The Critical Time Intervention Project is a three-year clinical trial which tests a time-limited, supportive intervention to reduce recurrent homelessness among mentally ill men moving from a shelter to the community. Along with a comparison of nights spent homeless and other outcomes, the evaluation of the Critical Time Intervention includes a comparison of the relative costs of the intervention, compared to usual treatment. Such cost effectiveness analyses are difficult to perform and are rarely applied to mental health treatments. This paper presents the general scheme of this analysis and discusses critical issues in the construction and measurement of cost variables. Preliminary results which have implications for the cost analysis are presented.