Abstract
The effects of data heterogeneity on the efficiency estimate by data envelopment analysis are evaluated here in terms of empirical applications in the computer industry. Scale or size variations of firms and heteroscedasticity are the two forms of heterogeneity considered here. Our empirical results show that the adverse effects of data heterogeneity can be considerably reduced by the methods suggested here.
Acknowledgements
I am sincerely thankful to two anonymous referees for their constructive suggestions that proved most useful in improving an earlier version. I alone am responsible for any remaining errors.