Online Program

Incorporating the Sampling Variability from an Employee Perception Survey into the Ranking Process of U.S. Government Agencies
*Taylor Hudson Lewis, U.S. Office of Personnel Management 

Keywords: ranking, linearization, replicate weights, Federal Employee Viewpoint Survey, FEVS

The Federal Employee Viewpoint Survey is administered yearly to U.S. Federal employees by the U.S. Office of Personnel Management to gauge a variety of factors pertaining to the workplace environment and employee satisfaction. Responses to four thematically-grouped subsets of attitudinal questions are combined to form indices upon which 37 agencies are ranked. The four themes are motivated by the Human Capital Assessment and Accountability Framework put forth by the Chief Human Capital Officers Act of 2002. Currently, there is no formal procedure to incorporate the sampling error inherent in the indices when assigning ranks. Instead, an ad-hoc rounding technique has historically been used. This paper proposes two methods to approximate index variability, one based on Taylor series linearization and one based on replication. Two alternative rank assignment procedures utilizing these measures of index variability are introduced. No major differences were found between the two variance approximation techniques, but evidence from the alternative rank assignment procedures suggests the current method may be imposing rank boundaries where, statistically speaking, they do not belong.