英语论文网

as well as the
problems due to generated mental environment leading to stagnancy in stock market, the coefficients
values and the models forecast power were subjected to considerable fall.
4.3. The Comparison of the Four Models Results Regarding the EFR Hypothesis
Taking a little consideration on the results shown in Table 4, we find that using this results may cause
errors arisen when comparing traditional and sign-oriented models because of the high similarity of
four models’ results. For the same purpose in this research other criteria were used. The criteria utilized
for comparing the four studied models are the Sum of Squared Residuals, Akaike and Schwarz
Criterions that in all three cases the lower criterion indicated the superior model. Based on these
criteria, the comparison results of four models during six estimation periods have been given in Table
5.
As specified in this table, the Sum of Squared Residuals for the third model in five periods and
for the fourth model in one period out of six estimation periods is lower than other models.
Consequently, with considering the Sum of Squared Residuals criterion, the third model is superior to
other models. Akaike criterion in all periods indicates the superiority of the fourth model. At last,
Schwarz criterion also introduces the fourth model as the superior model in three periods out of six
estimation periods. Finally, with taking into account all aspects above associated with the EFR
hypothesis, the fourth model can be titled as the superior model. Also, though the sign of conservatism
variable in the third as well as the fourth model – for firms having negative abnormal earnings- is unlike to the prior expectation, but Feltham-Ohlson (1995) based models have generally acted more
successfully than Ohlson (1995) based ones.
4.4. Temporary Persistence Test of Negative Abnormal Earnings
As previously explained, according to Giner and Iniguez (2006) investors assume the negative
abnormal earnings as more temporary than the positive ones. Their two reasons for this claim have
been given before. In order to study this issue in sign-oriented models of this research (second and
fourth models) Wald coefficients test was used. Since according to theory, the abnormal earnings
coefficient should be between zero and one, the more this coefficient differs from 1 (close to zero) the
more temporary will be the relevant abnormal earnings. Therefore, the null hypothesis of Wald test is
confirmed on these coefficients equaling to 1 and it is expected that this assumption is failed for
negative abnormal earnings and is confirmed for the positive type. This test results have been shown in
Table 6.
69 International Research Journal of Finance and Economics - Issue 36 (2010)
Table 4: The results of EFR hypothesis in four models
Estimation Model 1 Model 2 Model 3 Model 4
period 11 ω AR2
11 ω +
11 ω AR2 11 ω 12 ω AR2 11 ω +
11 ω 12 ω +
12 ω AR2
+
22 ω
1998-2002 0.716097* 0.80 0.885741* -0.187931** 0.80 0.2005* -32.12591* 0.73 0.652453* -0.126849 -11.00133* 16.7147* 0.77 1.253698*
1998-2003 0.720112* 0.78 0.861372* -0.171273* 0.78 0.135527* -32.27204* 0.70 0.623132* -0.120399 -10.29964* 16.57176* 0.75 1.264519*
1998本论文由英语论文网提供整理，提供论文代写，英语论文代写，代写论文，代写英语论文，代写留学生论文，代写英文论文，留学生论文代写相关核心关键词搜索。