Section 2 20 points We first estimate the nonlinear model AR

Section 2 (20 points) We first estimate the non-linear model AR-c+bi DISPERSION 2+b2 LOGSZE + Where AR is the abnormal return of funds\' holding stock in DISPERSION is measured the level of funds\' stock picking, and LOGSIZE is the natural logarithnm of stock market capitalization, e is an error term. By using Eviews8, the empirical results are given as follows: the earnings announcement date, Variable Coefficient Std. Error Statistic Prob. 0.174767 0.075654 2.310087 0.0220 68.21070 26.26439 2.597079 0.0101 -0.017789 0.007517 -2.366419 0.019 DISPERSIONA2 LOGSIZE 0.040964 Mean dependent var 4.036463Durbin-Watson stat 0.019204 0.001356 R-squared F-statistic Prob(F-statistic) 2.006614 From this table, we find (1) What is the final model? And Why? model by figure. 2

Solution

1. The final model is

Abnormal returns = 0.17 + 68.21 Dispersion ^ 2 -0.02 Log Size + Error term

The abnormal returns increases with the increase in funds\'s stock picking and decreases with the increase in the log of stock market capitalization.

2. Since the R Squared of the final model is 0.04 or around 4 per cent, and F statistic is also low, thus the variables do not fully explain the abnormal returns of fund\'s holding stock.