Physician-induced demand for childbirths.

A controversial technique for testing the hypothesis that physicians induce demand involves two stage least squares (TSLS) regression analysis of cross-section data on physician supply and utilization. This paper tests the power of TSLS by applying it where there is at most only a trivial amount of demand inducement--the demand for childbirths. We find 'evidence' of inducement of childbirths, calling into question the validity of the TSLS approach. This unlikely finding may be traced to at least two factors: The first stage regression is not identified and the second stage regression does not adequately address border crossing.