REFUSAL AND ATTRITION IN THE RAND HEALTH INSURANCE EXPERIMENT: A RESPONSE TO NYMAN Joseph P. Newhouse Harvard University Naihua Duan Columbia University New York State Psychiatric Institute Emmett B. Keeler RAND Arleen Leibowitz University of California (Los Angeles) Willard G. Manning University of Chicago Carl N. Morris Harvard University Charles E. Phelps University of Rochester John E. Rolph University of Southern California John Nyman (Nyman 2007) argues that the usual interpretation of the RAND Health Insurance Experiment – modest cost sharing reduces use of services with negligible effects on health for the average person − is an artifact that results from greater attrition by those in plans with cost sharing. In particular, he speculates that if those facing hospitalization on cost sharing plans differentially dropped out of the Experiment, the observed medical expenditure would be lower and health status would be better among those remaining in the cost sharing plans. When we analyzed the data from the Experiment, we were sensitive to the possibility that differential attrition by plan – and although Nyman does not note it, also differential refusal rates by plan at the time of enrollment – could have affected our results. We conducted several analyses that convinced us that Nyman’s speculation is unwarranted and that the usual interpretation of our results is correct. Those analyses are described on pages 17-26 of Free for All? Lessons from the RAND Health Insurance Experiment (Newhouse and the Insurance Experiment Group 1993), which is reproduced on Newhouse’s home page