5.5. Robustness analyses
5.5.1. Sample splitting
Following prior literature (La Porta et al. 1998, Francis et al. 2008), we next partition our sample
of 128 countries based on governance quality scores (GOVQUALITY) to examine the relative
importance of our independent variables in jurisdictions with strong and weak governance structures.
The logit regressions are estimated separately for these two groups. In the group of developed
countries, Model 8 in Table 8 shows that the coefficients have the correct signs and are
statistically significant. In the group of less developed countries, coefficients of the variables
NATGAAP, FULLIFRS, and LAW show the predicted signs and remain statistically significant.
Interestingly, the coefficient of GOVQUALITY is negative, but no longer statistically significant.
One explanation for this result is that the most poorly governed countries are less responsive
to international accounting standards (Ramanna and Sletten 2009), suggesting that some of these
countries suffer from corrupt, slow-moving, and ineffective governments that are unwilling or
unable to introduce international accounting standards (La Porta et al. 1999). Ramanna and
Sletten (2009) call these countries ‘failed states’ where adoption of IFRS for SMEs is not of
any interest.