The General Accounting Office (GAO) study of home care services in the Cleveland area is a prominent attempt to inform the selection of clients for these services. This paper critiques two major portions of the GAO study: the "break even" analysis relating level of client impairment to the comparative costs of care in home and institutional settings, and the analysis of risk of institutionalization as a function of client characteristics. Detailed review shows serious flaws in the GAO methodology which undermine the credibility of its estimates of relative service costs and risks of institutionalization. This paper proceeds to place the GAO study within a conceptual framework that permits more systematic consideration of the issue of client selection. Analysis of a model of institutionalization shows that the strategy recommended by GAO of targeting the highest risk individuals may be an inefficient way to use home care resources. Factors pertinent to client selection include not only risk of institutionalization but also level of private support, changes in risk and private support occasioned by an increase in public support, number of clients in the target group, and political attractiveness of the target group.