A procedure used to average several Mantel-Haenszel delta difference values for an item is described and evaluated. The differential item functioning (DIF) procedure used by the Educational Testing Service (ETS) is based on the Mantel-Haenszel statistical technique for studying matched groups. It is standard procedure at ETS to analyze test items for their differential difficulty with regard to gender and race/ethnicity of examinees. Test items used to perform the averaging procedure for this study were taken from the item bank of a large national testing program. Each form used in this study consisted of 80 items. Forms selected for the study had a large number of items in common. The delta difference average is the mean of single delta difference statistics across forms weighted inversely to their respective error variances. Overall, the results of differential item functioning analyses are similar to the weighted average procedure. Even if test forms were constructed to have items in common, as is the case in common item equating, the cost of test development and subsequent statistical analyses would be considerable. The weighted average procedure, on the other hand, is quite easy to perform, particularly with the use of a personal computer, and it does not require reformatting of the original data. Seven data tables are included. (TJH)