The Impact of Market Conditions on Insiders’ Behavior at
the Expiration of an IPO Lock-up Contract: Empirical
Evidence from the US Stock Market
Nadine Aguili S2396599
Supervisor: Prof. Dr. C.L.M. Hermes
EBM866B20: Master Thesis Finance (2013-2014-2) June, 26th 2014
Abstract
This thesis investigates the impact of market conditions on the behavior of insiders by analyzing the abnormal returns measured at the expiration of a lock-up contract. I find a -1.40 percent significant abnormal return at the event day. The drop in the share price is caused by the additional supply that insiders create by selling their shares. Comparing the abnormal return for a pre-crisis and crisis period I find that the former experiences larger negative abnormal return, but this result is not significant. With the regression analyses I find no evidence that market conditions influences the behavior of insiders.
2
1.
Introduction
When a firm decides to go public, through a so-called Initial Public Offering (IPO), this event is often combined with a lock-up contract. A lock-up contract is used to restrict insiders from selling their shares in the firm for a specific period of time. Existing literature find that after such a lock-up period, the firm’s share price tends to drop (Ofek and Richardson, 2000; Field and Hanka, 2001; Garfinkle, Malkiel and Bontas, 2002; Brau, Carter, Christophe and Key, 2004; and Nowak, 2004). There is consensus in the literature that this drop in the share price is caused by the additional supply that insiders are able to provide after the lock-up period. In this thesis, I want to investigate whether the level of additional supply by insiders is affected by market conditions, in particular whether during periods of increasing stock returns insiders supply more of their shares to the market. To do so, I compare the abnormal stock returns at the expiration day between bullish (pre-crisis) and bearish (crisis) market periods.
There are various reasons for the existence of lock-up contracts. The main reason is to alleviate informational asymmetries of IPOs. First there is an adverse selection problem; since insiders have superior knowledge of the firm, outside investors may not be able to assess the fair price of the firm, as they cannot distinguish between high- and low-quality firms. This may lead to a lower share price for high-quality firms and may be a reason for insiders of high-quality firms to signal their quality through a up contract (Brau, Lambson and McQueen, 2005). The lock-up contract forces the insiders to hold shares of equity of the firm, thereby revealing their trust in the firm to outside investors. Second, there are potential moral hazard problems after the IPO. The possibility of insiders to sell their shares, may lead to the situation that management and investors’ interest are no longer aligned. This can cause IPO under pricing, as investors may shy away from investing in firms with potential moral hazard problems. The lock-up contract commits insiders to the firm (Brav and Gompers, 2003) and leads to an aligned interest of management and investors.
3 undiversified portfolio because a relatively large proportion of their portfolio exists of shares from the IPO. By selling some or all of their shares in the firm, they will have a more diversified and thus less risky portfolio. Another reason may be explained by the market conditions during the expiration of a lock-up contract. Investors tend to sell their shares when stock markets experience increasing returns (Shefrin and Statman, 1985; and Odean, 1998; Henker and Owen, 2008). Thus, a bullish market can be a reason for insiders to sell their shares in the firm.
For this research I will be using US data. I have a sample of 579 firms of which 270 firms have a lock-up contract which expires during the crisis period (bearish market) and 309 firms have a lock-up contract which expires during the pre-crisis period (bullish market). Appendix 1: The S&P 500, the boom and crash market presents the performance of the S&P 500, a clear upward moving trend is shown from 2003 until October 2007, which I will refer to as the pre-crisis period. After October 2007 a clear downward moving trend is shown until March 2009, which I will refer to as the crisis period. For this research I will use a two-stage approach, first I will conduct an event study, to see whether there is abnormal return around the expiration day. By conducting a t-test and a non parametric test I will compare the abnormal returns from the pre-crisis and the pre-crisis period. For the second stage I will use a multiple regression analyses to see whether the abnormal return can be explained by the variables: volatility of the firm, the market value of the firm, whether the firm is traded on NASDAQ and whether the lock-up contract expires during the crisis.
4
2.
Literature review
2.1 Theoretical contribution
A lock-up contract is a legally binding contract between underwriters and insiders of a company, which prohibits insiders from selling their shares to the market. Lock-up periods are usually around 180 days. The reason why underwriters impose such a contract on insiders, is to minimize the risk of having to sell the shares for a lower price than they were bought. Insiders can be defined as the company’s founders, initial investors, executives, managers and employees. Insiders agree to such a lock-up contract, because it ensures an element of stability in the stock's price in the first few months of trading. This stability is very important for a firm, since the IPO is being executed to raise funds. When a company is stable during the first months of trading, investors will be interested to invest, while if the company is unstable investors will allocate their money to other projects.
5 The second problem associated with IPOs is moral hazard. Once the founders of a company decide to go public through an IPO, the interests of the founders and the management may start to diverge. Since an IPO is the first opportunity for the firm’s founders and initial investors to realize the value of their ownership, the personal wealth that can be manipulated from an IPO is very tempting and the long- term interests of founders may become short-term (Hurt, 2004). To prevent insiders from using the IPO to exploit outside investors the lock-up agreement is used as a commitment device. A firm whose moral hazard incentives in the aftermarket are likely to be large should accept a longer lock-up contract to convince the public to buy their shares. The insiders’ ability to take advantage of the investors is reduced by these lock-up contracts and investors will be more willing to buy the shares.
The last point of Brav and Gompers (2003) deals with the underwriter’s ability to extract additional compensation. As stated before the contract is an agreement between the underwriter and the insiders of a firm. The contract only allows insiders to sell their shares prior to the lock-up expiration date if the underwriter agrees to it. When the insider has the approval of the underwriter to sell the shares within the six months after the IPO the insider is forced to do a block trade (trading large quantities at once) or a Seasoned Equity Offering (SEO) through the lead underwriter. The underwriter will make additional fees when insiders make use of either one of these two options. The longer the lock-up contract between the underwriter and the insider, the more additional compensation the underwriter can extract.
6 known and expected it should not have a negative effect on the shares price. These negative abnormal returns can be explained by two theories. The first theory is the larger-than expected insider sales of Field and Hanka (2001), they find that the expectation and the realized insiders sales differ. The second theory is the downward sloping demand curve suggested by Ofek and Richardson (2000). The supply shock caused by the expiration of the lock-up contract shifts the equilibrium point to a higher quantity of shares sold and a lower share price. In both cases, the increase in the supply of shares is caused by the selling behavior of insiders.
As discussed in the introduction, reasons for insiders to sell their shares are to diversify their portfolio and to collect their profit. Since profits are collected in markets with increasing returns, market conditions play an important role in the decision making process of investors. In behavioral finance literature, some research is done about the decision making process of investors. These papers will be discussed in the empirical study following this paragraph. The main conclusion of these papers is that when investors experience a loss (crisis) they do not want to collect these losses and hold on to their shares, hoping the market will turn. When investors experience a win (bullish market) they are more willing to sell their shares and collect their profits (Shefrin and Statman, 1985; and Odean, 1998; Henker and Owen, 2008).
2.2 Empirical results
7 For the second reason moral hazard, Brav and Gompers (2003) develop a commitment hypothesis, which states that longer lock-up contracts commit insiders to the company and reduces moral hazard. They test this hypothesis by conducting a multiple regression analyses to see whether the lock-up contract is longer for firms with a higher probability of moral hazard. This probability is higher for companies which: (i) are not financed by venture capital, (ii) have low-quality underwriter, (iii) a small firm size, (iv) have low book-to-market ratios, and (v) have a high price volatility. They find that large firms, firms with higher-quality underwriters, and firms backed by venture capital all have shorter lock-up contracts. This result can be explained by the fact that firms with high-quality underwriters or venture capital financing are unlikely to take advantage of outside investors and therefore have less need to commit to longer lock-up contracts. Firms with lower book-to-market ratios and high price volatility have longer lock-up contracts. These results are in line with the commitment hypothesis. Bessler and Kurt (2007) research agency problems for IPOs in the German Market. In line with Brav and Gompers (2003) they find that venture capital firms are usually in a position to signal the superior quality of an IPO, thereby reducing moral hazard.
For the last reason rent extraction, Brav and Gombers (2003) develop the hypothesis that investment banks are able to extract additional compensation by imposing a longer lock-up contract. This could be through forcing early sales by insiders to be traded by the lead underwriter or by having the firm do a Seasoned Equity Offering (SEO) through the lead underwriter of the IPO. They find the profits made by underwriters on early sales are very low and that firms are more likely to switch underwriter when the underwriter of a SEO has a higher rank. So they find no evidence for this hypothesis.
8 transparency of a firm. Thus, the conclusion of Brav and Gompers (2003) for the commitment hypothesis that large firms, firms with higher-quality underwriters, and firms backed by venture capital all have shorter lock-up contracts, should also apply to the signaling hypothesis. Brau, Lambson and McQueen (2005) develop their own signaling model, which predicts that lock-up contracts will be shorter for transparent firms and firms with high idiosyncratic risk. They find evidence for their signaling model and thus for the fact that lock-up contracts can serve as a signaling device.
Now that the mechanisms of a lock-up contract are explained, I will discuss the expiration of this contract. This event is of great importance to this thesis, because insiders are able to sell their shares. As discussed in the theoretical contribution, the supply created by investors is associated with negative abnormal return. Ofek and Richardson (2000), Field and Hanka (2001), Garfinkle, Malkiel and Bontas (2002), and Brau, Carter, Christophe and Key (2004) investigate the price impact on stocks at the expiration date of lock-up contracts in the US market. They all find a negative abnormal return and a permanent increase in trading volume around the expiration day. Ofek and Richardson (2000) argue that this evidence is consistent with the downward sloping demand curve hypothesis. This hypothesis predicts that demand curves are downward sloping, which causes a permanent shift in supply to lead to a permanent drop in the share price. These effects are permanent because the shares sold by insiders stay in the market. Nowak (2004) investigates the impact of the expiration of the lock-up contract on price and trading volume of shares for the German market. He finds a significant negative abnormal return and an increase in trading volume of 25 percent. This result is in line with the downward sloping demand curve discussed by Ofek and Richardson (2000).
9 In the behavioral finance literature Kahneman and Tversky (1979), Shefrin and Statman (1985) and Odean (1998) researched the behavior of investors during periods of losses. Kahneman and Tversky (1979) presented the prospect theory, which describes the way people choose between two options involving risk, it states that people will base decisions on potential value of gains rather than potential value of losses. Shefrin and Statman (1985) state that investors have the tendency to sell shares whose price has increased, while keeping shares that have dropped in value, this is called the disposition effect. Investors are less willing to recognize losses, but are more willing to recognize gains. Odean (1998) agrees with the study of Shefrin and Statman (1985) and also states that investors realize their profitable stocks investments at a much higher rate than their unprofitable ones.
In the field of psychology some researchers discuss investor’s behavior during certain market conditions. Henker and Owen (2008) state that during a crisis investors will mainly hold on to their shares, hoping the trend will reverse. Another result of researches in this area, which were mainly lab experiments, is that investors who experience losses are more risk taking (Hold and Laury, 2005; Seo, Goldfarb and Barrett, 2010). Gong, Lei and Pan (2013) researched the effect of booms (pre-crisis) and crashes (crisis) in the asset market on investor’s trading behavior. They recruited active investors to participate either in the boom treatment or the crash treatment and observed their trading behavior. The crash treatment in their research is based on the financial crisis of 2007. Their research concludes that the number of trades increases when participants experience a booming market compared to participants that experience a crashing market.
10 2.3 Hypotheses
Despite the efficient market hypothesis, which states that all relevant and known information should be reflected in the share price, previous research found that at the expiration date of a lock-up contract, which is a fully known and expected event, stock prices experience negative abnormal returns. By conducting an event study, I will test the abnormal return of my data sample to see whether they experience abnormal negative return at the expiration day of a lock-up contract. This leads to my first hypothesis:
H1: There is abnormal return measured at the expiration day of the lock-up contract
Since abnormal return is related to the additional supply insiders can create during the expiration of a lock-up contract, I will divide my sample into a pre-crisis and crisis period. By comparing these two samples I can see whether the abnormal returns measured during the first test is higher for shares unlocked during the pre-crisis period. This leads to my second hypothesis:
H2: The pre-crisis period shows larger negative returns compared to the crisis period
Some researchers discuss the trading behavior of investors during certain market conditions. The main conclusion of these papers is that investors are more willing to sell their shares during a period of increasing returns (pre-crisis) compared to a period of decreasing returns (crisis). I will conduct a multiple regression analyses to see whether market conditions influence the behavior of insiders. This leads to my third, and last, hypothesis:
11
3.
Methodology
3.1 Event study methodology
For the event study methodology I used the paper of Brown and Warner (1980). They research various methodologies which are used in event studies to measure security price performance. They find that the market model and the simpler models that do not adjust for market factors or risk all perform well. Because most research about lock-up contracts use a market based model (Ofek and Richardson, 2000; Field and Hanka, 2001; Garfinkle, Malkiel and Bontas, 2002; and Nowak, 2004) I will use the Market adjusted return as explained in the paper of Brown and Warner (1980).
First, I obtained the daily adjusted1 stock prices and daily market returns from DataStream. These prices and returns are used to calculate the continuously compounded returns. The reason I used continuously compounded returns (instead of simple returns) is that the frequency of the compounding of the return does not matter, since it is continuously. This makes it easier to compare returns across assets. The returns are calculated by the following formula from Brooks (2008):
Where Rt denotes the continuously compounded return at time t, Pt denotes the asset price at time
t, and ln denotes the natural logarithm.
Secondly, I will use the continuously compounded returns to calculate the abnormal returns. As discussed above I will use the market adjusted method from the paper of Brown and Warner (1980). The formula is as follows:
Where is the return on stock i at time t and is the return on the market at time t. The market return used for this study is the return of the S&P 500.
12 Finally, to investigate the abnormal performance, a frequently used method is the cumulative abnormal return (CAR). Ofek and Richardson (2000), Field and Hanka (2001), Garfinkle, Malkiel and Bontas (2002) and Nowak (2004) all used the CAR to measure abnormal performance at the expiration day of the lock-up contract. The CAR is calculated as the Cumulative abnormal return of the previous event day plus the current value of the abnormal return. The formula obtained from Brown and Warner (1980) is as follows:
Where is the cumulative abnormal return of the previous event day and is the
abnormal return at day t.
To test whether the expiration of a lock-up contract affects the stock price, the CAR of the event will be compared with the “normal” returns. The normal returns are obtained by taking the average of the returns from the period before the event also known as the estimation window. The estimation window of this event study will be -180 (the first trading day) until -11 days. I will be using various event windows to see what the impact is before, after or during the event. The event windows used are:
- Over the whole event period (-10 to +10) - The period before the event (-10 to -1) - The three day event period (-1 to +1) - The period after the event (+1 to +10)
- The days around the event ( -2, -1, 0, +1, +2)
For testing the statistical significance of the abnormal returns, I will conduct two tests. 1) A t-test
2) The Wilcoxon Rank test
13 event window, to see if they are significantly different from zero. After the test for abnormal returns, I will repeat the two tests to compare the pre-crisis and crisis abnormal returns, to see whether the former has larger negative abnormal return.
3.2 Multiple regression methodology
For the third hypothesis I am using a multiple regression model. Previous research used several variables to explain the abnormal return measured during the expiration of a lock-up contract. The main variables used are: the market value, Nasdaq firms, percentage of locked shares (free float), the volatility of a share, venture capital financing, book-to-market ratio and underwriter reputation. For this paper I will only use the following variables as control variables in my regression: Market value, volatility, and Nasdaq firms. The reason that I did not use the other variables is because I could not find the data or the data I did find in the databases where incorrect. The regression will show whether the abnormal returns measured are influenced by the market conditions or if the effect can also be explained by the other control variables. The regression model is as follows:
Where is the cumulative abnormal return of the event day, market value is the firms value
according to the market, volatility is a risk measure for the changes in the value of the securities, the is a dummy variable which has a value of 1 if the company is listed on the NASDAQ and 0 otherwise and the is a dummy variable which has a value of 1 if the companies’ lock-up contract expires during the crisis and 0 otherwise.
14 3.3 Data description and sample collection
I will use the IPO showcase information from the Nasdaq website (NASDAQ, 2014) to identify my sample. I also use the website to identify the lock-up period and the lock-up expiration date. When the website of NASDAQ does not provide the information about the lock-up contract, I will hand-collect the information from the prospectuses, which can be obtained from the Orbis database. Stock return data are daily stock prices and daily market returns, these are collected from the database DataStream. An initial sample of 664 IPOs from the period January 2006 until March 2009 is collected.
For the pre-crisis period I use data from January 2006 until October 2007. The reason that I do not use the whole period of 2003 until 2007 is because the crisis period is much shorter. This may result in biased outcomes, i.e. the time period for the pre-crisis period would be five times longer than the crisis period. Within the initial sample there are some companies for which I cannot find an ISIN code, or for which the database (DataStream) cannot provide any information. These companies are excluded from the sample. After collecting all the data for which I can find the information the final sample consists of 309 companies for the pre-crisis period and 271 companies for the crisis period.
The variables for the regression analyses are obtained from the database DataStream or they were calculated manually. Market value is calculated by the following formula:
15 on the NASDAQ are expected to show higher abnormal returns. This is because NASDAQ double counts the trades made, which relaxes the rule 1442 restriction of not selling more than 1 percent of the trading volume. Firms with high volatility should experience higher abnormal returns surrounding the expiration day of the lock-up contract, because they cause larger uncertainties.
3.4 Descriptive statistics
Table 1 presents the descriptive statistics for the estimation window, the event windows and the explaining variables. At the event day (t0) the mean and median are negative, which indicates
that the majority of the returns on that day are negative. The Jarque-Bera test statistics are very large and significant for all variables, which means that the data is not normally distributed at a 95 percent significance level. The mean of the explaining variables show that 47 percent of the lock-up contracts expire during the crisis, 36 percent of the firms are traded on the NASDAQ, the firms in the sample have an average volatility of 23 percent and that they have an average market value of 8 (logarithmic value).
16
Table 1: Descriptive statistics
Descriptive statistics of the estimation window, event windows and the variables for the regression. There are 579 daily observations over the time period 01-2006; 02-2009. The estimation window is the average return of the “normal” period before the event day, the event windows are cumulative abnormal returns of different periods, denoted by the days from the event window (vb. t-10 to t+10). The
crisis dummy has a value 1 if the expiration day is during the crisis and 0 otherwise, the Nasdaq dummy has a value of 1 if the firm is traded on the NASDAQ and 0 otherwise, the log of market value is calculated by the log of the share price x shares outstanding at the IPO and the volatility is calculated as the standard deviation of the stock over the whole period -180 to +10.
Observations Mean Median Maximum Minimum
17
4.
Results
This part will present the results of the tests. I will start with the test output of the event study in paragraph 4.1. The event study is conducted for the whole period and for the pre-crisis and crisis period, I will use a t-test and a parametric test to see whether there is any abnormal return around the expiration day and to see whether the abnormal returns of the pre-crisis and crisis period differ. The results of the event study will give an answer to the first two hypotheses. After this the multiple regression results will be discussed in paragraph 4.2. First I will test the variables from the regression on multicollinearity to see whether the data can lead to biased results. After this the regression results are discussed.
4.1 Event study
This paragraph will present the results from the event study. Figure 1 shows a time series plot of the cumulative abnormal returns. This graph shows that the share price declines strongly around the lock-up expiration day. The continuously downward sloping trend in the line may be explained by the fact that insiders do not sell all their shares at ones at the expiration day, but that they do this over several days.
Figure 1: Cumulative abnormal return whole sample
For a sample of 579 observations the cumulative abnormal returns from day -10 until day +10 around the lock-up expiration day are shown below. The returns shown in this graph are calculated as: . The graph shows a negative abnormal return especially on
the expiration day (t=0).
18 The abnormal returns of various event windows are shown in Table 2. At the expiration day abnormal return is -1.40 percent and significant at the 1 percent level for the t-test and the 5 percent level for the parametric test. The percentage of negative return is 54 percent, thus a little more than half of the sample experiences negative returns. Furthermore, the table shows that for the three-day CAR (t-1 to t+1) the abnormal return is highly significant at a 1 percent level for
both tests.
Table 2: Abnormal returns whole sample
***
,** and *show the significance of the t-test at a 1, 5 and 10 percent level. III,I I and I show the significant of the Wilcoxon rank sum test (parametric test) at a 1, 5 and 10 percent level. The CAR is calculated as: , using daily abnormal returns and the fraction of negative returns is calculated by dividing the number of negative returns over the total number of returns. Because this is a one-tailed test the probabilities are divided by 2. The sample size consists of 579 daily abnormal returns.
event windows CAR in %
Fraction of negative returns in % Median t-10 to t+10 -2.35***III 51.38 -0.10 t-10 to t-1 -0.82*II 52.14 -0.14 t-2 -0.47 49.48 0.00 t-1 -0.82**III 57.59 -0.31 t0 -1.40 ***II 53.62 -0.17 t+1 -1.60 51.72 -0.05 t+2 -1.44 47.24 0.00 t-1 to t+1 -1.13 ***III 54.31 -0.18 t+1 to t+10 -0.96 **III 50.40 -0.05
19 significant, could be due to the fact that the sample sizes are smaller since the initial sample is divided in a pre-crisis an crisis sample.
In Figure 2, it is clearly shown that the pre-crisis period experiences larger negative returns than the crisis period. This result is in line with the behavioral finance theory discussed in the literature review. The studies discussed, all agree on the fact that investors hold on to their shares during a period of losses (crisis), in the hope the situation will turn. And that investors sell their shares in a period of wins (pre-crisis), to lock-in their profit (Shefrin and Statman, 1985; and Odean, 1998; Henker and Owen, 2008). This theory may explain why the pre-crisis period show larger negative abnormal returns compared to the crisis period.
Table 3: abnormal returns pre-crisis versus crisis period
***,** and *show the significance of the t-test at a 1, 5 and 10 percent level. III,I I and I show the significant of the Wilcoxon rank
sum test (parametric test) at a 1, 5 and 10 percent level. The CAR is calculated as: and the fraction of
negative returns is calculated by dividing the number of negative returns over the total number of returns. Because this is a one-tailed test the probabilities are divided by 2. The sample size consists of 579 daily abnormal returns.
Pre-crisis period Crisis period
20
Figure 2: Cumulative abnormal return pre-crisis versus crisis period
For a sample size of 579 observations, the cumulative abnormal returns from day -10 until day +10 around the lock-up expiration day are shown below. The returns shown in this graph are calculated as: . Showing that the pre-crisis period is more negative than the crisis period.
Table 4: Differences in abnormal return between pre-crisis and crisis period shows the differences in returns between the pre-crisis and the crisis period. At the event day (t0), the pre-crisis has a
larger negative return than the crisis period, only this difference of -0.13 percent is not significant. Two days before and two days after the event day, the differences in return are significant. At t-2 the crisis experiences significant larger negative returns and at t+2 the pre-crisis
21
Table 4: Differences in abnormal return between pre-crisis and crisis period
The differences are calculated by subtracting the crisis period from the pre-crisis period. Their significance is denoted by ***,** and * for the t-test at a 1, 5 and 10 percent level and III,I I and I for the Wilcoxon rank sum test (parametric test) at a 1, 5 and 10 percent level. Because this is a one-tailed test the probabilities are divided by 2. The sample size consists of 579 daily abnormal returns.
event window
differences between pre-crisis and crisis period in %
t-10 to t+10 0.53 t-10 to t-1 0.38 t-2 0.39 **II t-1 -0.14 t0 -0.13 t+1 0.13 t+2 -0.33 **II t-1 to t+1 -0.14 t+1 to t+10 0.28
The first hypothesis for this event study is:
H1: There is abnormal return measured at the expiration day of the lock-up contract
This hypothesis can be confirmed at a 1 percent and 5 percent significance level for the t-test and the parametric test, respectively. This result is in line with the findings of Field and Hanka (2001), Brav and Gombers (2003), Brau, Carter, Christophe and Key (2004), and Nowak (2004). As discussed in the literature review, this effect may be explained by larger-than expected sales of insiders or the downward sloping demand theory.
The second hypothesis is:
22 This hypothesis cannot be confirmed, since the results on the event day are not significant. This result implies that there is a larger negative abnormal return measured for the pre-crisis period, but that the tests could not find any evidence for this sample to accept the hypothesis.
4.2 Multiple regression
This paragraph will present the results from the regression. Before I present the results, I will perform a test to see whether the explanatory variables used in the regression cause any bias to the outcome of the results. Table 5 presents a correlation matrix. The correlation matrix shows that the variable volatility is significantly correlated with the other three explanatory variables, and that the crisis dummy is significantly correlated with the Nasdaq dummy. However, the correlations are relatively small, so there will not be a problem with multicollinearity.
Table 5: The correlation matrix
The correlation and their significance are shown, most of the variables are significant, but the correlations are relatively small.
Correlation
Probability Volatility
Log of market
capitalization High tech Crisis
Volatility 1 Log of market capitalization -0.2750 1 0.0000 Nasdaq dummy 0.2571 -0.0094 1 0.0000 0.8224 Crisis dummy 0.1925 0.0430 0.1184 1 0.0000 0.3026 0.0044
23 variables increases by the value of one, the cumulative abnormal return decreases by the value of the coefficient. Except for the market value, these results are in line with the theory discussed in paragraph 3.2 Multiple regression methodology.
The theory discusses that the market value of a firm, should be positively related to the abnormal return at the event day. The results of this regression are not in line with this theory. I do not have a clear explanation for this result. It could be that for this particular sample the relation is negative, but since the results are not significant I cannot draw any conclusions.
Table 6: The effect of market conditions on the cumulative abnormal returns at t0
The regression results are shown below. Coefficients are reported and in parenthesis the standard error, the statistical significance is denoted by ***, ** and * for the 1, 5 and 10 percent significance level. There are 6 models used to make a distinction between the effects of the variables. All variables show a negative impact on the abnormal return, but none of them are significant. The sample size consists of 579 daily cumulative abnormal returns.
Dependent variable: CAR t=0
coefficient (standard error) model 1 2 3 4 5 6 Crisis dummy -0.0046 -0.0041 -0.0043 -0.0041 -0.0040 -0.0036 (0.0032) (0.0033) (0.0033) (0.0033) (0.0033) (0.0034) Nasdaq dummy -0.0039 -0.0039 -0.0037 (0.0034) (0.0034) (0.0034)
Log of market value -0.0006 -0.0008 -0.0013
24 Table 7 presents the result for the regression analyses with the three-day CAR as dependent variable. I conduct this second regression because the event day has no significant abnormal return in table three, when a distinction is made between pre-crisis and crisis periods. For the three-day CAR the abnormal returns are significant. The results do not differ much from the regression based on the event day. The crisis dummy and Nasdaq dummy coefficients remain negative and are not significant. The market value has a positive relation to the three-day CAR, which means that larger firms have less negative abnormal return during this period. This result is in line with the literature discussed in paragraph 3.2 Multiple regression methodology. The volatility also remains negative, for model five this result is also significant at a 10 percent significance level.
Table 7: The effect of market conditions on the cumulative abnormal returns at t-1 to t+1
The regression results are shown below. Coefficients are reported and in parenthesis the standard error, the statistical significance is denoted by ***, ** and * for the 1, 5 and 10 percent significance level. There are 6 models used to make a distinction between the effects of the variables. All variables show a negative impact on the abnormal return, but none of them are significant. The sample size consists of 579 daily cumulative abnormal returns.
Dependent variable: CAR t-1 to t+1
25 The hypothesis tested with the regression is:
H3: The market conditions have an impact on the abnormal returns.
26
5.
Conclusion
Although the expiration of a lock-up contract is a fully known and expected event, I find a -1.40 percent significant abnormal return at the event day. This result is consistent with the first hypotheses that there is negative abnormal return measured at the expiration day of the lock-up contract. This effect can be explained by larger-than expected sales of insiders or the downward sloping demand curve. The results that I find are in line with the findings of Field and Hanka (2001), Brav and Gombers (2003), Brau, Carter, Christophe and Key (2004) and, Nowak (2004). When comparing the abnormal returns for the crisis and crisis period, I find that the pre-crisis experiences a larger negative abnormal return than the pre-crisis, but this result is not significant. Therefore, the hypotheses that the pre-crisis period shows larger negative abnormal returns compared to the crisis period cannot be confirmed. This result is not in line with the behavioral finance literature, which state that when investors experience a loss (crisis) they do not want to collect their losses and hold on to their shares, hoping the market will turn. When investors experience a win (bullish market) they are more willing to sell their shares and collect their profits (Shefrin and Statman, 1985; and Odean, 1998; Henker and Owen, 2008).
For the third hypotheses that market conditions have an impact on the abnormal returns. I find no evidence, this means that the market conditions have no influence on the abnormal return measured at the event day. Since this abnormal return is caused by an increase in supply, I cannot conclude that insiders sell more of their shares during the pre-crisis period compared to the crisis period. This result is in line with the results from the event study. The other variables of the regression: Volatility of the stock, the Nasdaq dummy and the market value of the firm, Show results that are in line with the theory, although the results are not significant.
Limitations
27 lock-up length for their contract of 180 days. Because of this reason I could not link the insider’s behavior with the informational asymmetries, adverse selection and moral hazard.
Future research
28
References
Bessler, W., & Kurth, A. (2007). Agency problems and the performance of venture-backed IPOs in Germany: Exit strategies, lock-up periods, and bank ownership. The European Journal of Finance, 13(1), 29-63.
Brau, J. C., Carter, D. A., Christophe, S. E., & Key, K. G. (2004). Market Reaction to the Expiration of IPO Lockup. Managerial finance, 30(2).
Brau, J. C., Lambson, V. E., & McQueen, G. (2005). Lockups Revisited. Journal of Financial and Quantitative Analyses, 40(3).
Brav, A., & Gompers, P. A. (2003). The role of lock-ups in initial public offerings. Review of Financial Studies, 16, 1-29.
Brooks, C. (2008). Introductory econometrics for Finance. UK: Cambridge university press. Brown, S. J., & Warner, J. B. (1980). Measuring Security Performance. The Journal of Financial
Economics, 8(3), 205-258.
Courteau, L. (1995). Under-Diversification and Retention Commitments in IPOs. Journal of Financial and Quantitative Analyses, 30(4).
Field, L. C., & Hanka, G. (2001). The expiration of IPO share lockups. The Journal of Finance, 56(2).
Garfinkle, N., Malkiel, B. G., & Bontas, C. (2002). Effect of Underpricing and Lock-up
Provisions in IPOs: Implications of a classic case of supply and demand. The Journal of Porfolio Management, 28(3), 50-58.
Gong, B., Lei, V., & Pan, D. (2013). Before and after: The impact of a real bubble crash on investors’ trading behavior in the lab. Journal of Economic Behavior & Organization, 95, 186-196.
29 Holt, C. A., & Laury, S. K. (2005). Risk aversion and incentive effects: new data without order
effects. American economic review.
Hurt, C. (2004). Moral Hazard and the Initial Public Offering. Cardoza Law Review, 25. Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk.
Econometrica, 47(2), 263-292.
Leland, H. E., & Pyle, D. H. (1977). Informational asymmetries, financial structure, and financial intermediation. The Journal of Finance, 32(2).
NASDAQ. (2014, 04). Retrieved from
http://www.nasdaq.com/markets/ipos/activity.aspx?tab=pricings
Nowak, E. (2004). The Expiration of Mandatory and Voluntary IPO Lock-up Provisions -
Empirical Evidence from Germany’s Neuer Markt. In G. Giudici, & P. Roosenboom, The Rise and Fall of Europe’s New Stock Markets (pp. 181–200). Elsevier.
Odean, T. (1998). Are Investors Reluctant to Realize Their Losses? The Journal of Finance, 53(5).
Ofek, E., & Richardson, M. (2000). The IPO Lock-Up Period: Implications for Market effciency And Downward Sloping Demand Curves. working paper, Stern School of Business. Seo, M., Goldfarb, B., & Barrett, L. F. (2010). Affect and the framing effect within individuals
over time: risk taking in a dynamic investment simulation. Academy of management journal, 53(2), 411-431.
Shefrin, H., & Statman, M. (1985). The Dispsition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence. The Journal of Finance, 15(3).
US Government. (2014, 06). Retrieved from https://www.sec.gov/investor/pubs/rule144.htm Yahoo finance. (2014, 02). Retrieved from
30
Appendices
Appendix 1: The S&P 500, the boom and crash market (Yahoo finance, 2014)
In this figure you can see clearly the upward moving market (boom) from 2003 until late 2007, after this the market crashes. Around October 2007 the market falls and late 2008 you can see that the market is really crashing.
Appendix 2: Rule 144: Selling Restricted and Control Securities
The rule 144: Selling Restricted and Control Securities, is a rule set by the Securities and Exchange Commission in the US. This institute sets the conditions under which restricted
securities can be sold (US Government, 2014). The five conditions that must be met before these securities may be sold are as follows:
1. Securities of companies that are subject to the reporting requirements of the Securities Exchange Act of 1934, need to be held at least six months. Securities of companies that are not subject to the reporting requirements need to be held at least one year.
2. There is an 'adequate' amount of current information about the issuing firm publicly available.
3. The amount to be sold may not exceed one percent of the shares outstanding or, for shares listed at a stock exchange, the amount to be sold must account for less than one percent of the average of the previous four weeks' trading volume.
4. All trades must be handled as routine trade transactions, so normal commissions must be paid to brokers.
