Providing services to handicapped children is more expensive than educating nonhandicapped children. Previous studies have estimated the cost of special education to be approximately twice that of regular education. However, these studies have produced a number of problems in providing accurate cost data including a lack of data at the local level to make meaningful determinations, difficulties in treating shared and indirect costs, problems in making cost comparisons across districts, and variances in the cost of resources over time. This study evaluated the Larson (1985) model, a new methodology for calculating special education program costs, by comparing it to the Rossmiller (1970) model, a widely used method for calculating special education program costs. Judgments were made by comparing the efficiency and effectiveness of each model to the other. Efficiency was appraised byl comparing input and process considerations in computing special education program costs in a select school district in Virginia. Effectiveness was appraised by comparing each model’s ability to produce comprehensive and accurate special education program costs from the sample school district. Findings indicated that the Larson model had several advantages over the Rossmiller model. First, the Larson model was more efficient as less information from the regular budget was needed to complete indirect cost calculations. Second, the Larson model was more efficient in dealing with shared costs as they could be prorated through the use of a multiplier. Third, the Larson model was considered more accurate in its treatment of related services costs. However, several qualifications needed to be made in Judging the Larson model as a better product over the Rossmiller model. Conducting a cost determination was a lengthy process no matter which model was used and is more dependent upon the availability of data in a school district than the model used. Also, both models tended to produce similar cost figures when related services costs were taken into account. / Ed. D. / incomplete_metadata