Prospects for reducing group mean differences on cognitive tests via item selection strategies.

Area of Science:

• Industrial-Organizational Psychology
• Psychometrics
• Human Resources Management

Background:

• Cognitive ability tests are prevalent in employee selection.
• Significant mean differences in test scores exist across racial and ethnic subgroups.
• These score disparities pose a challenge for equitable test use.

Purpose of the Study:

• To investigate an item-level procedure for reducing subgroup mean differences in cognitive ability test scores.
• To assess the impact of removing test items on score equity and test validity.

Main Methods:

• Three datasets were analyzed using item-level trimming procedures.
• Items were removed based on difficulty or subgroup mean differences.
• Test reliability was adjusted, and subgroup mean differences were compared before and after trimming.

Main Results:

• Item trimming cannot eliminate, but may modestly reduce, subgroup score differences.
• The extent of reduction increased with higher proportions of items removed.
• Impacts on Black-White and Hispanic-White comparisons varied.
• Removing items based on subgroup differences had minimal impact on criterion-related validity.

Conclusions:

• Item trimming is a potential method for modestly improving score equity in cognitive tests.
• The effectiveness and impact on validity depend on the proportion and type of items removed.
• Further research is needed to optimize this procedure for fair employee selection.