. Literature review

Underpricing and overvaluation of IPO stocks have been extensively researched and illustrated. Multiple reasons have been given by different authors for these phenomena. Some explanations follow from the efficient market theory, others from behavioral finance theory.

Efficient market theory

Following a number of investigations by the Securities and Exchange Commission in the 1960’s, Ibbotson and Jaffe (1975) were one of the first authors to prove that IPO’s have extraordinary returns in the first period after issuance and the

existence of periods with heightened activity of IPO’s, which they dub “hot issue” markets. Both the number of issues and the percentage jump in returns compared to non‐IPO stock are higher in these “hot issue” periods. A “cold issue” period is a period with a low number of issues and a lower return, but then IPO’s still perform above regular stock in the first period after issuance. They claim that issuers take advantage of these windows of opportunity, but are actually better off issuing in “cold issue” periods, since in such periods the issue premium between the offer price and the security price in the long run (the actual price of the stock) is lower. Issuers are thus receiving revenues closer to real value in the long run.

Rock (1986) argues in a “lemons” type of argument that there are two types of investors in the IPO market before going public. According to him these can be divided in informed investors who invest in information gathering themselves, and uninformed investors, who buy at random. If there is no underpricing, the

informed investors would invest in profitable IPO investments and crowd out uninformed investors. The underpricing makes sure that the uninformed investors would on average make a normal return on their investments and keeps them in

(7)

the market. The higher quality firms are hurt by this since they receive less than their real value.

Tiniç (1988) develops a theory he calls the “insurance‐against‐legal‐liabilities” hypothesis. He states that underwriters and issuers systematically underprice the securities to be on the safe side since they have responsibilities towards the possible investors regarding the facts and figures underlying the IPO issue. Any duped investor has the possibility by law to recover any losses, litigation fees and punitive damages if it's proven that preceding due diligence has not been carried out properly. Therefore, Tiniç claims that underwriters systematically underprice the securities to be on the safe side. Since insuring this potential problem will create moral hazard, premiums would be very high. Tiniç argues that issuing firms and underwriters set the offer price lower as an insurance against these costs.

Schiller (1990) poses a theory he dubs the “impresario” theory as a reason for underpricing of IPO’s. He compares the underpricing of concert tickets by an impresario of a musician creating the illusion of success with the underpricing of IPO’s by underwriting investment banks. An impresario might be inclined to do so to fill up subsequent shows. By underpricing an issue, the underwriter establishes his reputation as a good investment advisor, thereby assuring future business. Schiller states that underwriters are able to create “fads” or periods that investors are willing to invest extra in IPO’s, comparing this to the “hot issue” periods found by other researchers named in this thesis.

Ruud (1993) argues that underpricing is caused by underwriter price support. She describes the mechanism whereby underwriters support any IPO that falls below the offer price, thereby creating a positively skewed distribution of the first day closing price. She poses that stock prices can rise but are prevented to fall by underwriters, creating an upwards pressure on average return. According to her underwriters do this because of reputational and liability concerns.

(8)

Behavioral finance

Miller (1977) explains the occurrence of underpricing and overvaluation through his “divergence of opinion” theory. This theory states that when the future is uncertain and risky, meaning opinions are very divergent between optimistic and pessimistic investors, the most optimistic investors with the highest valuation of a stock will determine the price. Short selling keeps this in check for non‐IPO stock. For IPO stock the future is very uncertain and short selling is prohibited, hence the underpricing effect compared to non‐IPO stock, as well as the subsequent

underperformance.

Krigman, Shaw and Womack (1999) distinguish different types of IPO’s and their corresponding future performance. They found evidence of a correlation between the opening day return and the subsequent performance of the IPO stock. First day positive performers subsequently perform well and first day losers perform badly subsequently. They argue that the correlation can be explained by the fact that investor expectations are anchored on the first day result and do not adjust in the subsequent period. They find an exception for what they call “extra hot” IPO’s, these are the IPO’s with the highest first day return. These perform the worst of all IPO’s investigated. They explain this using the concept of “flipping” which is defined as the first day volume of IPO shares sold in block trades to the market or the underwriting syndicate. It is a sort of scam by underwriters, making use of hot periods and investor sentiments to flip low quality IPO’s.

Hirschleifer (2001) argues that the efficient market model which predicts that current and future value is incorporated in the price of a stock is incomplete. Similar to Miller’s hypothesis of diverging opinions, Hirschleifer states that the expectations incorporated in any stock price is a weighted average of all

expectations of rational traders as well as irrational traders. Arbitrage between these different types of traders would lead to an equilibrium price between the rational and the irrational. Furthermore, no trader can act with the speed of light, so there will always be a time lag between an occurrence and the reaction of

(9)

traders on this occurrence. He quotes Pliny the Elder to explain his general point: “The best plan is...to profit by the folly of others.”

Ljungqvist, Nanda and Singh (2006) describe the existence of the “sentiment” investor. Again this is similar to the “divergence of opinion” theory by Miller. They explain underpricing as a premium for regular investors who are assigned the IPO stock pre issue against the underpriced offer price. Sentiment investors will then bid up the price, but as time passes the IPO stock underperforms. The premium earned by underpricing is a compensation for this loss by the regular investors. Their model predicts that in hot periods IPO’s underperform due to the fact that in hot periods the demand for IPO stocks are high (because “sentiment” is high) and issuers will capture part of this sentiment by issuing against higher prices.

Subsequently, they will underperform.

To conclude the literary review, some special attention is paid to Jay Ritter, the author with the most extensive research and work done on underpricing and overvaluation effects of IPO’s. In a number of papers (e.g. 1991, 2002, 2004) he finds the occurrence of these effects using extensive methods like for instance matching each firm in a 1,500+ data set to a comparable firm, taking into account the different industries of the IPO firms and weighting the portfolios by letting IPO fall into the portfolio by date. Furthermore, he constructed a large database of IPO data ranging from the late 50’s of the previous century until recent years, which is publicly accessible. In his 1991 paper called “The Long-Run Performance

of Initial Public Offerings” he finds that investing in IPO’s during the period of 1975

to 1984 would have made an investor only 83 cents of every dollar earned from the comparable firms portfolio. In his 2002 paper called (co‐author I. Welch) “A

Review of IPO Activity, Pricing and Allocations” he finds IPO stock over the period

of 1980 to 2000 underperforming the market by over 23% and a comparable firms portfolio by over 5%. In his 2004 paper (co‐author T. Loughran) called “Why Has

IPO Underpricing Changed Over Time?” he looks at different time periods and

finds underpricing with resulting price jumps of 65% during the internet bubble period, which subsequently drops to 12% in the burst period of 2001 to 2003. His arguments for the effects have evolved over time. In 1991 his main argument was

(10)

the “fads” explanation, similar to the “hot issue” explanation other authors have given. In his 2002 paper Ritter considers share allocation issues (i.e. underwriter’s behavior) an additional reason for the effects. In his 2004 paper the focus shifts to issuer acceptance of underpricing, mainly due to market power of underwriters and changing ownership composition of issuing firms, as well as to the “spinning” concept. Spinning is the activity where underwriters set up personal brokerage accounts to allocate side payments to executives and venture capitalists of hot IPO’s. Ritter argues that the side payments make up part of the reward so the issuer will agree to higher underpricing.

The more mechanical arguments stemming from efficient market theory were presented by for instance Rock, Schiller and Krigman, Shaw and Womack. We have seen a number of behavioral finance arguments like the arguments of Ibbotson, Miller and more recent, Ljunqvist, Nanda and Singh, where investor sentiments or mind sets cause the effects of underpricing and overvaluation. Ritter has swayed from behavioral finance arguments to the mechanics of interaction between issuers, underwriters and primary investors.

As mentioned in the introduction, the way the effects under study behave during a crisis period could indicate the validity of one of the two types of arguments. If the behavior of the overpricing and overvaluation effects of IPO stock during the crisis period is similar to the behavior of these effects during a normal period, an argument can be made for the mechanical arguments of the believers of the efficient market hypothesis. If on the other hand we see a drop in the effects during the crisis period compared to the normal period that is similar to the enlarging effect on underpricing and overvaluation of the bubble period, the sentiment arguments of the behavioral finance approach is more likely to be the cause of the effects.

(11)

3. Data

The most widely used database for IPO research is the Securities Data Company (SDC) database. Since the University of Amsterdam regrettably does not have access to this database, a work‐around had to be made to construct a data set. The length of the period for accepting an IPO in the data set was determined as follows. For the normal period the start date is January 1st_{2003, this is regarded as}

the end of the recession following the internet bubble. As a starting point for the recent financial crisis the period around the bankruptcy of Lehman Brothers is taken, so the end date of the normal period is June 30th_{2008 and the start of the}

crisis period is July 1st_{2008. The European debt crisis was still going strong in}

2011, hence the end date of the crisis period was selected as December 31st_2011.

The Zephyr database was used to find all IPO’s from January 1st_{2003 until}

December 31st_{2011. The data of the New York Stock Exchange (NYSE) and the}

NASDAQ National Market were taken. The first result was 1,140 IPO’s. Like Krigman, Shaw and Womack (1999) as well as Ritter (1991), all financial

corporations, investment funds, REIT’s (Real Estate Investement Trusts), mutual funds, mezzanine companies, and special purpose vehicles were removed from the set, since these listings do not perform the same as regular companies. This filter resulted in 824 remaining companies. Subsequently, the companies with no ticker or CUSIP (identification number giving to securities by the Committee on Uniform Security Identification Procedures) were removed, these could be IPO’s that never listed in NYSE or NASDAQ, or were removed after bankruptcy or merger, thus creating a survivorship bias in the sample, which would skew the distribution in a positive direction. The resulting number of firms was 794. This set was split up in the periods indicated, creating a set of 580 IPO’s for the normal period and 214 IPO’s for the crisis period.

The trading data then needed to be downloaded from the Center for Research in Security Prices (CRSP). For the trading data the list of CUSIP’s created from the ISINs (International Securities Identification Number) from the Zephyr results, was used. Matching CRSP with Zephyr leaves 493 matches for the normal period,

(12)

mostly due to IPO’s from Zephyr which were on other stock markets than NASDAQ or NYSE, listed within 6 months of IPO. Matching CRSP with Zephyr for the crisis period leaves 196 matches. In total, around 900.000 data points were downloaded from CRSP. In the data set downloaded from CRSP are the daily price and return data for each stock and the daily return on the S&P 500 for comparison.

The data were uploaded in a database constructed especially for this thesis. In this database a number of adjustments was made to the data. The Zephyr results containing regular IPO data like the IPO date, offer price, and number of shares given out, were combined with the CRSP trading data. The daily returns of the IPO’s and S&P 500 had to be converted to cumulative returns. Then separate tables were constructed to obtain the data points for the results after 1 week, 1 month, 1 year, 2 years and 3 years.

Table 1: Descriptive statistics of the first day return for the normal period and the crisis period

Normal period First day return Crisis period First day return

Mean 0,136 Mean 0,081

Standard Error 0,009 Standard Error 0,011

Median 0,083 Median 0,055

Mode 0,000 Mode 0,000

Standard Deviation 0,206 Standard Deviation 0,160

Sample Variance 0,042 Sample Variance 0,025

Kurtosis 3,658 Kurtosis 1,116 Skewness 1,537 Skewness 0,451 Range 1,531 Range 0,992 Minimum ‐0,410 Minimum ‐0,375 Maximum 1,121 Maximum 0,617 Count 493 Count 196

As can be seen in table 1, the mean of the normal period is higher than the mean of the crisis period. Later on statistical techniques will be used to test if the

difference between these means is significant. The standard deviation of the crisis sample is smaller than the standard deviation of the normal sample. One would expect this to be the other way around and to see that prices during the crisis

(13)

period are volatile. The difference could be explained by looking at the difference in kurtosis of the samples and the difference in maximum values. The combination of those values indicates that the normal period sample is more positively skewed with higher values in the maximum. The higher standard deviation is in this case a positive or upward risk which seems higher for the normal period sample.

4. Methodology

The first day abnormal return of a stock is measured by the difference between the first day closing price and the offer price of the IPO divided by the offer price.

= ( − )/

where is the first day abnormal return of stock i, is the price at the end of the first day of trading the stock in the secondary market and is the offer price of the IPO.

To calculate the average first day abnormal return per period, all the returns of the stock in that period are summed up and divided by the number of IPO’s in the period.

= ∑

where is the first day abnormal return of sample i and is the number of observations for sample i.

The standard deviations of both data sets are calculated in the following way.

(14)

The averages are compared using a two sample t‐test. Independence and

approximately normal distribution of the returns is assumed because of the large sample size.

= ( − )/ +

The hypothesis to be tested is that the average first day abnormal return in the crisis period is smaller than the average first day abnormal return in the normal period. We formulate the following null hypothesis to be tested:

: =

where is the average of the normal period and is the average of the crisis period. A single‐tailed test is conducted and the following alternative hypothesis can be formulated:

: >

To estimate the 3 year buy and hold return for IPO stocks two measures are used. The first measure is similar to the method used to calculate the first day abnormal return. The holding period return (HPR) of stock i is calculated as follows:

= −

The HPR’s for all selected IPO’s in both periods are calculated and then the average HPR for the period is calculated:

= ∑

To estimate the performance of the stocks compared to market over a 3 year period after the IPO date, linear regression analyses are run on the data points of daily trading data compared to the market data for the 1‐week, 1‐month, 1‐year, 2‐year and 3‐year data points found. For this we will use the ordinary least

(15)

risk) of the stock. CAPM can be used since it can be expected that the same other factors (like Fama‐French factors) are of influence on both data sets. The model will thus take this form:

= + ( − ) +

where is the stocks total return, is the stocks abnormal return, is the risk premium for the stock sensitivity to market fluctuations, is the market return,

is the risk free rate for which we use the interest rate on the 3‐month US Treasury Bill which is adjusted according to the period tested, and is the stocks residual error which is assumed to be normally distributed with an expected mean of 0.

The next step is to compare the resulting alphas to see if they significantly differ from each other. We use a student t‐test for comparing means and as before we assume the abnormal returns are normally distributed. The hypothesis to be tested is that the average abnormal 1‐week, 1‐month, 1‐year, 2‐year and 3‐year buy and hold return in the crisis period is larger than the average abnormal 3‐year buy and hold return in the normal period. We formulate the following null

hypothesis to be tested:

: =

where is the regression alpha of the normal period and is the regression alpha for the crisis period. The alternative hypothesis is that the abnormal return is lower (more negative) in the normal period than in the crisis period.

: <

To compare the regression coefficients a Z‐test will be used with the following formula:

(16)

where is the standard error of coefficient of squared and is the

standard error of coefficient squared. The z‐tests will be run for all regression coefficients.

To be able to further study the effect of the crisis, a regression is done on the total sample with the following formula:

= + ( − ) + +

where is a dummy variable which is set to 1 if the IPO is in the crisis period and set to 0 if it occurred in the normal period. The value of will show the sensitivity of the abnormal return to the IPO having taken place in the crisis period. The assumption that is made in this model is that the risk premium in both periods is not significantly different from each other. Like in the earlier model we assume that is the stocks residual error which is assumed to be normally distributed with an expected mean of 0.

5. Results

Table 2: Results of comparing means first day return

Results first day jump

Period Normal Crisis

Average first day return 13,59% 8,08%

Standard deviation 20,61% 15,96%

Number of obs 493 196

t‐statistic 3,748

degrees of freedom 687

p‐value 0,000

CI normal period upper 16,526

CI normal period lower 10,655

As we have seen in table 2, the average first day return means of the two samples differ substantially. The t‐statistic for comparing means is 3,748 with a p‐value smaller then 0,001. We therefore reject the null hypothesis that these means are

(17)

the same. The mean of the normal period is larger than the mean of the crisis period at the 1% level. The lower confidence bound of the normal sample is 10,655 and the mean of the crisis period sample is well under this bound.

Table 3: Regression results normal period

Regression results normal period

Period Alpha Alpha t-stat Beta Beta t-stat R square No obs

1 week 0,013*** 3,178 1,008*** 4,070 0,034 470 1 month 0,018*** 2,947 1,546*** 7,493 0,107 470 1 year 0,147*** 6,021 1,476*** 9,740 0,168 470 2 year 0,289*** 8,643 1,217*** 8,320 0,132 456 3 year 0,362*** 9,596 0,699*** 3,305 0,025 424 Significance levels *<0,10, **<0,05, ***<0,01

As can be seen in table 3, the abnormal return (Alpha) for the normal period is a positive value, indicating the portfolio of IPO stock outperformed the market in the first 3 years after the IPO. The alphas rise through the period and a peak abnormal return of 36,2% can be seen after the 3 year period. All alphas are significant at the 1% level as can be seen from the high t‐statistics and indicated by the stars. The betas are larger than one for the first part of the period,

indicating that the systematic risk or volatility of the IPO returns is high during this period. The beta peaks after the 1 month period and then a beta decay can be seen, a decline in the beta to even below the volatility of the market at the 3 year period. The betas are all significant at the 1% level. The low R‐squared value for the samples indicate that the explanatory value of the CAPM model for the IPO returns is limited. The variance in the return of the 1 week sample is only

explained for 3,4% by the model. Also, the R‐squared for the 3 year period is very low. We see the number of observations drop due to some companies dropping out of the sample. This is caused by bankruptcies or mergers, the first case causing a survivorship bias in the sample, skewing the results positively.

(18)

Table 4: Regression results crisis period

Regression results crisis period

Period Alpha Alpha t-stat Beta Beta t-stat R square No obs

1 week 0,000 0,000 0,852*** 4,049 0,078 195 1 month ‐0,011 ‐1,048 1,336*** 6,205 0,166 195 1 year ‐0,158*** ‐2,630 2,031*** 4,276 0,087 193 2 year ‐0,282* ‐1,972 1,781*** 3,187 0,053 185 3 year ‐0,100 ‐0,304 1,328 1,601 0,015 176 Significance levels *<0,10, **<0,05, ***<0,01

As can be seen in table 4, for the crisis period a negative value for the abnormal return is found, indicating the portfolio of IPO stock underperformed the market in the first 3 years after the IPO. The alphas fall over the first two years to a bottom value of ‐28,2% and then rise to ‐10% after the 3 year period. The 1 year alpha is significant at the 1% level and the 2 year alpha is significant at the 10% level. The other alphas are not statistically significant from zero, due to their high standard deviation. In Attachment 2, the regression results for the crisis, a

standard deviation of 0,33 can be found for the abnormal return value of ‐0,10 for the 3 year period for instance. The betas show an upward trend over the first 3 periods to a peak of 2,031. The betas are all significant at the 1% level except for the beta of the 3 year period, which does not cross the 10% threshold. The last fact can be explained by high volatility, a standard deviation of 0,83 can be found for the 3 year beta of 1,328. The R square values for the samples are even lower then in the normal period. Only the 1 month sample has an R‐square higher than 10% with 16,6% of the variance of the return explained by the model. The same drop in number of observations can be observed due companies dropping out of the sample.

Table 5: Z-values for comparing the regression coefficients of both periods

z-values

Period Alpha Beta

1 week 1,634 0,479

1 month 2,394 0,706

1 year 4,708 ‐1,113

2 year 3,888 ‐0,976

(19)

The z‐test mentioned earlier is used to test if the alphas are significantly different from each other. The threshold for rejecting the null hypothesis that the alpha of the normal period subsamples for 1 week, 1 month, 1 year, 2 year and 3 year is significantly different from the alphas for the corresponding periods of the crisis sample is a value of 1,96 or ‐1,96. As can be seen in table 5, the alphas for the periods of 1 month, 1 year and 2 years significantly differ from each other. The 1‐ week and the 3‐year values do not pass the significance level. The betas are not significantly different from each other over all periods.

Table 6: Regression results total period with crisis dummy

Regression results total period

Period Alpha Alpha t-stat

Beta Rm-Rf Beta t-stat Gamma Crisis Crisis t-stat R square No obs 1 week 0,013*** 3,104 0,913*** 5,810 ‐0,012* ‐1,669 0,053 665 1 month 0,018*** 2,879 1,436*** 9,811 ‐0,030*** ‐2,538 0,132 665 1 year 0,149*** 6,071 1,529*** 10,554 ‐0,258*** ‐5,317 0,146 664 2 year 0,229*** 7,111 1,079*** 8,632 ‐0,352*** ‐5,131 0,106 640 3 year 0,363*** 9,548 0,740*** 3,586 ‐0,233** ‐2,114 0,023 600 Significance levels *<0,10, **<0,05, ***<0,01

For the total period we find positive abnormal returns, moving up to 36,3% after the 3 year period. All alphas found are significant at the 1% level. The upward moving betas for the first three periods show a beta decay over the 2 year and 3 year period. A peak beta of 1,529 is found after 1 year. The betas are also

significant at the 1% level. The effect of the crisis is isolated by the dummy

variable gamma crisis. The negative effect on the return due to the IPO occurring during the crisis period drops down to ‐35,2% after the 2 year period, after which the return recovers a bit to ‐23,3% after 3 years. The significance levels for the gamma variable is at the 1% level for the 1 month, 1 year and 2 year period.

The full results of the regression analyses can be seen in attachment 1 for the normal period, in attachment 2 for the crisis period and in attachment 3 for the total period.

(20)

6. Discussion

The first day jump in the crisis period is significantly different from the first day return in the normal period preceding the financial crisis. This would indicate that the sentiments or mind set arguments for the IPO underpricing effect of

behavioral finance supporters are the main drivers behind the effects under research. As discussed before, the “fads” or “hot issue” periods create the driving force to determine the effects through heightened expectations and investor psychology, the “cold” or crisis period under study here shows a reverse effect. The fact that we see a magnification of this effect in the 1999 to 2001 bubble period and we see a diminishing effect during the crisis period indicate that this effect is not likely manufactured by market agents like underwriters.

The abnormal returns in the normal period actually outperformed the market to an extent of 36,2% positive abnormal return after the 3 year period. We see a positive abnormal return of the same value in the total sample with crisis dummy. This effect has not been documented before in IPO research. It indicates that IPO abnormal returns of which the IPO occurred during a relatively steady part of the business cycle have positive values. It seems that during this period the higher average risk of IPO’s compared to seasoned equity offerings, which can also be deducted from the market beta larger than 1 for most of the period, was actually rewarded with a positive abnormal return. Most research done on the subject, found a negative abnormal return. The results found in this thesis, indicate that the correlation between underpricing and the underperformance of IPO stocks is not that clear‐cut. Most studies focused on abnormal parts of the business cycle, either bubbles or crises, or took a sample over a very long period with both upswings and downturns in the sample. Apparently, isolating a normal period, defined in this thesis as a period during which the business cycle is not disrupted by any significant disturbance due to a social or economic phenomenon, reveals effects that have not been documented.

The underperformance effect during the crisis period was found in the same magnitude as other researchers did find. The minimum of the negative effect on

(21)

the return in the crisis sample was ‐28,2% and in the total sample, the minimum gamma value was ‐35,2%. This is between the bounds that previous research has found, ‐14% to ‐50% as mentioned in the introduction. As other researchers found, the effect diminishes towards the 3 year period after the IPO.

As can be seen from the z‐values for comparing the regression coefficients, the alphas significantly differ from each other between the normal and crisis period. This indicates that the crisis period has a negative impact on abnormal return of IPO stock. From a behavioral finance point of view, when market sentiment is good, people are willing to take more risk. Since IPO stocks are more risky in

general than non‐IPO stock, the argument can be made that the crisis period has a negative impact on market sentiment. The betas do not significantly differ from each other for the normal period and the crisis period. This result leads to the fact that the effect found in the crisis dummy is significant, since our assumption for this model was that the betas over both periods do not differ.

Even though the underpricing effect was lower in the crisis sample than in the normal period, the underperformance is not lower. The fact that we find similar effects of underperformance in the crisis period as other researchers found in the bubble period, indicates that market sentiment is a driving force behind the effects. The normal period investigated in this thesis, shows opposite effects, indicating further that the effects are not conjured by agents. The similar market betas over both periods under investigation, indicate that the volatility did not change significantly over both periods, but the returns compared to the market returns are very different from each other. The survivorship bias in similar as well in both samples, the normal period sample loses 9,78% of the companies over a 3 year period and the crisis period sample loses 9,74% if the sample, so this effect is not likely to be the driver.

The regression with the dummy variable controls for the crisis period while

assuming the betas over the whole period of IPO stocks are similar. We saw in the z‐test for comparing betas between the samples that this assumption is not

(22)

rejected. We find very similar effects in the total period model, strengthening the results found from the normal period sample and the crisis period sample.

The low R squared seen in all the models indicates that IPO dynamic differs from market dynamic. The market fluctuations only explain a small part of the

fluctuations in the returns of the IPO’s.

7. Conclusion

In conclusion, the research in this thesis indicates that the sentiment arguments made by the behavioral finance approach are more valid than the arguments of the believers of efficient market hypothesis for the underpricing effect. In the differences between the underpricing effects as shown through the first day jumps during a bubble of around 40%, a normal period of around 14% and in a crisis period of around 8%, the sentiments of the investors shines through. As Miller (1977) states, the price of the stock on the first day of trading will be determined by the most optimistic investor, and the research in this thesis indicates that the mood is influenced by the business cycle. The performance in the 3 year period after the IPO seems to be governed by market sentiment as well, as the returns between a normal period and a crisis period differ significantly.

The normal period was distinguished by a large liquidity build‐up that eventually led to the global financial meltdown. An explanation for the results that were found in this sample is that there was a high appetite for risk and this liquidity was searching for any and all investment possibilities, like for instance highly risky assets like subprime mortgages. The high demand for investment opportunities combined with the positive market sentiment could explain the positive abnormal return over this period of the risky IPO’s. The testing of this hypothesis is beyond the scope of this thesis and the abnormal returns found in the normal period justify additional research.

(23)

The flight to safety during the crisis period, whereby investors mostly invested in risk free government bonds and tangible assets like gold, could explain the results we see in this sample, i.e. that investors shy away from risky IPO stock. From the high standard deviations of the regression coefficients of the crisis period

compared to the normal period, the high volatility of the returns in this period can be observed.

The findings in this thesis warrant further research into the mechanics and development of IPO price effects during different parts of the business cycle. A comparison could be made between the bubble period of 1999 to 2001 with subsequent performance and the crisis period. Also, loading up the regression models with additional explanatory power by adding Fama‐French or other factors to create a multifactor model could yield more information. IPO companies are relatively smaller than average companies and usually have more growth

potential, meaning a high market to book value, so these factors will better help explain of the volatility of the stock prices. As mentioned earlier, for different IPO’s from different periods the factors are not expected to differ, so the basic analysis results presented in this thesis are not expected to differ.

(24)

8. Literature

‐Agarwal, S., Liu, C. and Ree, S.G., “Investor demand for IPOs and aftermarket performance: Evidence from the Hong Kong stock market”, Int. Fin. Markets, Inst.

and Money, Vol 18 (2008), pp. 176-190

‐Dong, M. and Michel, J‐S., “Does Investor Heterogeneity Lead to IPO Overvaluation?”, Working paper (March, 2009)

‐Hirshleifer, D., “Investor psychology and asset pricing”, The Journal of Finance,

Vol. 56, No. 4 (Aug., 2001), pp. 1533-1597

‐Ibbotson, R. and Jaffe, J., “Hot Issue Markets”, The Journal of Finance, Vol. 30,

No. 4 (Sep., 1975), pp. 1027-1042

‐Krigman, L., Shaw, W. and Womack, K., “The persistence of IPO Mispricing and the Power of Flipping”, The Journal of Finance, Vol. 54, No. 3 (Jun., 1999), pp.

1015-1044

‐Ljungqvist, A., Nanda, V. and Singh, R., “Hot markets, investor sentiment, and IPO pricing”, The Journal of Business, Vol. 79 (2006), pp. 1667-1702

‐Miller, E., “Risk, Uncertainty and Divergence of Opinion”, The Journal of Finance,

Vol. 32, No. 4 (Sep., 1977), pp. 1151-1168

‐Purnanandam, A.K. and Swaminathan, B., “Are IPOs really underpriced?”, The

Review of Financial Studies, Vol 17, No. 3 (2004), pp. 811-848

‐Ritter, Jay R., “The Long-Run Performance of Initial Public Offerings”, The Journal

of Finance, Vol. 46, No. X (1991), pp. 3- 27

‐Ritter, J. and Welch, I., “A Review of IPO Activity, Pricing and Allocations”, The

Journal of Finance, Vol. 57, No. 4 (Aug. 2002), pp. 1795-1828

‐Ritter, J. and Loughran T., “Why Has IPO Underpricing Changed Over Time?”, The

Journal of Financial Management (Autumn 2004), pp. 5-37

‐Rock, K., “Why new issues are underpriced”, The Journal of Financial Economics,

(25)

‐Ruud, J., “Underwriter price support and the IPO underpricing puzzle”, The

Journal of Financial Economics, Vol. 34 (1993), pp. 133-151

‐Schiller, R., “Speculative Prices and Popular Models”, The Journal of Economic

Perspectives, Vol. 4, No. 2 (Spring 1990), pp. 55-65

‐Tiniç, S., “Anatomy of Initial Public Offerings of Common Stock”, The Journal of

(26)

Attachment 2: Regression results normal period Results 1 week Regression Statistics Multiple R 0,18491 R Square 0,034192 Adjusted R Square0,032128 Standard Error0,085762 Observations 470 ANOVA df SS MS F Significance F Regression 1 0,1218622 0,121862 16,5682 5,51E-05 Residual 468 3,4422269 0,007355 Total 469 3,564089

CoefficientsStandard Error t Stat P-value Lower 95%Upper 95%Lower 95,0%Upper 95,0%

Intercept 0,012575 0,0039568 3,177995 0,001581 0,004799 0,02035 0,004799 0,02035 Rm-Rf 1,008326 0,2477212 4,070405 5,51E-05 0,521542 1,495109 0,521542 1,495109 Results 1 month Regression Statistics Multiple R 0,327303 R Square 0,107127 Adjusted R Square0,105219 Standard Error0,133387 Observations 470 ANOVA df SS MS F Significance F Regression 1 0,9990387 0,999039 56,15072 3,38E-13 Residual 468 8,3266984 0,017792 Total 469 9,3257371

Intercept 0,018136 0,0061546 2,946694 0,003372 0,006042 0,03023 0,006042 0,03023 Rm-Rf 1,546099 0,2063288 7,493379 3,38E-13 1,140654 1,951545 1,140654 1,951545 Results 1 year Regression Statistics Multiple R 0,410159 R Square 0,168231 Adjusted R Square0,166457 Standard Error0,520533 Observations 471 ANOVA df SS MS F Significance F

(27)

Regression 1 25,702293 25,70229 94,85832 1,55E-20

Residual 469 127,07768 0,270955

Total 470 152,77998

Intercept 0,147266 0,0244592 6,020904 3,5E-09 0,099203 0,19533 0,099203 0,19533 Rm-Rf 1,475813 0,1515283 9,739524 1,55E-20 1,178055 1,773572 1,178055 1,773572 Results 2 year Regression Statistics Multiple R 0,363742 R Square 0,132308 Adjusted R Square0,130397 Standard Error0,668059 Observations 456 ANOVA df SS MS F Significance F Regression 1 30,896314 30,89631 69,22728 1,04E-15 Residual 454 202,62137 0,446303 Total 455 233,51768

Intercept 0,289367 0,0334788 8,643292 9,43E-17 0,223575 0,35516 0,223575 0,35516 Rm-Rf 1,217146 0,1462864 8,320293 1,04E-15 0,929663 1,504628 0,929663 1,504628 Results 3 year Regression Statistics Multiple R 0,158842 R Square 0,025231 Adjusted R Square0,022921 Standard Error0,771113 Observations 424 ANOVA df SS MS F Significance F Regression 1 6,4949445 6,494944 10,92293 0,001031 Residual 422 250,92769 0,594615 Total 423 257,42263

Intercept 0,362362 0,0377604 9,596348 7,41E-20 0,28814 0,436584 0,28814 0,436584

(28)

Attachment 1: Regression results crisis period OUTPUT 1 week

Re gres s i on Sta ti s ti cs Mul ti pl e R 0,27979363 R Squa re 0,07828447 Adjus ted R Squa re0,07350875 Sta nda rd Error 0,09204913 Obs erva ti ons 195 ANOVA

df SS MS F Si gni fi cance F Re gres s i on 1 0,1388914 0,1388914 16,39215 7,45E-05 Re s i dua l 193 1,6352971 0,008473

Tota l 194 1,7741885

Coe ffi cients Sta nda rd Errort Sta t P-val ue Lowe r 95%Upper 95%Lowe r 95,0%Upper 95,0% Intercept -2,23E-05 0,0065987 -0,0033722 0,997313 -0,01304 0,012993 -0,01304 0,012993 S&P 0,85234196 0,2105212 4,0487226 7,45E-05 0,437124 1,26756 0,437124 1,26756 OUTPUT 1 month Re gres s i on Sta ti s ti cs Mul ti pl e R 0,40755734 R Squa re 0,16610299 Adjus ted R Squa re0,16178228 Sta nda rd Error 0,146158 Obs erva ti ons 195 ANOVA

Tota l 194 4,9441322

Coe ffi cients Sta nda rd Errort Sta t P-val ue Lowe r 95%Upper 95%Lowe r 95,0%Upper 95,0% Intercept -0,0111538 0,0105149 -1,0607565 0,290126 -0,03189 0,009585 -0,03189 0,009585 S&P 1,33446025 0,2152259 6,2002781 3,35E-09 0,909963 1,758957 0,909963 1,758957 OUTPUT 1 year Re gres s i on Sta ti s ti cs Mul ti pl e R 0,2961079 R Squa re 0,08767989 Adjus ted R Squa re0,08290335 Sta nda rd Error 0,52858479 Obs erva ti ons 193 ANOVA

Tota l 192 58,494555

Coe ffi cients Sta nda rd Errort Sta t P-val ue Lowe r 95%Upper 95%Lowe r 95,0%Upper 95,0% Intercept -0,1595644 0,0602369 -2,648949 0,00875 -0,27838 -0,04075 -0,27838 -0,04075 S&P 2,03285174 0,4744742 4,2844301 2,90E-05 1,096969 2,968734 1,096969 2,968734 OUTPUT 2 years Regression Statistics Mul ti pl e R 0,2361354 R Squa re 0,05575993 Adjus ted R Squa re0,05060015 Sta nda rd Error 0,7036167 Obs erva ti ons 185

(29)

ANOVA

df SS MS F Significance F

Re gres s i on 1 5,3501154 5,3501154 10,80664 0,001213 Re s i dua l 183 90,598992 0,4950765

Tota l 184 95,949107

CoefficientsStandard Error t Stat P-value Lower 95% Upper 95%Lower 95,0%Upper 95,0%

Intercept -0,3006091 0,1421183 -2,1152035 0,035766 -0,58101 -0,02021 -0,58101 -0,02021 S&P 1,81430659 0,5519062 3,2873461 0,001213 0,725389 2,903224 0,725389 2,903224 OUTPUT 3 years Re gres s i on Sta ti s ti cs Mul ti pl e R 0,12044467 R Squa re 0,01450692 Adjus ted R Squa re0,00884317 Sta nda rd Error 0,7935455 Obs erva ti ons 176 ANOVA

df SS MS F Si gni fi cance F Re gres s i on 1 1,6129262 1,6129262 2,561361 0,111318 Re s i dua l 174 109,57032 0,6297145

Tota l 175 111,18324

Coe ffi cients Sta nda rd Errort Sta t P-val ue Lowe r 95%Upper 95%Lowe r 95,0%Upper 95,0% Intercept -0,1052596 0,3335169 -0,3156051 0,75268 -0,76352 0,553 -0,76352 0,553 S&P 1,33481207 0,8340358 1,6004253 0,111318 -0,31132 2,980941 -0,31132 2,980941

(30)

Attachment 3: Regression results total period Results 1 week Regression Statistics Multiple R 0,230142 R Square 0,052965 Adjusted R Square 0,050104 Standard Error 0,087594 Observations 665 ANOVA df SS MS F Significance F Regression 2 0,284073 0,142036 18,51202 1,5E-08 Residual 662 5,079299 0,007673 Total 664 5,363372

CoefficientsStandard Error t Stat P-value Lower 95% Upper 95%Lower 95,0%Upper 95,0%

Intercept 0,012543 0,004041 3,104202 0,001989 0,004609 0,020477 0,004609 0,020477 Rm-Rf 0,912535 0,157069 5,809767 9,72E-09 0,604121 1,220949 0,604121 1,220949 Crisis -0,01246 0,007463 -1,66887 0,095615 -0,02711 0,002199 -0,02711 0,002199 Results 1 month Regression Statistics Multiple R 0,363118 R Square 0,131854 Adjusted R Square 0,129232 Standard Error 0,137183 Observations 665 ANOVA df SS MS F Significance F Regression 2 1,89218 0,94609 50,27246 4,72E-21 Residual 662 12,45834 0,018819 Total 664 14,35052

Intercept 0,018218 0,006329 2,878557 0,004124 0,005791 0,030644 0,005791 0,030644 Rm-Rf 1,435667 0,14633 9,811175 2,63E-21 1,14834 1,722993 1,14834 1,722993 Crisis -0,02969 0,011699 -2,53826 0,011368 -0,05267 -0,00672 -0,05267 -0,00672 Results 1 year Regression Statistics Multiple R 0,382373 R Square 0,146209 Adjusted R Square 0,143626 Standard Error 0,523005 Observations 664 ANOVA df SS MS F Significance F Regression 2 30,96249 15,48124 56,59714 2,05E-23 Residual 661 180,806 0,273534 Total 663 211,7685

Intercept 0,148936 0,02453 6,071468 2,14E-09 0,100769 0,197103 0,100769 0,197103 Rm-Rf 1,528586 0,144834 10,55408 3,59E-24 1,244197 1,812976 1,244197 1,812976 Crisidummy -0,25767 0,048462 -5,3169 1,45E-07 -0,35283 -0,16251 -0,35283 -0,16251 Results 2 year

(31)

Regression Statistics Multiple R 0,325457 R Square 0,105922 Adjusted R Square 0,103115 Standard Error 0,68207 Observations 640 ANOVA df SS MS F Significance F Regression 2 35,10815 17,55408 37,73291 3,26E-16 Residual 637 296,3447 0,465219 Total 639 331,4528

Intercept 0,229458 0,032267 7,111161 3,1E-12 0,166095 0,292821 0,166095 0,292821 Rm-Rf 1,078825 0,12498 8,632006 4,81E-17 0,833403 1,324246 0,833403 1,324246 Crisis -0,35151 0,068502 -5,13133 3,82E-07 -0,48602 -0,21699 -0,48602 -0,21699 Results 3 year Regression Statistics Multiple R 0,151199 R Square 0,022861 Adjusted R Square 0,019588 Standard Error 0,777441 Observations 600 ANOVA df SS MS F Significance F Regression 2 8,442162 4,221081 6,983749 0,001004 Residual 597 360,8356 0,604415 Total 599 369,2777

Intercept 0,363289 0,03805 9,547661 3,36E-20 0,288561 0,438018 0,288561 0,438018 Rm-Rf 0,739691 0,2063 3,585513 0,000364 0,334529 1,144852 0,334529 1,144852 Crisis -0,23325 0,110358 -2,11359 0,034964 -0,44999 -0,01651 -0,44999 -0,01651

Underpricing and overvaluation of IPO’s during the crisis : behavioral finance vs efficient market theory

University of Amsterdam

Bachelor thesis Economics and Finance