Performance Ratings and Career Advancement in the US Federal Civil Service

Abstract

A strong link between performance and rewards in the U.S. federal civil service could raise top performers to positions of power and responsibility and motivate employees to greater productivity. Federal employees, the general population and scholars all express doubts about the strength of that link, however, though few have estimated it empirically. Using random-effects panel data models on a one per cent sample of federal personnel records for 1988–2003, we examine whether performance ratings meaningfully influence promotion probabilities and annual salary increases. With an average annual promotion rate of 17.8 per cent over this period, we estimate that employees with ‘outstanding’ and ‘less than fully successful’ ratings were one-fourth more likely and one-fifth less likely, respectively, to receive promotions than those with ‘fully successful’ ratings. Average salary impacts were smaller but still significant. Patterns held up across agencies and stages of the federal careers. Performance ratings continued to affect career advancement one or two years later. We speculate on whether these links are strong enough to motivate performance and advance the most qualified federal employees.

Key words:

• Performance appraisals
• pay for performance
• salary increase
• promotion
• federal civil service

Notes

Wooldridge (2009: 490) notes that the random effects model uses ‘quasi-demeaned data on each variable. The fixed effects estimator subtracts the time averages from the corresponding variable. The random effects transformation subtracts a portion of that time average’. The average of a time-constant variable is constant, of course, so that variable drops from the model. This also applies to omitted variables.

Several federal agencies have moved from five-level rating systems to three-level or pass/fail rating systems since 1996. As both systems are too crude to provide a basis for pay and promotion decisions (Montoya and Graham 2007; Liff, 2007), we dropped them from the analysis. Although Liff (2007) writes that the goal of pass/fail system is to ‘weed out’ poor performers, analysis on a one per cent sample of CPDF shows that among those who were under pass/fail system only 0.12 per cent of federal employees received ‘fail’ ratings and that there is no significant difference in exit rate between those who received ‘pass’ and those who received ‘fail’. Only a few federal employees are under three-level rating system and among those, only a few or no (in some years) employees received the lowest rating. In addition, there is no evidence that the highest rating lead to higher salary increase or higher promotion probabilities than mid-level rating, meaning three-level rating system is not much different from pass/fail system in terms of the use of the results.