5.5.3. Multi-period logit
Next, we use a multi-period logit model instead of a single-period logit model. Note that the
majority of our independent variables in our main model do not change over time since we use
binary variables (NATGAAP, FULLIFRS, and LAW). Therefore, we use a multi-period
random effects model that is able to estimate the effects of time invariant variables.11 We
obtain data on the year of IFRS for SMEs application from the IASB’s jurisdiction profiles.
Further, we use country-specific webpages, included in the jurisdiction profiles, to collect the relevant
information.
Our multi-period logit model incorporates each country-year as a separate observation, where
countries can adopt IFRS for SMEs. Adopters are coded ‘0’ in the years before adoption and ‘1’ in
the year when IFRS for SMEs is optional or required the first time; Non-Adopters are coded ‘0’every year they are in the sample (Ramanna and Sletten 2009, Bassemir 2012). This approach
results in 640 country-year observations for 128 sample countries over the sample period
2009–2013. Moreover, we use lagged values (t21) for the continuous independent variables.
In addition, we rerun our analysis including year-dummies to examine whether IFRS adoption
does depend on time.
Model 13 in Table 9 shows that NATGAAP, FULLIFRS, GOVQUALITYas well as LAWare
statistically significant at the 1% level including time dummies. Furthermore, there is evidence
that larger economies measured by the variables LogAREA and LogGDP are less likely to
adopt IFRS for SMEs. The estimated values of the year dummies become larger over time,
suggesting that the probability of IFRS adoption is increasing over time. Taken together, the coefficients
of the multi-period random effects model are larger in magnitude than the single-period
logit coefficients.