Sample selection and source of data Descriptive Statistics

2432 total of 161 companies have been taken as samples for the purpose of this study Please refer to Table 1.

3.3 Statistical Method

Logit analysis is used in a multivariate setting to investigate the factors associate d with income smoothing. The logit model is considered appropriate because the dependent variable is a dichotomous variable and the independent variables are either intervally or nominally measured. The logit model can be expressed as follows: Smooth i =  +  1 NED i +  2 DUAL i +  3 REM i +  4 INST i +  5 OWN i +  6 CCC i +  7 AUD i +  8 IND i +  9 PROP i +  10 SIZE i +  11 PROFIT i +  i 2433 where: Smooth = income smoothing status of company where 1 is for smoothers and 0 for non-smoothers Conclusion can be drawn by looking at the significance and numerical sign + - of the regression coefficients The details on the definitions for the independent variables are as per Table 2.

4.0 ANALYSIS AND FINDINGS

The results of running descriptive statistics, univariate tests and logit analyses are reported below.

4.1 Descriptive Statistics

The descriptive statistics of the 161 sample companies are presented in Table 3. These results indicate the existence of income smoothing practices among companies listed on Bursa Malaysia Berhad. 2434

4.2 Univariate Test Results

Two univariate tests were conducted for this study. One is the Wilcoxon Signed Ranks Test, while the other is T-test. In the former, the results obtained for DUAL, AUD, IND and PROP, are all very significant at the 0.01 significance level, for all income smoothing objectives. This is indicated in Table 4. Under the T-test, all the variables, except for NED H1and REM H3, do affect the incidence of income smoothing activities among the listed companies on Bursa Malaysia Berhad boards. Though not all of these income smoothing objectives are affected by each of the independent variables that are included in this study, yet their occurrences in one or more of the income smoothing objectives is sufficient to highlight the fact that each of these variables does have an impact on the income smoothing activities that prevail. To investigate the results further in a multivariate context, logit analyses are performed. The following section discusses the results of the multivariate analysis. 2435

4.3 Logit Analyses Results.

The results of logit analyses are summarized in Table 5 for each income -smoothing objective. In particular, the estimated model beta, the associated significance test results and the holdout accuracy rates of the model are reported. The logit model for Income from Operations and Profit before Tax as income- smoothing objectives are found to be insignificant with a p -value of 0.845 and 0.791. As can be seen from Table 5 Panel A, at significance level above 0.1 one -tailed test, only the variables for the industry sectors namely IND and PROP are significant p - values equals 0.095 and 0.081, respectively. The findings suggest the company in the hotel and property industry would tend to smooth their income more than the IndustrialCommercial industry.