The Post-Offering Performance of IPOs in the Health Care Industry
Guirguis, Hany S., Onochie, Joseph, Rosen, Harry, Journal of Economics and Finance
This paper investigates the post-offering performance of initial public offerings in the health care industry in a sample of 223 IPOs issued between 1985 and 1996. Statistically insignificant abnormal returns for IPOs relative to matched control firms and risk-adjusted health care index are evident for the whole sample. Thus, our empirical results support the overall information efficiency in the IPO market. However, numerical and statistical differences of the IPOs' abnormal returns are documented in every subgroup specified according to the issuance years and sectors. We conjecture that such differences are due to the growing threat of government intervention and the significant structural changes. (JEL 11, Cl)
An Overview of the IPO Performance Literature
The post-offering performance of IPOs has been examined in many academic studies in the last two decades. It is of significant interest because the existence of any long-run return patterns in the IPO market can affect investment strategies. In addition, the existence of abnormal long-run returns casts doubt on the informational efficiency of the IPO market.1 Many studies have documented the tendency of IPOs to underperform the market in the long run. For example, Ibbotson (1975) reported negative performance of the IPOs between the second and fourth years for IPOs during 1960-1969. In addition, Stem and Bornstein (1985) showed that over a period of 10 years the IPOs underperformed the Standard & Poor's 500 stock index by 22 percent. Aggarwal and Rivoli (1990) documented that the IPOs' 250-day return underperformed the matching NASDAQ return by 13.73 percent during 1977-1987.
For more details see Shiller (1990).
2 We thank an anonymous referee for pointing us in this direction.
3 We first tried to match each IPO to a control firm from the health care industry based on the market and the book values as explained earlier. However, we were successful in matching only 12 IPOs from the final sample with 223 IPOs.
4 As in Ritter (1991), we truncate both the returns for the IPOs that are delisted before their three-year anniversary and the returns for the IPOs with initial return period greater than one day.
5 Unlike Barber and Lyon (1997), we resort to non-parametric techniques to account for the non normality of the series that is augmented by the limited number of observations.
6 This procedure can be described as follows. First, draw 1,000 random samples without replacement from the IPOs buy-and hold-returns, (0), In (ME), and (ln(BE/ME)). Second, run the regression and calculate the coefficients (8'A at each draw (i) where n=1, 2, 3, and 4. Then calculate the standard deviation (a',) of each 8 '- from the 1000 regressions. Third, at each draw conduct another round of 200 bootstrapping using the draws of the four series as your original series and run the regression at each draw. Then calculate estimated coefficients (S ".) and their standard deviation (a"Ji from the 200 regressions. Fourth, calculate t-statistics of the (8 'n) as follows:
Fifth, order the 1000 tj and find the alt and I - aI2 percentile values of t, where a is the 10 percent significance level. Finally develop the 90 percent confidence interval around the estimated coefficients from the original data as follows:
If the zero line is located between the Sth and 95th percentile confidence intervals, the abnormal returns are statistically insignificant at the 10 percent level. For more details see Mooney and Duval (1993) and Effron and Tibshirani (1986).
' Subsequent empirical tests are restricted to the abnormal returns of the IPOs based on the control firm and the riskadjusted returns (portfolios 2 and 4). Adjusted returns based on portfolios I and 3 were subsequently excluded from the remaining tests.
Aggarwal, Reena, and Pietra Rivoli, P. …