Each regression was tested using the fixed effects model and the random effects model
All the models were estimated considering the inclusion of six dummy variables corresponding to the years studied. Six variables were included (and not seven, from 2008 to 2014), to avoid the problem of multicollinearity, and the Wald test was used to verify the relevance of those variables in the model. The aim of this test is to verify whether the parameters of the dummy variables are together equal to 0, which would indicate there being no statistical importance to keep them in the model.
The significance from the Wald test resulted in a value greater than 0.05 in all cases, indicating that the parameters of the dummy variables are together equal to 0 and, therefore, should be excluded from the model. That is, these variables were not able to capture macroeconomic factors that could have affected the companies in the period analyzed. This indicates that in no year was there discordant behavior compared to the others, which may be related to relevant external factors from a particular year, such as a serious crisis or specific economic events.
All the models were also estimated considering the classification of the auditors into two groups (“Big Four” vs. others), not resulting in different conclusions from the five groups (each one of the “Big Four” and others) considered.
4.1 Model A
To carry out the regression, only the periods of the companies in which there was a difference between the amount of CFO disclosed and the encouraged CFO were considered, resulting in the use of 171 companies and 585 data items.
Rio de Janeiro, RJ: Elsevier
To evaluate which of the models (fixed effects or random effects) is the most adequate, two tests were used: the Breusch-Pagan Lagrange multiplier (Breusch-Pagan test) and the Hausman test. Their results can be found in Table 5.
As the significance (p-value) of the Breusch-Pagan test was lower than 0.05, it is concluded that between the pooled model and the random effects model, the random effects one is the most adequate. Read more…?