拿数据(WIND)试了试,样本没选得很细,也没用Fama–Macbeth调整t值:
reg X i.D R i.D#c.R if INDUSTRY!="J"&LIST!=0 /*X: EPS/P */
reg X_w i.D R c.R#c.LTA_w c.R#c.MTB_w c.R#c.LEV_w i.D#c.R i.D#c.R#c.LTA_w i.D#c.R#c.MTB_w i.D#c.R#c.LEV_w LTA_w MTB_w LEV_w i.D#c.LTA_w i.D#c.MTB_w i.D#c.LEV_w if INDUSTRY!="J"&LIST!=0, vce(cluster ID)
matrix a=e(b)
gen u1=a[1,3]
gen u2=a[1,4]
gen u3=a[1,5]
gen u4=a[1,6]
gen lima1=a[1,8]
gen lima2=a[1,10]
gen lima3=a[1,12]
gen lima4=a[1,14]
gen C_Score= lima1+ lima2* LTA_w+ lima3* MTB_w+ lima4* LEV_w
gen G_Score= u1+ u2* LTA_w+ u3* MTB_w+ u4* LEV_w
原作者(Khan and Watts, 2009)算出来的是:
C_Score, mean 0.105 median 0.097
G_Score, mean 0.048 median 0.044
Cor(C, G)<0
我的结果是:
C_Score, mean .0472 median .0447
G_Score, mean .0036 median .0032
Cor(C, G)<0
大致还是Make sense吧
忽略我笨笨的代码
