A significant difference among the learning algorithms for the case of LAADR and LABANKS did not exist (see Table 2).
For the LAADR dataset, the relevant variables are market capitalization, law and order tradition, capital expenditure to long-term assets ratio, operating expenses to sales ratio, and debt to total assets ratio.
Comparing the main variables selected for LAADR and LABANKS (see Table 3), the main distinctive variable is the size of the company measured by the logarithm of market capitalization for LAADR and the equity index for LABANKS.
The accounting indicators, capital expenditures to long-term assets ratio, operating expenses to sales ratio, and debt to total assets ratio play a central role in the case of LAADR, and long-term assets to deposits ratio in the case of LABANKS.
Corporate governance variables seem to be more important for LABANKS than for LAADR. Government entities supervise and regulate banks intensely.
The participation of outside directors is the second most importantvariable for LAADR. The finance literature indicated that outside directors supervise managers (Weisbach, 1988; Shivdasani 1993).
In LAADR and LABANKS, insiders' equity ownership is also a relevant variable that affects performance.
The law and order tradition also has a negative relationship with performance for the LAADR. Large Latin American companies probably perform better in environments with a weak legal structure, with less of a tradition of law and order, and with higher levels of corruption because of the close family relationships that help influence government decisions in their favor.
In the case of LAADR and LABANKS, the limited impact of the size of the board of directors and the composition of the board (percent of outsiders) on performance and efficiency give findings similar to what previous studies have indicated for the USA.