As expressed by Hypothesis 1, the event study on abnormal returns at lockup expiration allows us to test the semi-strong form of the EMH:
Hypothesis 1: We expect zero cumulative abnormal return at lockup expiration
Furthermore, we will test Hypothesis 1 according to the following null hypothesis:
𝐻0: 𝐴𝐶𝐴𝑅𝜏1,𝜏2= 0
Initially, we consider the daily AAR of the full sample for a broad selection of days surrounding the event window. The development of the daily AAR is shown in Figure 2.
49 Figure 2 shows the daily AAR for a 13-day period (-4, +8) and the upper- and lower bounds for the 95%
confidence interval for AAR. The graph cannot span a broader selection of trading days, since it is not possible to include trading days from the estimation window (which ends at 𝑡 = −5) nor any trading days that take place after 21 March 201828. One can observe that the AAR is positive at the lockup expiration date (𝑡 = 0), and immediately drops to a negative value on the two following dates (𝑡 = +1, 𝑡 = +2). The initial drop in AAR immediately after lockup expiration slightly rebounds over the following trading days. The development of AAR does not show a clear trend but does indicate a noteworthy market reaction immediately after lockup expiration. The bounds of the 95% confidence interval suggest that AAR varies between a positive and negative value of 1% throughout the depicted trading days. Furthermore, the confidence-bounds suggest that AAR is significant at a 5% level on day -2, +2, and +6, since the 95% confidence interval excludes 0% on these trading days. The results for the daily AAR are tabulated in Table 1 below.
28 The event study is performed as of 22 March 2018. Therefore, the latest adjusted closing price that can be attained at the time of analysis is for 21 March 2018. The sample includes several recent IPOs where the lockup expiration takes place close to the time of the analysis. For example, the lockup for Paxman AB expires on 9 March 2018. Therefore, we cannot include trading days that are later than t = 8, since this date is 21 March 2018 for the case of Paxman AB.
-1.5%
-1.0%
-0.5%
0.0%
0.5%
1.0%
1.5%
-4 -3 -2 -1 0 1 2 3 4 5 6 7 8
AARt
Trading day (t)
FIGURE 2: Daily development of average abnormal returns (AARt) (t= -4, +8)
The solid line represents the daily development of average abnormal returns throughout an expanded selection of trading days surrounding lockup expiration, including the event winow (-2, +2). The dotted lines represent the upper- and lower bounds of the 95% confidence interval for average abnormal returns.
50 The AAR is found to be negative on 5 out of the 13 (= 38%) depicted trading days surrounding the event date.
There are three degrees of significant observations in Table 1, namely 1) a negative AAR at 𝑡 = −4, which is significant at the 10% level, 2) a negative AAR at 𝑡 = −2 and a positive AAR at 𝑡 = +6, which are significant at the 5% level, as well as 3) a negative AAR at 𝑡 = +2, which is significant at the 1% level. Before we consider the significantly negative price reaction that occurs immediately after lockup expiration (𝑡 = +2), these findings are furthermore noteworthy since they indicate a premature and significantly negative abnormal return several days before the date of lockup expiration (𝑡 = −4, 𝑡 = −2), as well as a significantly positive and delayed abnormal return after lockup expiration (𝑡 = +6). However, the observations on day -4 and day +6 are arguably too far from lockup expiration to be considered as directly affected by the event. This notion is supported by Armitage (1995), who emphasizes that shorter event windows are recommended when the event itself is easy to determine, as it will ensure optimal representation of the effects that stem from lockup expiration.
Before we establish inferences from the results, we calculate CAR according to different event windows to test the sensitivity of our event study. Table 2 shows the results for CAR when applying our preselected 5-day event window (-2, +2) and other alternative event windows.
TABLE 1: Daily development of average abnormal returns (AARt) (-4, +8)
Trading day AAR SD N SE t-stat p-value Lower 95% Upper 95%
-4 -0.4% * 2.9% 141 0.002 1.83 0.069 -0.9% 0.0%
-3 0.4% 4.5% 141 0.004 1.15 0.252 -0.3% 1.2%
-2 -0.5% ** 2.3% 141 0.002 2.60 0.010 -0.9% -0.1%
-1 0.1% 2.7% 141 0.002 0.37 0.715 -0.4% 0.5%
0 0.2% 3.9% 141 0.003 0.65 0.518 -0.4% 0.9%
1 -0.2% 2.6% 141 0.002 0.88 0.378 -0.6% 0.2%
2 -0.6% *** 2.2% 141 0.002 3.50 0.001 -1.0% -0.3%
3 0.1% 2.1% 141 0.002 0.74 0.460 -0.2% 0.5%
4 -0.3% 2.6% 141 0.002 1.36 0.177 -0.7% 0.1%
5 0.2% 2.3% 141 0.002 1.02 0.308 -0.2% 0.6%
6 0.6% ** 3.1% 141 0.003 2.36 0.020 0.1% 1.1%
7 0.0% 2.5% 141 0.002 0.23 0.821 -0.4% 0.5%
8 0.3% 2.6% 141 0.002 1.26 0.211 -0.2% 0.7%
The coloured rows represent trading days where AAR is significant with at least 10% significance. The table depicts trading days surrounding the event, including those that are considered in the event window (-2, +2).
*p<0.1; **p<0.05; ***p<0.01
51 In Table 2, we apply various symmetric and asymmetric event windows surrounding the date of lockup expiration. Firstly, one can observe that nearly all event windows yield an ACAR that is negative. The ACAR ranges from -1.2% to 0.1%, which is narrower than what has been observed by previous research. The US studies found significantly negative ACARs that range between -1% to -3% (Ofek and Richardson, 2000;
Bradley et al., 2001; Field and Hanka, 2001; Brav and Gompers, 2003; Brau et al., 2004), whereas the European studies observe insignificantly negative ACARs of -0.5% to -2.5% (Ahmad, 2007; Espenlaub et al., 2001; Espenlaub et al., 2013; Goergen et al., 2006). Our results thus suggest that the variation in ACAR is less sensitive to the applied event window than what has previously been observed. Furthermore, our results indicate that the percentage of stocks that have a negative CAR varies between 52% and 61% among the different event windows. This is comparable to the finding of Field and Hanka (2001) who observe that 63%
of their observations have a negative CAR.
The analysis on ACAR yields a p-value of 0.031 and 0.035 for the 5-day (-2, +2) and 7-day (-2, +4) event windows, respectively, thus substantiating significance at the 5% level. It is highly noteworthy that these results comply with the preselected event window of 5 trading days (-2, +2), which was proposed as the most optimal event window in the methodology section. Due to the limited difference between the magnitude and significance of ACAR for the two event windows, we still adhere to apply the symmetric 5-day event window (-2, +2).
In this manner, our event study on abnormal returns suggests that Nordic IPOs on average experience a significantly negative CAR of -1.1% at lockup expiration, which is deemed to be significant at the 5% level.
We can thus reject the null hypothesis for the event study on abnormal returns and conclude that CAR on average is significantly different than zero:
𝐻0: 𝐴𝐶𝐴𝑅𝜏1,𝜏2= 0 The null hypothesis is rejected
TABLE 2: Average cumulative abnormal return (ACAR) according to different event windows
Event window ACAR SD N SE t-stat p-value Lower 95% Upper 95%
(-1, +1) 0.1% 5.79% 141 0.005 0.21 0.838 -0.86% 1.06%
(-2, +2) -1.1% ** 5.75% 141 0.005 2.18 0.031 -2.02% -0.10%
(-3, +3) -0.5% 6.91% 141 0.006 0.85 0.399 -1.64% 0.66%
(-4, +4) -1.2% * 8.61% 141 0.007 1.70 0.091 -2.67% 0.20%
(-2, +3) -0.9% * 6.10% 141 0.005 1.80 0.073 -1.94% 0.09%
(-2, +4) -1.2% ** 6.81% 141 0.006 2.13 0.035 -2.36% -0.09%
(-1, +2) -0.5% 5.79% 141 0.005 1.12 0.266 -1.51% 0.42%
(-1, +3) -0.4% 6.12% 141 0.005 0.80 0.424 -1.43% 0.61%
(-1, +4) -0.7% 6.81% 141 0.006 1.23 0.219 -1.84% 0.43%
The coloured rows represent trading days where ACAR has been observed with a minimum of 10% significance.
The table depicts various event windows to represent the sensitivity of our analysis. The applied event window is highlighted by the dotted lines.
*p<0.1; **p<0.05; ***p<0.01
52 Our event study on abnormal returns at lockup expiration provides significant evidence against the semi-strong form of the EMH in the Nordic market. It should not be possible to observe abnormal price reactions in a semi-strong form efficient market, as rational expectations are based on publicly accessible lockup-specific information from the IPO prospectus. Our findings thus suggest that the market fails to fully incorporate public information related to a forthcoming lockup expiration date. This result is highly important since previous research on lockup expirations in Europe has failed to observe significantly abnormal returns with similarly narrow event windows (Ahmad, 2007; Espenlaub et al., 2001; Espenlaub et al., 2013; Goergen et al., 2006).
The underlying cause of the significantly negative ACAR of -1.1% cannot be fully determined before we have assessed the results of the cross-sectional analysis. However, the development of AAR in Figure 2 is somewhat compatible with theory on the downward-sloping demand curve (Ofek and Richardson, 2000; Brav and Gompers, 2003). The graph indicates a sudden price reaction after lockup expiration, with a significantly negative AAR of -0.6% on the second trading day (𝑡 = 2). This can partly be attributed to new shares entering the market, which implies that investors seem to underestimate the number of shares that are sold after lockup expiration.
There are however certain characteristics of the graph that slightly contradict the theory on downward-sloping demand curves. Firstly, we observe that a significantly negative AAR occurs on two of the trading days prior to lockup expiration, namely at 𝑡 = −2 and 𝑡 = −4. This premature reaction cannot be explained by the downward-sloping demand curve unless some of the locked up shares have been subject to an early release.
Secondly, we observe a significantly positive AAR of 0.6% several days after lockup expiration, namely at 𝑡 = +6. This delayed reaction can arguably be attributed to induced demand from investors that are exploiting the sudden price drop. Due to the delayed reversal of the stock price, one could argue that the abnormal price reaction at lockup expiration is a result of a temporary price pressure29 rather than a permanent consequence of a downward-sloping demand curve (Field and Hanka, 2001).
With respect to information asymmetry, the significantly negative AAR after lockup expiration (𝑡 = +2) could possibly be a result of a misalignment in the quality of information held by insiders and outside investors (Field and Hanka, 2001; Brav and Gompers, 2003; Brau et al., 2004). If one were to suppose that insiders have possessed negative private information prior to lockup expiration (thus implying that outside investors have overvalued the stock due to their informational disadvantage), then the significantly negative AAR after lockup expiration could be a consequence of an unexpected value realization. However, we cannot ascertain such implications based on the event study’s results. This will be elaborated further upon when we have attained the results from the cross-sectional analysis.
29 A fundamental attribute of price pressures is that they only have a temporary effect and are reversed within a few days (Barclay and Litzenberger, 1988; Mikkelson and Partch, 1985).
53