Columns (ix) and (x) present the results for non-outsourcing firms. In
general, the corrected results are more precisely estimated and
provide higher adjusted R². This allows us to interpret our
estimates more confidently. While we do not find drastic
changes in the magnitude of coefficients, one notable exception
is the IT knowledge coefficient whose magnitude rises
from 0.051 in the switching regression to 0.069 (35 percent
change) under OLS (see columns (i) and (v)). We conjecture
that firms with comparative advantage in benefitting from a
vendor’s IT knowledge may be more likely to outsource and
to be represented in our sample. Therefore, the OLS results
that do not adjust for selection bias overestimate the impact of
IT knowledge.