6.3 Summary of results
6.3.1 BootRegression (1)
Without Outliers BootRegression(1) Regression(1) BootRegression(2) Regression(2) Intercept -0.7425 ** (-3.59)*** -0.7442 -0.1599
* -0.1414
(-1.68)*
Cold Issue 0.0084 0.0048 (0.17) -0.0039 -0.0041 (-0.34) Market Condition 0.0183 0.0285 (0.68) 0.0081 0.0129 (0.73) Energy -0.0118 -0.0111 (-0.33) -0.0055 -0.0124 (-0.79) Materials -0.0854 ** -0.0853 (-1.51) -0.0341 ** -0.0344 (-1.48) Industrials -0.0034 -0.0072 (-0.24) -0.0008 (0.242) 0.0032 Consumer_disc -0.0216 -0.0218 (-0.70) -0.0070 -0.0013 (-0.59) Consumer_stap 0.0467 (0.93) 0.045 0.0116 0.0107 (0.47) Health Care 0.0016 -0.0005 (-0.01) -0.0077 -0.0096 (-0.65)
IT 0.0506 0.0563 (1.5) 0.0068 0.0078 (0.51)
Utilities -0.0277 -0.0173 (-0.27) -0.0165 -0.0009 (-0.35) Log(market_cap) 0.0312 ** (3.32)*** 0.0316 0.0062 0.0054 (1.39) Log(age) 0.0256 ** (2.46)** 0.0238 0.0126 *** (2.97)*** 0.0123 Underwriter 0.0335 0.0289 (1.10) 0.0075 0.0052 (0.47) Spillover -0.1303 -0.0445 (-0.33) -0.0672 -0.0002 (-0.00) Ownership -0.0580 -0.0611 (-1.18) -0.0065 -0.0089 (-0.41)
Bank 0.0395 * (2.01)** 0.0401 0.0099 (01.22) 0.0105
Table 13: Regression results without outliers
6.3 Summary of results
10% level and appears with a positive sign. The results implies that a company operating within the information technology industry experience a higher underpricing on average relative to the financials industry. The result is in line with our expectations as IT was classified as being a risky industry.
In addition, the variables Log(Marketcap) and Log(Age) are statistically significant. Log(Marketcap) is significant at a 5% level with a coefficient of 0.0281. This indicates that as the market capitalization of a company increases by 1%, the average underpricing increases by 0.0281%. Log(Age) is also statistically significant on a 5% level. The estimated coefficient of the variable is 0.0234, implying that as the age of a company increases by 1%, the average underpricing increases by 0.0234%. Both of these results are contrary to our expectations, and a thorough discussion of possible reasons for this will follow in section 6.4.
The variable Bank was estimated as having a statistically significant positive effect on underpricing, in line with our expectations. It is significant at a 10% level with a coefficient of 0.0418, indicating that if a company has no or multiple banking relationships it will experience a 4.18% higher underpricing on average.
Several variables are estimated to have signs that are in line with our expectations, but not statistically significant at any traditional significance level. Firstly, Cold_Issue appear with a positive sign in
BootRegression(1), indicating that there is higher underpricing on average when there is a period of few IPOs being performed. The results are opposite of previous research.
The industry dummies Energy, Materials and Utilities are estimated to have a negative impact on average underpricing which is in line with our classification of these industries as being low risk. Our expectations was that the variables were estimated to be zero, indicating no significant difference in underpricing relative to Financials as this is also classified as a low risk industry. Consumer_Stap is the last low risk industry and appears positive in our regression, indicating that the average underpricing is higher in this industry relative to that in the Financials industry. The results are in line with our
expectations as the classification of risk categories is broad and there may be small differences in the level of underpricing within them. Industrials, Consumer_Disc and Health_Care enter the regression
85
with positive signs, inferring that a company operating within one of these industries will experience a higher underpricing on average. This in also in line with the risk assessment of the industries;
Industrials, Consumer_Disc and Health_Care were deemed to be moderately risky industries, indicating higher risk than Financials and thereby a higher expected underpricing. Unfortunately, neither of the industry dummies are statistically significant at any traditional significance level, hence we cannot say that there is any significant difference in the level of underpricing between these industries and the financials industry.
The variables Market_Condition, Underwriter and Ownership were estimated as having the opposite effect on underpricing than what we expected. Market_Condition appear with a marginal negative sign in BootRegression(1), in line with previous research. The results indicate that underpricing is higher when the market is classified as performing better than average. Furthermore, the expectation was that underwriter reputation would have a negative impact on underpricing; that is using a reputable underwriter should lower underpricing. In BootRegression(1) the variable appears with a positive sign.
Finally, having a higher degree of insider ownership were expected to increase underpricing, but entered the regression with a negative sign indicating that insider ownership reduces underpricing.
Neither of the variables is statistically significant at any traditional significance level. Thus, we cannot conclude if the timing of the IPO, the use of a reputable underwriter or the level of insider ownership has any significant impact on the level of underpricing in an IPO.
To check the robustness of our regression, we excluded outliers defined through a Cook’s D above 0.0449. The regression was then run again with the new dataset and the results are somewhat
different. Market_Condition appear positive in the regression without outliers and falls in line with our expectations. The variable is still not statistically significant. IT, which was significant at a 10% level in the original bootstrap regression, is not significant in the regression without outliers. The variable still has a positive impact on underpricing, indicating higher underpricing within IT compared to Financials.
The loss of significance might be because one of the outliers is the company with the highest
underpricing in the sample, Tobii which had an underpricing of 38.45%, operates within the IT industry.
86
The industry variable Materials becomes significant at a 5% level when estimating BootRegression(1) without outliers. The significant coefficient implies that companies operating within this industry are less underpriced on average relative to that of companies within the Financials sector. The result is not in line with our expectations as the industry is classified as having low risk, and the statistically
significant estimate indicates that it has significantly different level of underpricing relative to the base industry. Lastly, the sign of Industrials and Consumer_Disc changed from positive to negative. The change in Consumer_Disc might be a result of the industry losing one of the observations, Besquab.
Neither of these variables is statistically significant in the original nor in the regression without outliers.
As mentioned previously, all outliers have been checked and found to be correctly registered and not unlikely to occur again, so our analysis will mainly focus on the regressions performed on the original, full dataset.