Tunisia promulgated law n° 2005-96 which comes to control the companies, improve the transparency of financial information and reduce the manipulation. Before integrating the variable "law" like explanatory variable to test its impact on the REM, we carried out a test of CHOW which test the stability of the coefficients of regression on two different groups (before and after law n° 2005-96). We have found that the calculated statistics are lower than the theoretical ones. Consequently, the two groups are statistically different. Therefore, the test proves that there are differences between the two groups. The results of regression (Table 9) show that the variable "law" haven’t any effect on the reduction of the extent of the accruals earnings management in the Tunisian context (p-value = 0.543). In addition, the financial security law of 2005 doesn’t associate with the increasing of REM (p-value = 0.169). This result is inconsistent with Zang (2007), Graham et al. (2005), Gunny (2010) and Demers and Wang (2010).