Sensitivity Test

H. Sensitivity Test

To check the robustness of my findings, I run two sensitivity tests are below.

1. Reexamining with Normal Data

The first sensitivity test is done with the trimming data (reduced data) of main sample so that the samples showed normal distribution in 538 observations (64.04% from the selected sample). There are no autocorrelation and multicollinearity problems but the problem of heteroscedasticity can‟t be treated.

Accounting information of NBEn, EPS, DAKOT, and INV has a value relevance and significant at 1%, and the coefficient is positive. The DATW and ALOPS don‟t indicate a value relevance. The results of statistical tests show the

four variables (NBEn, EPS, DAKOT, and INV) are consistent with the results of the regression test of the main data (840 observations).

The test results show that conservatism interaction to accounting information indicates that conservatism weaken the value relevance of the EPS, DAKOT, INV, and DATW and strengthen the value relevance of the NBEn, but has no effect on the value relevance of ALOPS. The test results are consistent with the results of testing the main data (840 observations) in the case of NBEn, EPS, INV, and DATW. Although statistically significant, coefficient of Cn*DAKOT is negative, in contrast to the direction of the Cn*DAKOT in main data.

2. Validation with the Price of Common Stock at the Time of Publications of

Financial Statement on Newspapers

The second validation test is done by replacing the dependent variable to the price at the time of publication of financial statement in the newspapers that are restricted until the end of the submission of financial statements to Bapepam, i.e. until the end of April (for 2002 data and earlier) and the end of March (for 2003 data and after). Stock prices in both kinds of samples have a very strong correlation (100%). Stock prices on both kinds of samples have minimum and average values are not much different.

Regression results of accounting information on stock prices suggest that the accounting information of NBEn, EPS, and DAKOT has value relevance with probability value is 0.000 (significant at 1%), whereas DATW has value relevance with probability is 0.030 (significant at the 5 %) in positive direction. The ALOPS is significant at 1 percent in negative direction. The growth of fixed asset (INV) does not indicate value relevance.

The test results of the 6 variables are consistent with the results of hypothesis testing first-fourth and sixth. Coefficients (β) of the five variables are significant and show the same sign with the result of main sample (840 observations). Different result is on the fifth variable, which INV shows no significant.

The test of influence of conservatism to accounting information indicates that it weakens the value relevance of the EPS, INV, and DATW and strengthens the value relevance of the NBEn and DAKOT. The impact of conservatism on the value relevance of the growth in current operating asset is not significant.

The results of both sensitivity tests report that the results are quite consistent with the results of regression of main data (840 observations). Their consistency are: (1) the book value of equity and net income have value relevance, (2) dirty surplus items (OCI) has value relevance, (3) the dynamic accounting information in the form of the growth of fixed asset has value relevance, (4) conservatism as moderating variable, weaken the value relevance of net income, dirty surplus items, growth of fixed asset, and intangible asset and reinforce the value relevance of equity book value.