This study takes the following steps to test hypothesis 1. First, countries are grouped by
investor-protection levels using a K-means cluster analysis based on MacQueen (1967).6
Second, each of the four earnings attributes is measured for the companies in the sample
using a pooled regression per country per year in order to get 217 country-year accounting
observations, calculated from a product of 31 countries and 7 years. These 217 country-year
earnings-attribute observations are then averaged by country. Then, this country-mean
value is assigned to its investor-protection cluster. The study investigates the hypothesis
that clusters with low levels of investor-protection should have relatively less favorable
values of these attributes.