not be reasonable to assume that the relationship between firm characteristics and borrowing is the same across ownership categories.
The same questions, in fact, apply to the time trends noted above. What is the source of the increased reliance on retained earnings of firms in TEs? Is it because of changing sample characteristics (size, sector, etc.)? Or because the relationship between characteristics and borrowing has changed? Or has it been an autonomous change, unrelated to characteristics of firms?
The decomposition approach is again a natural approach to addressing these issues. Applied here, the method consists of identifying two samples of firms. The regression relating financing shares to firm characteristics is estimated for the two samples for a particular category of financing. The decomposition then separates the total difference in financing shares into the amount attributable to “endowments” (firm characteristics); coefficients (the relationship between characteristics and financing); and the shift coefficient (the “unexplained” or “autonomous” difference between the two categories).11
We apply the decomposition to the following:
Private sector firms in developed market economies vs. private sector firms in
transition economies in 2005.
Privatized vs. state-owned firms in TEs in 2005.
Privatized vs. new private firms in TEs in 2005.
New private firms in TEs in 2005 vs. 1999.
Privatized firms in TEs in 2005 vs. 1999.
State-owned firms in TEs in 2005 vs. 1999.
In each case, we decompose the total difference in the share of financing via a particular source into the amount attributable to characteristics (endowments, E), coefficients (C) and the remainder (the dummy variable defining the first-named group of firms). Each cell in the main part of the table reports both the total amount attributable to a particular
characteristic, as well as the amount into E+C (underneath, in parentheses).