Group Comparability: A Multiattribute Utility Measurement Approach to the Use of Random Assignment with Small Numbers of Aggregated Units

It is not always possible, especially in large-scale evaluation research, to ensure that random assignment will produce groups that are comparable on any number of potentially important factors. Typically, gaining comparability has been achieved only at the expense of random assignment. A method is presented that allows multivariate comparability while making only minimal restrictions on randomization. The procedure is demonstrated in the context of assigning 63 aggregated units (schools) to 28 experimental and control conditions. Good comparability of groups for all primary main effects and interactions was venfied for 15 individual variables.