The study estimated per pupil expenditures for a stratified random sample of 571 special education students in three metropolitan school districts--Charlotte-Mecklenburg (North Carolina), Milwaukee (Wisconsin), and Rochester (New York). The study used data collected in 1982-83 from parent interviews, teacher interviews, and the student's school record. The distribution of estimated per pupil expenditures was examined overall and by primary handicap, by placement, by study site, and by background characteristics. Factors associated with variations in estimated per pupil expenditures were identified. Main findings included: (1) the mean per pupil expenditure for all special education students was approximately twice that of regular students; (2) instruction accounted for 68% of expenditures, related services for 8%, and indirect services for 24%; (3) expenditures for instruction and related services varied considerably; (4) hearing impaired and physically/multiply handicapped students had higher per pupil expenditures than other groups; (5) within primary handicap groups there was considerable variation in per pupil expenditures for related services; (6) the least restrictive environment was also the least expensive and special classes were the most expensive placement; (7) there was no evidence of bias based on race, socioeconomic status, or gender in the levels of resources provided for instruction. (DB)