row is the total attributable to measured characteristics (E+C). We also provide, for information, two additional rows giving the raw means for the two groups of firms (the difference is the amount to be explained, in the “Total” row); and two additional columns giving the mean characteristics (endowments) for the two groups of firms. The explanatory variables are size (log L), location (a “big city” dummy), export activity (dummy variable), majority foreign ownership (dummy variable), and sector dummies. The treatment of sector dummies is different – for reporting purposes, manufacturing is chosen as a benchmark category and the “Sector” row reports the total decomposition for non-manufacturing vs. manufacturing. The results are based on fixed-effects regressions with country dummies.

We begin with Table V.5, which reports the decomposition applied to our basic benchmarking question: how do firms in transition economies (TEs) differ from firms in developed market economies (DMEs) in 2005? The base category is TEs, so the differences are what make DMEs different from TEs. The “DME mean E” and “TE mean E” columns show that the two sample of firms do differ significantly in basic characteristics: the TE firms are substantially larger (employment is about 35% higher); are less concentrated in manufacturing than the DME sample; and are more likely to be found in big cities. The big difference between the two groups of firms is the lower reliance of DMEs on retained earnings (14.2 percentage points) and the higher use of bank credit (7.6 percentage points). It turns out that these raw differences reported in the “Total” row are only modestly different from those in the “Region (DME dummy)” row, which is the difference between the groups after account for the impact of different endowments E and coefficients C. The biggest single identifiable impact is size: although the smaller size of the DME firms make them slightly more likely to rely on retained earnings (1.8 percentage points), this is reversed by the smaller (in absolute value) negative size effect for DMEs. But this is dwarfed by the large shift coefficients that remain after accounting for measurable characteristics: firms in TEs rely more on retained earnings (9.1 percentage points), more on family and informal sources (5.6), and less on bank financing (-6.6) and other sources (-6.7). The fact that the differences

11

We used Ian Watson’s (2005) “decomp” addin for Stata for the decompositions in this section.

33