NILG Conference - OFCCP: An Overview of Statistical Applications

OFCCP: An Overview of Statistical Applications
Speaker: Dr. Javaid Kaiser, Director of Statistical Analysis, OFCCP

In this session, Dr. Kaiser introduced himself as the new head statistician for OFCCP. As the head of the department his focus in the presentation was to discuss how he wanted to shape the OFCCP and what he felt were the key things that Federal contractors should be focusing on when it comes to conducting statistical analyses. He began with a few statistical procedures that would benefit the contractor community for self-evaluation of employment data, then he went into changes that are happening at the OFCCP and he closed with audit tips.

There is a new name for the statistical analysis division at the OFCCP – the Division of Statistics and Technology (DST). The DST has adopted a new directive that all audit-related statistical work must reviewed by the heads of the division before a Notice of Violation can be issued. The reason for this is to ensure that the analyses are conducted correctly.

The statistical analyses that the DST uses during selection reviews (and those that contractors are urged to use) are the binomial test for large sample sizes, Fisher’s Exact for small sample sizes, Logistic regression, Mantel-Haenzel, and the Breslow-Day tests. Dr. Kaiser mentioned that Logistic Regression, Mantel-Haenzel and Breslow-Day tests are more complicated than the other statistical analyses and should be conducted by statistical experts.

Listed below are some current challenges that the OFCCP faces when conducting compliance reviews:
Submittal of incomplete data.
Developing SSEGs is a difficult process – the OFCCP might change the process.
It’s difficult analyzing data that has small sample sizes. OFCCP’s solution is to use non-parametric tests.
Handling pay grade, department, location, shift, etc. differences when analyzing certain groupings.

Some suggestions that Dr. Kaiser gave regarding data submittal during a compliance evaluation are:
· Clearly label all 12 factors for the compensation data request and explain why any are missing.
· Compensation data should be the same as the snapshot data for the plan.
· Disclose any bargaining union agreements

Dr. Kaiser finished the session discussing DST’s plans for improvement in the following areas in order to decrease the time spent on reviews:
Development of the SSEGs.
Review the triggers for compensation.
Data verification protocol.
Explore parametric statistics.
Standardize operating procedures throughout all OFCCP offices.
Uniform business rules nationwide.
Automate business process.