Our matching analysis requires the identification of nonoutsourcing
firms with attributes similar to outsourcing firms.
However, since the analysis requires IT data for all firms, the
set of non-outsourcing firms is also limited to those in the CII
database.
The matching analysis is based on the assumption
that all of the firm-specific attributes that lead to heterogeneity
in value creation from IT outsourcing are observed.
However,
several diagnostic statistics give us confidence that our results
are not driven by unobserved heterogeneity.
While switching
regression explicitly assumes normality in the error terms that
cannot be tested, it does allow us to account for unobserved
heterogeneity in choosing the mode of service delivery.
Although the matching technique and switching regression
require different assumptions, our results from both methods
are consistent, increasing our confidence in the results. In our
analysis of the effect of spillovers, we ignore the possibility of
other forms of knowledge transfer, such as employee turnover.
Finally, we acknowledge that we reconstructed the IT stock
variable starting in 1995 due to a redefinition of IT capital in
the dataset during the sample period. Further research would
benefit from a richer dataset. Despite these limitations, our
study provides robust evidence and insights into whether and
how IT outsourcing benefits client firms.