Y = X[beta] + [y.sub.1][I.sub.d] hospital and ratings + [y.sub.2][I.sub.s] + [mu]where X represents the vector of predisposing, needs, and other enabling factors. Incorporating the selection using the control function model, we haveY = X[beta] + [y.sub.1][I.sub.d] + [y.sub.2][I.sub.s] + [[alpha].sub.0](1 - [I.sub.d] - [I.sub.s])[[lambda].sub.0] + [[alpha].sub.d][I.sub.d][[lambda].sub.d] + [[alpha].sub.s][I.sub.s][[lambda].sub.s] + vThe [lambda]''s are the transformations hospital of the probabilities predicted from the ratings first stage (Heckman 1978; 1979; 1990; Lee 1982; 1983). They are the correction factors similar to the correction hospital factor (inverse Mill''s ratio) in a Heckman correction model where the response variable in the first stage is binary rather than multicategorical (Heckman 1979; 1990). Vella and Verbeek (1999) showed that testing for the presence of endogeneity is equivalent to testing whether [alpha]''s are zero. With the incorporation of the endogeneity control, ratings the effects of having a regular doctor ([I.sub.d]) and having a regular site ([I.sub.s]), conditional on the choice of a usual source of hospital care, are [y.sub.1]+([[alpha].sub.d][[lambda].sub.d]- [[alpha].sub.0],[[lambda].sub.0]) and [y.sub.2]+([[alpha].sub.s][[lambda].sub.s] - [[alpha].sub.0][[lambda].sub.0]), respectively.* A single-equation estimation ratings that regresses utilization over having a usual source of care would provide biased results of the effect of having a usual source hospital of care because of the omitted correction terms.

Based ratings on the conceptualization illustrated in Figure 1, this study focused on how having a regular doctor differs from having hospital and ratings a regular site in affecting the use of five preventive services: mammograms, pap smears, blood pressure checkups, cholesterol level checkups, and flu shots. The probabilities of having a usual doctor, a regular site, and no usual source of care were analyzed first in the selection stage. It was followed by the examination of the impact of having a regular doctor and having a regular site on the utilization of each above-mentioned preventive service, while incorporating individuals'' unobservable preferences in the selection stage.

©2003 www.hospital-ratings.com. All rights reserved.