Are investors in Initial Public Offerings affected by
their past performance?
*Bachelor Thesis
Sophie Heijenberg
10326642
BSc in Economics & Business
University of Amsterdam
Thesis supervisor: dr. J.E. Ligterink
Abstract
This study finds significant evidence that past returns on investment influence the future decision to invest in IPO shares for investors with a trading account at online broker Lynx (n=166). This result is consistent with the reinforcement learning theory: personally experienced outcomes outweigh rational decisions. Besides that, this study finds evidence that the deal size of an IPO and the underpricing at the first day are significant predictors of the probability to reinvest. Moreover, an extended dataset (n=203) finds enough evidence to assume there exists a gender difference: women are more likely to reinvest in IPOs. This study confirms the importance of psychological learning processes in finance and so contributes to the existing literature on
behavioral finance.
*I want to thank Sander Schwanen and Evert van der Hoorn from online broker Lynx for giving me the opportunity to use and for their assistance in retrieving data from Lynx to make this study possible.
Statement of Originality
This document is written by Sophie Heijenberg who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents
page nr.
I. Introduction 4
A. Theories of learning 5
B. Overconfidence and Risk aversion 6
C. Other factors that influence investment decisions 8
II. Research question and Hypotheses 9
III. Methodology 10
IV. Data and Statistical analysis 11
A. Data collection 11
B. Model and Statistical analysis 13
V. Results and Discussion 19
A. Results complete dataset 19
B. Interaction effects in the complete dataset 22
C. Limitations 23
D. Comparison to the extended dataset 24
VI. Conclusion 26
VII. Recommendations 27
I. Introduction
2014 was the year of a historic initial public offering (IPO). The IPO of Alibaba on the 18th
of September 2014 became the all time largest IPO with a deal size of 25 billion dollars (Barreto, 2014).
Inspired by the IPO of Alibaba, I wonder why investors invest in IPOs usually without knowledge of the future performance of the stock. A study by Miller (2000) finds evidence that IPO stocks typically underperform the indices in the first three years. This raises the question whether investors are affected by their past
performance and if future investment decisions are influenced by earlier experiences. There is not much empirical research about the way in which investors learn. For this reason, studying whether investors are affected by past performance is an interesting subject. Especially investments in IPOs are suitable for this analysis because on the moment of offering, there is no historical information available on the performance profile of the stock. Furthermore, nobody has experience with trading these stocks before.
The learning effects for investors have been previously studied. Kaustia and Knüpfer (2008) find that investors with positive returns on Finnish IPOs are more likely to invest in an IPO again. Their results suggest that investors outweigh their personal experience more than could be expected by rational learning. A comparable study by Chiang, Hirshleifer, Qian and Sherman (2011) shows that investors on the Taiwanese stock market tend to bid more aggressively in new IPOs when they had more previous experience in IPO investments.
Both studies find evidence that past performance influences future decisions. These studies are limited to respectively Finnish and Taiwanese investors with merely access to their national stock exchange. It can be questioned whether this effect is present in a different population that has access to a wider IPO market.
For this study I use a dataset from online broker Lynx. This firm, established in Amsterdam in 2006, facilitates a trading platform to thousands of clients. Lynx gives me access to a unique dataset of investors with the possibility to trade on a wider IPO market. This leads to a database of 166 individual investors that invested in 39 unique IPOs on NYSE and Nasdaq in the time period between January 2, 2012 and November 28, 2014.
By combining the results of previously reported learning effects and comparing these with the results of learning experiences present at Dutch IPO
investors, this study evaluates the existence of learning effects and possibly increases the external validity of the studies done in Finland and Taiwan.
This study adds additional insights to the existing literature because there is not much information about the way in which investors learn. These insights can help us in analyzing and better understanding decision-making of individual investors on the stock market. This study clarifies the psychological effects of learning and decision-making and shows their importance in behavioral finance.
The remainder of this paper I) discusses the theoretical part of learning from past performance, II) presents the research question and hypothesis, III) describes the methodology, IV) describes the data collection and performs the statistical analysis, V) presents the main results and discusses several explanations for them, VI) presents the conclusion, and VII) discusses brief remarks and subjects for later studies.
A. Theories of learning
It is interesting to study the way in which investors learn because it can tell us something about their future decisions. There are several theories about the ways investors can learn from experience.
Kaustia and Knüpfer (2008) describe the reinforcement learning theory. This theory suggests that personally experienced outcomes, such as incurring a positive return on investment, have a greater effect on behavior than outcomes without personal involvement, such as reading about positive returns on a certain investment. This means that the outcomes of investment decisions made in the past have an influence on future decisions. These future decisions are only affected by actual and directly experienced outcomes. Thus, in reinforcement learning, personal experiences outweigh rational Bayesian learning as discussed below.
Chiang et al. (2011) describe investors’ expectations: investors expect that the gains or losses they experienced in previous investments will recur, even when these expectations are not logically justified. This means that an investor who experienced a gain on a single IPO investment expects to make a gain in a future IPO investment as well. Vice versa for investors that incur losses on their IPO investment. Chiang et al. (2011) find that in reinforcement learning, personal experience is more important than information obtained by observation or communication with others.
If this study finds a correlation between positive returns and new future IPO investments, this would support that investors practice reinforcement learning and that the reinforcement learning theory is present in financial markets.
The second theory Kaustia and Knüpfer (2008) describe is the rational
Bayesian learning theory. A rational investor does not use information on investments in previous stocks when he invests in a new stock. He focuses on actual and forgone payoffs besides personally experienced outcomes (Kaustia & Knüpfer, 2008).
Rational investors only trade if their expected gains exceed the transaction costs they have to bear (Barber & Odean, 2001). This assumption forms the hypothesis that investors with less skill that incur losses should exit the stock market and stop investing.
Kaustia and Knüpfer (2008) describe recent research that suggests that investors with more trading experience are less likely to make behavioral errors and thus use more sophisticated trading strategies. This looks like a shift from reinforcement learning to more rational learning, in which investors include more objective information.
B. Overconfidence and Risk aversion
Barber and Odean (2000) find that the average individual investor realizes net trading losses. For this reason, profit-maximizing risk-averse Bayesian investors should not want to enter a market where the expected profits are negative. This raises the question: why do individuals want to invest and keep investing if their expected payoffs are negative?
The first theory is based on the potential similarities between investing and gambling. It can be stated that investors are ordinary sensation-seekers that, just like gamblers, “invest”, while their expected return is negative (Fong, 2014). Grinblatt and Keloharju (2009) describe that a trigger for sensation-seekers may be the novelty of new stocks. Because there is no information on past or future performance of IPOs, these are especially suitable for investors most prone to sensation seeking. This study also shows that these types of investors tend to trade more frequently than non sensation-seeking investors (e.g. buy-and-hold investors/long-term investors).
Wang (2012) finds that investors overreact to new events that occur in the market. As a result, they lose their rational judgment ability and make irrational trades
in IPOs. According to Wang (2012) investors pay too much for IPO shares due to this overly optimistic behavior.
Barber and Odean (2000)argue that the realized losses are too large for just being “sensation-seeking”. They study the overconfidence of investors and their belief that they can forecast prices. They find that investors tend to remember successful trades and their profits, but forget their mistakes and losses.
One of the factors that possibly influences overconfidence is gender. A study by Schubert, Brown, Gysler and Brachinger (1999) investigates the stereotypical view that women are more risk-averse than men because women are less trusted than men to make risky decisions that may be necessary for success. They show that in abstract gambles there exist gender-specific risk propensities. Men tend to be more risk-prone toward gains but women are more risk-prone toward losses. An IPO investment can be seen as an abstract gamble because there is not much information whether you will gain or lose money with this investment.
Barber and Odean (2001) find evidence for a potential gender difference: men are more overconfident than women. As a consequence, men trade 45 percent more and perform worse than women do.
There exists a difference in the literature between the gender differences in both overconfidence and risk aversion. Pålsson (1996) does not find a significant relation between the rate of risk aversion and being female, concluding that there does not exist a gender difference.
Besides gender, age could influence the rate of risk aversion for a particular investor. Pålsson (1996) finds evidence that risk aversion increases with age. As a result, people with a higher age should be less likely to (re)invest in IPO shares. They prefer “safer” shares with a clear performance profile instead of IPOs.
Korniotis and Kumar (2011) confirm that investors with a higher age and more trading experience hold less risky portfolios. They also find that these investors trade less frequently and have strong preferences for diversification.
Furthermore, there may be a difference between employed and retired traders. In general, retired traders have more spare time to follow the news and make
predictions about IPOs. Barber, Lee, Liu and Odean (2011) find that trading intensities are positively correlated with performance. More active investors
outperform less active investors. Thus, retired people with more spare time should be able to perform better than investors that have to work beside their trading activities.
C. Other factors that influence investment decisions
As stated above, the most obvious factor that should influence future investment decisions is the personally experienced performance. This is measured by the returns obtained in the past. Another factor that may influence any decisions on reinvestment is the initial return an investor made on the offering date of his IPO shares. This is defined as the difference between the first day closing price and the price at which the investor bought the IPO shares that day.
When including the total return on investment as well as the initial return on investment of each investor, the disposition effect bias is reduced. The disposition effect is the tendency for investors to immediately sell their winning stocks and hold on to losing shares until the moment that these shares realize a (small) gain (Barber & Odean (2007); Kaustia & Knüpfer (2008); Korniotis & Kumar (2011)). By including the initial return on investment for each investor, the outcome does not rely solely on the total return on investment.
Besides the initial return the investor makes, the initial return of the IPO itself on the offering date can also be important. This initial return equals the underpricing of the IPO because it uses the difference between the first day closing price and the offer price as stated by the underwriters.
There are two kinds of IPOs at the first trading day: hot and cold IPOs. A hot IPO is an IPO that appeals to many investors and for which there is great demand. They are often oversubscribed, meaning that the demand of the market exceeds the amount of shares supplied. As a result, the stock price skyrockets as soon as it is offered on the market. Because the stock price usually exceeds the offer price that was set before, these types of IPOs normally have a positive initial return. The contrary, an IPO with a negative return on the first trading day, is called a cold IPO.
Kaustia and Knüpfer (2008) confirm the importance of the initial IPO return by finding that more than twice as many investors reinvest in a future IPO if their first IPO investment was in a hot IPO than if it was in a cold IPO.
Though, returns are not the only factor to calculate for. What is it exactly were investors base their decision on? As discussed above, gender and age can be relevant
factors. Another factor to take in account is the deal size of an IPO, as larger IPOs are more likely to get media attention.
Barber and Odean (2008) find evidence that because of the large quantity of stocks and limited time, investors buy stocks driven by attention. They prefer to buy stocks that have gotten their attention due to the news.
Cook, Kieschnick and Van Ness (2006) find a positive correlation between the trading activity of individual investors on the first day of trading in an IPO and the publicity the IPO got before issuing. Thus, IPOs that get more media attention usually attract more investors. In addition to that, Cook et al. (2006) find evidence that the initial returns of an IPO are positively correlated with the amount of media attention it got before the IPO.
Consistent with Cook et al. (2006), Wang (2012) finds that the more attention an IPO gets, the higher the short-term returns of IPOs at the Taiwanese stock market. This implies that investors in large deal size IPOs are likely to obtain a higher return on the short-term. Taken into account that there is not too much time (max. 4 months) between every IPO in my dataset, this should positively influence the investment decisions of an investor to reinvest in IPO shares.
II. Research question and Hypotheses
To answer the research question “are investors at Lynx that invest in IPOs affected by
their past performance?”, I set up the following hypotheses:
I. Investors that incur a positive return on their first IPO investment invest with a higher probability in future IPOs than investors that incur a negative return II. Investors that incur a negative return on their first IPO investment are more
likely to quit IPO investing
III. Investors with a higher age are less likely to reinvest in IPO shares IV. There exists a gender difference: women are less likely to reinvest in IPO
shares
V. Investors with a first IPO investment in a hot IPO are more likely to reinvest in a future IPO
VI. Investors with a first IPO investment in a large deal size IPO are more likely to reinvest in a future IPO
III. Methodology
Lynx is an online brokerage firm established in 2006. Because Lynx was one of the first so-called discount brokers in The Netherlands, the firm saw a rapid growth of clients in the past eight years. At this moment, Lynx has about 8,500 clients and is still growing. The firm cooperates with Interactive Brokers U.S. and therefore uses a more complex trading platform than most Dutch banks and other competitors. Consequently, Lynx requires investors to have experience with trading before they can open a trading account.
This study focuses on IPOs on the New York Stock Exchange (NYSE) and Nasdaq in the three-year period between January 2, 2012 and November 28, 2014, as more experienced investors tend to be more United States-orientated. These
exchanges accommodate the major portion of the American trading market. The Lynx clients were able to trade in all IPOs on these exchanges in the specified period. In addition, there will not exist a problem with changing exchange rates between currencies in the three-year period of the sample.
It is important to notice that Lynx does not offer the possibility to investors to subscribe for IPO shares before the offering date, as some brokerage firms do. Thus, investors are only eligible for buying IPO shares after the moment at which the shares actually go to the market. Knowing this, every investor has a fair chance to obtain IPO shares and there will not exist any bias in my results caused by ex-ante IPO subscriptions. Moreover, investors at Lynx are all private clients, which, I assume, do not have any insider information that would give them any privileges compared to other investors.
I choose the period between January 2, 2012 and November 28, 2014 for several reasons. First of all, Lowry and Schwert (2002) find a negative relation between IPO volume and future initial returns. They find that periods with high and rising initial returns lead to more IPOs, which are themselves followed by periods of lower initial returns and less IPOs. In this way, there exists an IPO cycle. This cycle is more profound in crisis periods. Consequently, including IPOs from recent financial crisis period could bias the results.
Furthermore, Benninga, Helmantel and Sarig (2005) describe the “hot issue-phenomenon”. This is the phenomenon that many firms tend to go public at the same
time. They explain that good circumstances in the economy positively affect the firm’s cash flows. Thus, a high cash flow is the optimal timing to issue stock.
From the beginning of 2012, the United States Gross Domestic Product (GDP) has been positive except for the first quarter of 2014 (U.S. Bureau of Economic Analysis, 2014). This poses a good environment for IPOs and thus enlarges the sample and the chance of unbiased results.
Due to these positive growth rates and profits, the financial crisis is getting less severe the past years. This results in an increase in IPOs. This is an example of the “hot issue-phenomenon” described above. A positive effect enhances a more positive effect, so this is a self-fulfilling prophecy.
IV. Data and Statistical analysis
A. Data collection
For this study, I collect data from two databases. First, I retrieve all completed IPOs on the NYSE and the Nasdaq in the three-year period between January 2, 2012 and November 28, 2014 out of the Zephyr database (n=670). I also collect data on the deal size, the offer price and the first day closing price of the IPOs. The Zephyr database has several missing values on the offer price and the first day closing price. I
manually retrieve these values from Yahoo Finance and the Renaissance Capital IPO center.
The second database I use is the private server of Lynx. I retrieve all the trades in IPOs on the offering dates in the specified period. I specifically choose to only use trades done on the offering date because at this moment, there is no information available about past or future performance and in this way, there is no difference in experience with the stock between investors.
I use the buy and sell prices of investors to calculate the initial return on investment and the total return on investment of the first IPO in which they invested. If an investor buys an amount of shares on the offering date and sells this amount in parts, I use the weighted average sell price to calculate his return on investment. If an investor buys additional shares after the offering date and starts selling after that, I use the prices of the first sell transactions up to the amount of shares the investor bought at the offering date following the first in, first out principle.
Investors that still have a part or all of their IPO shares in their portfolio on November 28, 2014 (n=37) are excluded from the main analysis. After the main
analysis (n=166; IPOs=39) I perform an additional analysis (n=203; IPOs=45) as a robustness check. To calculate the total return on investment for the 37 additional investors, I use the closing price at Friday November 28, 2014. This day is the last day of my sample. I assume that all investors sell their shares on the last day of my sample. The comparison of both analyses will show if adding these 37 investors results in a different outcome.
The descriptive statistics of the complete dataset are displayed below. Table 1
These tables give the descriptive statistics for the investors and the IPOs in the sample (complete dataset) and the correlations between the main variables in the model. The observations do not equal 166 in every criterion because of missing values.
Investors, N = 166 percentiles
Obs Mean Median SD 25% 75%
Return on investment (%) 166 -3.85 -2.71 25 -12.80 0.78
Initial return (%) 166 -3.01 -2.69 6.29 -5.94 0.61
Age 163 47 45 14.49 35 57
Female dummy 163 0.07 0 0.26 0 0
Negative return dummy 166 0.66 1 0.47 0 1
The sample group consists of 166 Lynx clients of who 12 are female, 151 are male and there are three missing values. Their mean (rounded) age is 47.
The investors invested in 39 different IPOs on either NYSE or Nasdaq. The mean deal size of these 39 IPOs is around 11.5 billion dollars. The deal size is the most divergent variable in this study.
IPOs, N = 39 percentiles
Obs Mean Median SD 25% 75%
Deal size ($mm) 165 11,478.09 16,006.88 8,323.04 1,023.18 16,006.88 Initial return/underpricing (%) 165 20.01 0.61 29.09 0.61 38.07 Correlation matrix Return on investment Initial return Age Female dummy Negative dummy Size dummy Hot IPO dummy Return on investment 1.00 Initial return 0.44 1.00 Age 0.09 0.12 1.00 Female dummy 0.02 0.08 0.22 1.00
Negative return dummy - 0.55 - 0.42 - 0.06 - 0.05 1.00
Deal size dummy - 0.24 - 0.25 - 0.03 - 0.06 0.19 1.00
The descriptive statistics show that the average return on IPO investments is negative (-3.85%). This confirms the study of Miller (2000) that suggests that IPOs on average underperform the indices in the first three years after offering.
IPOs are on average underpriced with 20.01%. This means that the closing price at the first day is on average 20.01% higher than the offer price as estimated by the underwriters. This confirms the presence of underpricing in financial markets as described by Miller (2000).
The average initial return of investors on the first trading day is -3.01%. Miller (2000) suggests that this could be explained by the divergence of investors’ opinions. Because there are different expectations, the initial market price raises. At the end of the day, opinions are more convergent because there is a performance profile. Consequently, the market price lowers. This explains why the initial return of investors on the offering date is negative.
B. Model and Statistical Analysis
I use a nonlinear logistic regression to predict if an investor that once invested in an IPO at the offering date will invest again. I use a dependent binary variable (𝑦!) that
has a value of 0 when an investor does not reinvest in an IPO. The dependent variable has a value of 1 when there is at least one new investment in IPO shares made after the first IPO investment. These reinvestments tell something about the presence of learning effects and if past performance affects future decision-making.
For this study, I use the following regression equation:
𝟏 𝑦!= 𝛽!+ 𝛽!𝑟𝑒𝑡𝑢𝑟𝑛𝑖𝑛𝑣! + 𝛽!𝑖𝑛𝑖𝑡𝑟𝑒𝑡𝑢𝑟𝑛! + 𝛽!𝑎𝑔𝑒!+ 𝛽!𝑓𝑒𝑚𝑎𝑙𝑒!+ 𝛽!ℎ𝑜𝑡𝑖𝑝𝑜!+
𝛽!𝑑𝑒𝑎𝑙𝑠𝑖𝑧𝑒! + 𝜀!
The first two coefficients (𝛽!; 𝛽!) contain investment-specific information and say something about the returns an investor made. The coefficient returninv reflects the total return on investment the investor made on the first IPO stock he invested in. I calculate the return on investment for each investor by the capital gain yield formula:
𝟐 𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑖𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡 =𝑃!− 𝑃! 𝑃!
In which 𝑃! is the original price for which the investor bought the shares on the IPO
date and 𝑃! is the selling price for which the investor sold the shares at a particular
The coefficient initreturn shows the return an investor made at the offering date of the first IPO stock he invested in. The formula I use to calculate this return equals:
𝟑 𝐼𝑛𝑖𝑡𝑖𝑎𝑙 𝑟𝑒𝑡𝑢𝑟𝑛 𝑖𝑛𝑣𝑒𝑠𝑡𝑜𝑟 =𝐹𝑖𝑟𝑠𝑡 𝑑𝑎𝑦 𝑐𝑙𝑜𝑠𝑖𝑛𝑔 𝑝𝑟𝑖𝑐𝑒 − 𝐵𝑢𝑦 𝑝𝑟𝑖𝑐𝑒
𝐵𝑢𝑦 𝑝𝑟𝑖𝑐𝑒 ∗ 100
The next two coefficients (𝛽!; 𝛽!) reflect something about the personal characteristics of a particular investor. The control variable age is the age of the investor as an integer. The second control variable in the model is the dummy for
female. This dummy has a value of 0 for males and a value of 1 for females.
The last two coefficients (𝛽!; 𝛽!) tell something about the specific
characteristics of the first IPO in which an investor invested. Hotipo is a dummy variable that has a value of 1 when the initial return (underpricing) on the IPO is positive and a value of 0 when the initial return is equal to zero or negative. The initial IPO return is calculated with the following formula:
𝟒 𝐼𝑛𝑖𝑡𝑖𝑎𝑙 𝐼𝑃𝑂 𝑟𝑒𝑡𝑢𝑟𝑛 =𝐹𝑖𝑟𝑠𝑡 𝑑𝑎𝑦 𝑐𝑙𝑜𝑠𝑖𝑛𝑔 𝑝𝑟𝑖𝑐𝑒 − 𝑂𝑓𝑓𝑒𝑟 𝑝𝑟𝑖𝑐𝑒
𝑂𝑓𝑓𝑒𝑟 𝑝𝑟𝑖𝑐𝑒 ∗ 100
The hotipo dummy makes a distinction between hot IPOs and cold IPOs. In the sample of 39 IPOs there were 30 hot IPOs and 9 cold IPOs. The reason for the high initial return is linked to IPO underpricing as described by Miller (2000).
The coefficient dealsize is a dummy variable that can have either the value 0 or 1. The value 0 represents firms with a deal size below 250 million dollar and the value 1 represents a deal size equal to or larger than 250 million dollar. I base this value on the IPO statistics report (Ritter, 2014). In this report, the average deal size of NYSE and Nasdaq IPOs in the time period from 2012 until 2014 equals 247 million dollar.
After the regression model from equation (1) is tested in parts and as a total, I use the following regression equation to test the coefficient negative:
𝟓 𝑦!= 𝛽! + 𝛽!𝑖𝑛𝑖𝑡𝑟𝑒𝑡𝑢𝑟𝑛! + 𝛽!𝑎𝑔𝑒!+ 𝛽!𝑓𝑒𝑚𝑎𝑙𝑒!+ 𝛽!ℎ𝑜𝑡𝑖𝑝𝑜!+𝛽!𝑑𝑒𝑎𝑙𝑠𝑖𝑧𝑒!+ 𝛽!𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 + 𝜀! Negative is a dummy variable that has a value of 0 when the total return on investment for an investor is positive and a value of 1 when the total return on investment of an investor is negative. This coefficient gives the relationship between incurring a negative return and quitting IPO investing. The coefficient total return on
caused by the high correlation between total return on investment and negative (Table 1).
To be able to measure the interdependent effect from the dummies age,
female, and deal size, I include interaction terms in separate regressions. For example,
being female can influence the probability of reinvestment, but also influences the return on investment made by an investor on the IPO shares. The interaction terms are added to the original model as stated in equation (1). The correlations between the interaction terms and the return on investment can be found in Table 2.
Table 2
Correlations between the interaction terms and the total return on investment.
The first variable I add is return x age: a continuous-continuous interaction term that shows the relation between the total return on investment and the age of the investor. Return x female shows the continuous-binary relation between the total return on investment and being female and return x size the continuous-binary relation between the total return on investment and the deal size of the first IPO in which the investor invested. The results of the logistic regressions without and with interaction terms can be found in Table 3.
Correlation matrix
interaction terms Return on investment
Return x age Return x female Return x size
Return on investment 1.00
Return x age 0.96 1.00
Return x female 0.33 0.35 1.00
Table 3
Logistic regression with the probability of reinvestment as the dependent variable. Regression 1 contains only the characteristics on the investment itself. Regression 2 includes the personal characteristics of each investor. Regression 3 also adds the characteristics of each specific IPO. Regression 4 adds the negative return dummy and leaves out total return of investment (imperfect multicollinearity). Regressions 5, 6 and 7 add interaction terms one-by-one. Robust standard errors between parentheses.
* is significant at the 10% level, ** is significant at the 5% level, *** is significant at the 1% level
The next statistical analysis (Table 5) uses the sample extended with the 37 investors in IPO shares that still keep their shares in their portfolios. To be able to fully
compare both analyses, I do the same logistic analysis as in Table 3. For
completeness, the descriptive statistics and correlation matrices of the extended dataset are given in Table 4.
Logit (0 = no future IPO investments, 1 = at least one future IPO investment)
Probability of reinvestment 1 2 3 4 5 6 7
Total return on investment 1.682*
(0.912) 1.756** (0.833) 1.694* (0.875) - 3.153 (2.499) 1.419 (0.950) -0,172 (1.302) Initial return for investor 2.615
(4.013) 2.243 (3.853) -0.396 (4.937) -0.288 (4.405) 0.077 (4.961) -0.721 (4.881) 1.090 (4.880) Age - -0.013 (0,015) -0.012 (0.016) -0.010 (0.016) -0.011 (0.015) -0.011 (0.016) -0.007 (0.015) Female dummy - 0.431 (0.714) 0.527 (0.636) 0.460 (0.594) 0.503 (0.628) 0.249 (0.762) 0.627 (0.630)
Hot IPO dummy - - -1.326*
(0.730) -1.428** (0.715) -1.316* (0.738) -1.332* (0.723) -1.263 (0.781)
Deal size dummy - - -1.113**
(0.542) -1.141** (0.542) -1.149** (0.551) -1.053* (0.553) -1.225** (0.533)
Negative return dummy - - - -0.902*
(0.469)
- - -
Return on investment x age - - - - -0.028
(0.051) - - Return on investment x female - - - 4.708 (3.257) -
Return on investment x size - - - 2.911*
(1.578)
Pseudo-R2 0.044 0.048 0.119 0.116 0.120 0.128 0.135
Table 4
These tables give the descriptive statistics for the investors and the IPOs in the sample (extended dataset) and the correlations between the main variables and the interaction variables in the model. The observations do not equal 203 in every criterion because of missing values.
Investors, N = 203 percentiles
Obs Mean Median SD 25% 75%
Return on investment (%) 203 0.99 -1.43 28.83 -10.26 7.00
Initial return (%) 203 -2.85 -2.26 6.03 -5.81 0.96
Age 200 46.02 44.5 14.04 34.5 56
Female dummy 200 0.065 0 0.24 0 0
Negative return dummy 203 0.596 1 0.492 0 1
IPOs, N = 45 percentiles
Obs Mean Median SD 25% 75%
Deal size ($mm) 202 11,850.76 16,006.88 8,645.66 576.87 16,006.88 Initial return/underpricing (%) 202 21.20 0.61 28.54 0.61 38.07 Correlation matrix Return on investment Initial return Age Female dummy Negative dummy Size dummy Hot IPO dummy Return on investment 1.00 Initial return 0.31 1.00 Age 0.08 0.10 1.00 Female dummy - 0.02 0.08 0.22 1.00
Negative return dummy - 0.60 - 0.40 - 0.06 - 0.03 1.00
Deal size dummy - 0.12 - 0.13 - 0.01 - 0.03 0.02 1.00
Hot IPO dummy 0.08 0.06 0.07 0.08 - 0.14 0.36 1.00
Correlation matrix
interaction terms Return on investment
Return x age Return x female Return x size
Return on investment 1.00
Return x age 0.96 1.00
Return x female 0.26 0.29 1.00
Table 5
Bivariate logistic regression with the dataset extended with returns of investors that have still shares in portfolio (n=203; IPOs=45). Sell price used is closing price on November 28, 2014. Same order of regressions as in Table 3, interaction terms added in Regression 5, 6, and 7. Robust standard errors between parentheses.
* is significant at the 10% level, ** is significant at the 5% level, *** is significant at the 1% level
Logit (0 = no future IPO investments, 1 = at least one future IPO investment)
Probability of reinvestment 1 2 3 4 5 6 7
Total return on investment 0.396
(0.579) 0.453 (0.558) 0.457 (0.705) - 0.985 (2.369) 0.229 (0.764) -0.596 (0.933) Initial return for investor 3.354
(3.954) 2.892 (3.821) 1.583 (4.200) 0.665 (4.210) 1.742 (4.247) 1.142 (4.306) 2.524 (4.265) Age - -0.008 (0.014) -0.009 (0.015) -0.008 (0.015) -0.008 (0.015) -0.008 (0.016) -0.005 (0.015) Female dummy - 0.463 (0.689) 0.617 (0.606) 0.580 (0.584) 0.611 (0.608) 0.197 (0746) 0.663 (0.607)
Hot IPO dummy - - -1.190*
(0.702) -1.257 (0.692) -1.187* (0.705) -1.187* (0.696) -1.106 (0.744)
Deal size dummy - - -1.271**
(0.505) -1.309 (0.490) -1.294** (0.517) -1.203** (0.513) -1.383*** (0.511)
Negative return dummy - - - -0,478
(0.453)
- - -
Return on investment x age - - - - -0.010
(0.041) - - Return on investment x female - - - 4.915** (2.031) -
Return on investment x size - - - 1.718
(1.223)
Pseudo-R2 0.010 0.012 0.107 0.111 0.107 0.119 0.116
V. Results and Discussion
A. Results complete dataset
The first statistical analysis in this study is the logistic regression on the complete dataset (Table 3). In the first three models (Regression 1–3), the coefficient total
return on investment has a significantly positive effect on the probability that an
investor reinvests in an IPO. In the models with added interaction terms (Regression 5–7), total return on investment is no longer significant because of high correlation with the interaction variables (Table 2).
The economic effect of this result is determined by calculating the probability of reinvestment with the logit model:
𝟔 1
1 + 𝑒!(𝛽0+𝛽1𝑋𝑖+𝛽2𝑋𝑖+…+𝛽𝑘𝑋𝑖)
Assume a male investor with the average age of 47 making the average initial return of -3.01% in a hot IPO with a deal size larger than 250 (Regression 3). Firstly, I assume his total return on investment is 1%. In this case, his probability of reinvestment equals:
𝟕 1
1 + 𝑒!(1.11∗+1.69∗0.01−0.40∗−0.03−0.01∗47+0.53∗0−1.33∗1−1.11∗1)=13.41%
* in which 1.11 is the constant intercept of Regression 3 of Table 2
When his total return on investment increases to 2% and everything else is held equal, the probability of reinvestment increases to 13.61%. This is an increase in probability of 0.20 percentage points per additional percent of return on investment.
From the first three regressions and from the economic effect I can conclude that the total return that an investor has made on his investment in IPO shares is a significant predictor of the probability that he will reinvest. This is consistent with the reinforcement learning theory that describes that personally experienced outcomes have the greatest impact in decision-making. Furthermore, this outcome is consistent with the outcome of the studies of Kaustia and Knüpfer (2008) and Chiang et al. (2011) in respectively Finland and Taiwan.
The significant coefficient and the positive economic effect for total return on
investment confirm hypothesis I of this study: investors that incur positive returns will
invest in future IPOs with a higher probability than investors that incur negative returns.
The coefficient initial return for investor is not significant in one of the models. In Regression 1, 2, 5 and 7 the positive value of the coefficient shows that a positive return on the offering date increases the probability that an investor invests in a future IPO. This is consistent with the study of Kaustia and Knüpfer (2008) but not at the significant level. In the remaining regressions (3, 4, 6) initial return for investor has a negative coefficient. This means that a positive initial return decreases the probability of reinvesting in IPO shares. This contradicts any literature. Thus, the initial return is not a decisive determinant in decision-making.
The personal characteristics in the model consist of age and female. The negative coefficient on age shows that people with a higher age are less likely to reinvest in an IPO. This is consistent with the study of Pårsson (1996) on risk aversion and the study of Barber et al. (2011) on the correlation between trading intensities and performance. Though, this effect is not significant.
Against my expectations and contrary to the existing literature, the coefficient for women in the sample is positive. That means that being female increases the probability of reinvesting in a future IPO. Though, this coefficient is not significant. This means that there is no evidence that there is a gender difference in probabilities for reinvestment holding everything else equal.
In addition to that, there was a small amount of women in the sample (n=15 of n=166). As a consequence, it is hard to measure any difference caused by gender.
To summarize, the personal characteristics are not significant. This means that those coefficients have in themselves no significant influence on investment
decisions. The interdependent influence of these effects on the return on investment and the probability of reinvestment are discussed in the interaction terms section below.
The next coefficient is the dummy for hot IPOs, or the underpricing. This dummy is negatively significant in all the regressions except for Regression 7. This implicates that an investor that invested in a hot IPO for the first time is less likely to reinvest in any IPO in the future.
The economic effect of investing in a hot IPO is measured assuming an investor with similar characteristics as above obtaining a total return on investment of 1%. The probability assuming a hot IPO (hot IPO dummy = 1) equals 13.41% (eq. (7)).
The probability of reinvestment in Regression 3 assuming a cold IPO (hot IPO
dummy = 0) calculated as in equation (6) equals 36.84%. Investing in a hot IPO
decreases the probability of reinvestment with 23.43 percentage points.
This is the contrary outcome as I hypothesized in hypothesis V and contradicts the literature of Kaustia and Knüpfer (2008) that suggest that investors that once invested in a hot IPO are more likely to reinvest in the future. Therefore, hypothesis V is not accepted.
The dummy on deal size has a significantly negative influence on the probability of reinvestment in all the bivariate regressions. This implicates that a larger deal size (>250 million) reduces the probability that an investor reinvest in an IPO in the future.
The probability of reinvesting for the average investor as stated above
investing in an IPO with a deal size larger than 250 million dollar equals 13.41% (eq. (7)). The probability of reinvesting when the first IPO was an IPO with a deal size smaller than 250 million dollar (deal size dummy = 0) equals 32.04%. Investing in an IPO with a deal size larger than 250 million dollar decreases the probability of reinvestment with 18.63 percentage points. This is contrary to hypothesis VI.
The literature from Cook et al. (2006) and Wang (2012) describes higher initial returns for IPOs that got more media attention. Possibly this attention drives up the prices. Wang (2012) describes that investors overreact to this by buying too expensive IPO shares and leaving money on the table. This is an explanation for the negative influence of the deal size on the probability of reinvestment.
In Regression 4 the dummy for negative returns is tested. The coefficient for
negative returns is negatively significant. That means that a negative return on
investment reduces the probability of reinvestment.
The economic effect of this outcome is calculated with Regression 4: a male investor with the average age of 47 invests in a hot IPO with a deal size larger than 250 million dollar and obtains the average initial return of -3.01%. When he incurs a positive return on his first IPO investment (negative dummy = 0), the probability of reinvestment equals:
𝟖 1
1 + 𝑒!(1.65∗−0.29∗−0.03−0.01∗47−1.43∗1−1.14∗1−0.90∗0)=9.23%
The probability of reinvestment when he incurs a negative return (negative
dummy = 1) equals 20.05%. Thus, a negative total return on investment decreases the
probability of reinvestment with 10.82 percentage points. This result is consistent with hypothesis II: investors that incur a negative return on their first IPO investment quit IPO investing with a higher probability.
B. Interaction terms in the complete dataset
To estimate the interdependent effect of the dummy variables age, female, and size on the total return on investment, interaction terms are added to the model one by one.
In Regression 5, an interaction term for total return on investment and age is added. The interaction term return on investment x age is not significant. This means that there is no difference from the regression without this interaction effect. The coefficient for return on investment x age is negative, consistent with the literature of Pårsson (1996) and Barber et al. (2011), but this effect is not significant. Age does not significantly influence the return on investment and the interaction between those terms does not significantly influence the probability of reinvestment, thus hypothesis III is not accepted.
In Regression 6 an interaction term for return on investment and female is added to the complete model from Regression 4. This coefficient is not significant. Though, the coefficient has a positive value, meaning that being female positively influences the return on investment. This is consistent with the study of Barber & Odean (2001) that finds that women perform better than men, but the effect is not significant. This result shows that hypothesis IV is not accepted: there is not enough evidence to assume that there is a significant gender difference.
In Regression 7 I add an interaction term for return on investment and size. This interaction term is significant at the 10% level when added to the model. Therefore, there is evidence that the deal size of an IPO affects the return on investment and that the interaction between those terms has a positive influence on the probability that an investor reinvests in IPO shares at a future date.
The economic effect is measured with Regression 7. The probability of reinvestment for the investor with the same characteristics as above with the interaction term included equals:
𝟗 1
1 + 𝑒!(0.97∗−1.17∗0.01+1.09∗−0.03−0.01∗47−1.26∗1−1.23∗1+2.91∗1)=75.29%
If the interaction term return on investment x size is omitted from the regression, the probability of reinvestment equals 13.24. Thus, a positive return on an IPO with a deal size larger than 250 million dollar increases the probability of reinvestment with 62.05 percentage points. This is consistent with the study from Wang (2012). This result is contrary to the significantly negative result of the deal size dummy. This means that an investor is less likely to reinvest in IPO shares when he invests in an IPO with a deal size larger than 250 million dollar, though, when he obtains a positive return with this large deal size IPO investment, he is more likely to reinvest in a future IPO. The latter conclusion is consistent with hypothesis VI. The hypothesis is partly correct.
The conclusion from Table 3 is that the total return on investment, the dummy on hot IPOs, the dummy on deal size, and the interaction effect between the total
return on investment and the deal size are significant predictors of the probability that
an investor reinvests in a future IPO.
C. Limitations
Table 6
Statistics retrieved from total Lynx database with all trades within the period January 2, 2012 until November 28, 2014. 39 IPOs used from complete dataset (see descriptive statistics Table 2). First week is defined as the six days after the offering date, first month as the 30 days after the offering date.
Table 6 shows statistics of the additional IPO investments done in the first week after the offering date as well as the first month after the offering date. The results show that a lot of investors do not invest on the offering date itself, but in the days or the month after that. Especially the IPOs with large deal sizes have additional IPO investments in this time period. This implicates that a lot of investors impose a “wait and see what happens”-strategy. This could be linked to the rational Bayesian learning theory (Kaustia & Knüpfer, 2008). Most investors tend to wait until there is more information available on the stock. In this way, they can include this information to
Additional IPO investments (IPOs=39)
First week after offering date 84 First month after offering date 324 Total additional investments 408
make a more rational decision. This is consistent with the general assumption in finance that the average investor is risk-averse.
Other factors that are not included in the model but could have an impact on decision-making are liquidity constraints of particular investors, the time since their previous IPO investment, their investment strategy, and the industry of the IPO firm. Moreover, a sample with an equal amount of men and women can give more precise results regarding to a possible gender difference.
Besides that, it is possible that Lynx clients have an additional trading account at a different bank or brokerage firm. I am not able to retrieve the trades the investors do at other investment firms.
D. Comparison to the extended dataset
Table 5 reflects the results of the logistic regressions done on the extended dataset (n=203; IPOs=45). These regressions are a robustness check and serve as comparison for the results of Table 3.
I start with the consistencies with the complete dataset and discuss the inconsistencies after that. In both models, the initial return for investor is not a significant predictor of the probability of reinvestment. Though, in the extended model (Table 5) the coefficient stays positive, as expected. A positive initial return increases the probability of reinvestment, but not significantly.
The personal characteristics age and female are not significant in the extended model either. The dummy on hot IPOs stays a significant predictor (except for
Regressions 4 and 7). Furthermore, the dummy on deal size stays very significant in the extended model (expect for Regression 4) and has negative coefficients just like the coefficients in Table 3. In addition to that, the interaction term return on
investment x age is not significant in the bivariate model of the extended dataset. This
is also the case in the complete dataset.
There are also inconsistencies in the results between the complete and the extended dataset. The first is that the total return on investment is not significant in any of the regressions in the extended model.
Furthermore, the interaction terms lead to different results in both analyses. The interaction term return on investment x size is not longer significant, as it was in Table 3.
Contrary to the first analysis, return on investment x female is significant in Regression 6 of the second analysis (Table 5). This is consistent with the study of Barber & Odean (2001) mentioned above.
To obtain the economic effect of return on investment x female, I calculate the probability of reinvestment for a female investor with the average age of 47. She makes a total return on investment of 1% and an initial return on investment of -3.01%. If she invests in a hot IPO with a deal size larger than 250 million dollars with the interaction term return on investment x female included, her probability of reinvestment equals:
𝟏𝟎 1
1 + 𝑒!(0.69∗+0.23∗0.01+1.14∗−0.03−0.01∗47+0.20∗1−1.19∗1−1.20∗1+4.91∗1)=95.25%
* in which 0.69 is the constant intercept of Regression 6 of Table 4
The outcome of this regression assuming an otherwise comparable man (female and
total return on investment x female = 0) equals 9.96%. Thus, the probability of
reinvestment for a female investor that incurs a positive return on her first IPO investment increases with 85.29 percentage points compared to a male investor with similar characteristics.
To find the reason for the inconsistencies in the extended dataset, I compare the descriptive statistics. The mean return on investment increases from -2.85% to 0.99% in the extended dataset. More than two-third of the (created) total returns of investment for the 37 investors I add in the extended dataset are positive returns yielding from a range of 8.22% to 155.63%. Miller (2000) finds that most IPOs underperform the indices in the first years. After that, the shares are slowly better performing. This better performance is shown by the positive total returns on investment calculated with the closing price at November 28, 2014.
The descriptive statistics show that these results positively bias the statistics for the coefficient total return on investment. The average return in the extended sample is even positive. This difference in results partly explains why the total return
on investment is upward biased and not significant in this regression anymore.
Also more than two-third of the 37 added investors invests in IPOs with a deal size larger than 250. The deal size variable also experiences an upward bias, which makes it not significant anymore.
The amount of women in the sample increases only from 12 to 13 in the extended sample. Because of the on average positive total return on investment, there
is a significant relationship in the extended dataset between female investors incurring a positive return and the probability of reinvestment (interaction term).
The conclusion for the extended model is that the dummy on deal size, the dummy on hot IPOs, and the interaction term for return on investment x female (gender difference) are significant predictors of the probability of reinvestment.
VI. Conclusion
This study investigates whether the future investment decisions of investors in IPOs at online broker Lynx are affected by past performance. The results of this study have implications for the literature on behavioral finance and show the importance of psychology in finance.
In the first part, I use data on 166 investors that invested in 39 IPOs on either the New York Stock Exchange (NYSE) or Nasdaq in a three-year period. The logistic regression shows that there is significant evidence that investors that incur positive returns on investment will invest in future IPOs with a higher probability than investors that incur negative returns. The investors that incur a negative return on their first IPO investment are more likely to quit IPO investing.
Moreover, investors that invest in an IPO with a positive initial return on the offering date (hot IPO) are less likely to reinvest in the future. Also investors with a first investment in an IPO with a deal size larger than 250 million dollar are less likely to reinvest, except for the case in which they incur a positive total return on
investment. Thus, the interaction between a large deal size and the total return on investment has a significantly positive influence on the probability to reinvest.
From these results I can conclude that there exists reinforcement learning. The probability of reinvestment, and thus the decision for future investment, is influenced by the return an investor obtained in the past. This result is consistent with the studies of Kaustia and Knüpfer (2008) and Chiang et al. (2011) in respectively Finland and Taiwan.
In the extended dataset (n=203; IPOs=45), the results are upward biased. Consequently, the total return on investment is no longer a significant predictor of the probability of reinvestment in the extended dataset. Though, the hot IPO dummy and the dummy for a large deal size stay significantly negative predictors.
Inconsistent with the complete dataset, the interaction effect between the total return on investment and being female is a significant predictor. The extended dataset
yields enough evidence to assume that there is a gender difference: women that incur positive returns on their first IPO investment are more likely to reinvest in a future IPO than otherwise comparable men. This result significantly predicts the probability of reinvestment in an IPO and is consistent with the study of Barber & Odean (2001).
VII. Recommendations
A. Further research
This study has several implications for the literature on behavioral finance and
especially on decision-making of individual investors. For future studies, a model that takes into account all the returns a specific investor makes on all his IPO investments ever done can give more precise results. In this study only the first IPO investment is included in the model.
The more developed model can analyze the interdependent influence of the past returns on the following and so onwards. In this way it can be studied if frequent IPO investors gain increasing returns and thus can tell even more about the presence of learning effects and the importance of past performance.
Another recommendation for future research is a larger and more diversified dataset. There may be a difference between learning effects between countries, races or even brokerage firms. Because there is no Central Security Depository in The Netherlands, it was not possible for me to retrieve data on all Dutch investors. Doing so increases the external validity of this study and the other studies on this subject.
B. Directions for future investigations
As a result of this study, there are topics that can be studied in the future to fill in gaps in our understanding. It would be interesting to, besides looking at the numerical data, perform a questionnaire on investment strategy for IPO investors. In this way, we can obtain a better understanding for the motivation of investment decisions that do not have a numerical logic. This information can help further explain several (bad) choices investors made and can clarify why investors keep investing in IPOs.
In addition that, a study that includes the limitations mentioned in section V.C. (page 23/24) can enrich the knowledge on this subject.
VIII. References
Barber, B. M. and Odean, T. (2000). Trading is hazardous to your wealth: The common stock investment performance of individual investors. Journal of
Finance, 55(2), pp. 773–806.
Barber, B. M., & Odean, T. (2001). Boys will be boys: gender, overconfidence, and common stock investment. Quarterly Journal of Economics, 116(1), pp. 261–
292.
Barber, B. M., & Odean, T. (2007). All That Glitters: The Effect of Attention and News on the Buying Behavior of Individual and Institutional Investors. Review of Financial Studies, 21(2), pp. 785–818.
Barreto, E. (2014). Alibaba IPO ranks as world’s biggest after additional shares are
sold. Retrieved on November 11th, 2014 from http://www.reuters.com/article/
2014/09/22/us-alibaba-ipo-value-idUSKCN0HH0A620140922
Benninga, S., Helmantel, M. & Sarig, O. (2005). ‘The Timing of Initial Public Offerings’, Journal of Financial Economics, 75(1), pp. 115–132.
Chiang, Y., Hirshleifer, D., Qian, Y., Sherman, A.E. (2011). ‘Do Investors Learn from Experience? Evidence from Frequent IPO Investors’, The Review of
Financial Studies, 24(5), pp. 1560–89.
Cook, D. O., Kieschnick, R., & Van Ness, R. A. (2006). On the marketing of IPOs. Journal of Financial Economics, 82(1), pp. 35–61.
Fong, W.M. (2014). The Lottery Mindset: Investors, Gambling and the Stock Market.
Hampshire: Palgrave Macmillan.
Grinblatt, M. & Keloharju, M. (2009). Sensation Seeking, Overconfidence, and Trading Activity, The Journal of Finance, 64, pp. 549-578.
Hsu, Y., Shia, C. (2010). ‘The overconfidence of investors in the primary market’,
Pacific-Basin Finance Journal, 18(2), pp. 217–239.
Kaustia, M. and Knüpfer, S. (2008). ‘Do Investors Overweight Personal Experience? Evidence from IPO Subscriptions’, The Journal of Finance, 63(6), pp. 2679–
2702.
Korniotis, G.M. and Kumar, A. (2011). ‘Do Older Investors Make Better Investment Decisions?’, The Review of Economics and Statistics, 93(1), pp. 244–265. Linnainmaa, J.T. (2011). ‘Why Do (Some) Households Trade So Much?’, The Review
of Financial Studies, 24(5), pp. 1630–1666.
Lowry, M. and Schwertipo, G.W. (2002). ‘Market Cycles: Bubbles or Sequential Learning?’, The Journal of Finance, 57(3), pp. 1171–1201.
Miller, E.M. (2000). ‘Long run underperformance of initial public offerings: an explanation’, working paper, University of New Orleans
Pålsson A. (1996). ‘Does the degree of relative risk aversion vary with household characteristics?, Journal of Economic Psychology, 17(6), pp. 771–787. Pástor, L. and Veronesi, P. (2005). ‘Rational IPO waves’, The Journal of Finance,
60(4), pp. 1713–1757.
Ritter, J.R. (2014). Initial Public Offerings: Updated Statistics; Table 1 (updated
November 30, 2014). Retrieved on January 15th
, 2015 from http://bear.warrington.ufl.edu/ritter/IPOs2013Statistics.pdf
Schubert, R., Brown, M., Gysler, M. & Brachinger, H. W. (1999). 'Financial Decision-Making : Are Women Really More Risk-Averse ? The American
Economic Review, 89(2), pp. 381–385.
United States Bureau of Economic Analysis (2014). Table 1.1.1. Percent Change
From Preceding Period in Real Gross Domestic Product. Retrieved on
December 14th
, 2014 from http://www.bea.gov/iTable/
iTable.cfm?ReqID=9&step=1#reqid=9&step=3&isuri=1&903=1
Wang, J. (2012). ‘Investment behaviours and IPO returns: evidence from Taiwan’,
Applied Financial Economics 22(16), pp. 1385–1394.
Websites used to fill missing values in database: - Yahoo Finance:
http://finance.yahoo.com
- Renaissance Capital IPO Center:
