skip to main content skip to footer

Test Bias DIF

Author(s):
Beller, Michal
Publication Year:
2005
Report Number:
RM-04-09
Source:
ETS Research Memorandum
Document Type:
Report
Page Count:
13
Subject/Key Words:
Differential Item Functioning (DIF), Fairness, Predictive Bias, Test Bias

Abstract

Much of the criticism of psychological tests is derived from the observation that certain groups differ extensively in test performance. However, much of the polemic concerning the issues of test bias rests on two main confusions: (a) differences in test performance among groups are often regarded, in and of themselves, as an indication of test bias, ignoring performance on the external criterion that the test is designed to predict. Often, groups that perform poorly on tests also tend to perform poorly on measures of the criterion. Furthermore, analysis of the relationship between tests and criteria often reveals similar regression lines for the various groups; and (b) the issue of test bias is often confused with the possibility of bias in the content of some individual items included in a test. Group differences in test performance are attributed to the content of specific items, and rather than eliminating problematic items in a systematic way (i.e., checking all items for differential performance), this confusion has, in some cases, led to suggestions of a wholesale rejection of reliable and valid test batteries.

Read More