April 14, 2011Dear Chairman Issa:
I read in a recent Federal Times column that you as chairman of the House Oversight and Government Reform Committee asked the Government Accountability Office to explore why studies of federal pay differ widely in their conclusions. I might be able to save them some time. I've had several occasions to study just such issues.
I am aware that you and fellow Republican Rep. Dennis Ross of Florida want to replace the General Schedule system "with a merit-based, market-sensitive system that recognizes and rewards individual employee performance." I agree with your goal. It's consistent with an ongoing global trend among public employers.
My background includes involvement in three National Academy of Public Administration studies that focused on replacing the GS system. I also worked recently with both the Office of the Director of National Intelligence and the Defense Department to address pay and performance issues (though unfortunately, the plans for those pay systems were locked in before I got involved.) My federal experience runs more than 20 years.
Two themes come out of what by now must be 100 or more conversations. First, the GS system is broken and is no longer serving the needs of federal agencies. Second, the next salary system should be market-sensitive, with salary increases based on employee performance. There is also agreement that the next system needs to be more flexible. The list of individuals who agree with your goal is lengthy.
Reasons for the Differing Conclusions
The reason for the wildly differing conclusions reached in the analyses by the Office of Personnel Management and the Bureau of Labor Statistics as well as those by the Heritage Foundation and the American Enterprise Institute is simple -- they are "measuring" different things. They share a basic problem in that their conclusions are based on black-box statistical methods that defy easy understanding.
OPM is required by statute to determine market pay levels for jobs comparable to those assigned to the grades in the General Schedule. BLS data collectors assign survey jobs to grades and averages are calculated. Significantly, they are not working to determine market rates for federal jobs; their data do not show whether federal jobs are overpaid or underpaid. Their mandate and methods differ starkly from anything industry uses.
Another statutory requirement is that OPM uses BLS data. The source of the data is the Employment Cost Index. That is a time series developed to track the percentage increase in pay levels across the country. In reporting percentage increases, it is similar to the Consumer Price Index.
As a BLS acquaintance once put it in a meeting, "You wouldn't use the CPI to set the price of a jar of mustard." His point was they do track mustard prices for the CPI, but with all the sizes and varieties, the data are not appropriate for setting prices -- or salaries. No private sector employer would consider using ECI data to plan salaries.
AEI and Heritage share a different purpose. Their focus is the pay of individual employees. As their studies point out, other researchers have completed similar analyses and also concluded federal employees are paid above market levels. The word researcher is important -- normally the analyses are reported only in academic journals. Their results have virtually no practical value to managers addressing salaries.
They rely on a very different database, the Current Population Survey. The survey is never used for salary management. A fundamental problem is this database fails to include information that is central to analyzing market pay levels (e.g., job level). The think tank analyses reflect several assumptions at odds with the dynamics of labor markets.
Both OPM and the think tanks rely on "sophisticated" statistical analyses, but their focuses are very different. Both could be correct in their claims -- it all rides on the assumptions. Neither, however, focuses on understanding market pay levels for federal jobs. Nor can they identify the jobs or the individuals who are overpaid or underpaid. My recommendation is to assume neither generates usable market data, and to initiate a conventional market analysis. I am confident GAO will reach the same conclusion. (You should be aware that when GAO planned its own salary system, its analysts relied on a conventional approach to "market pricing" -- they compiled market data from benchmark salary surveys.)
Methodology to Determine Annual Pay Adjustment
The questions about the OPM methodology started shortly after the BLS radically changed its survey strategy to strengthen the ECI. That was in the mid-1990s. The changes undermined the long-established logic to adjust the GS ranges and reportedly were planned without asking for OPM's input. For several years, the agency refused to use BLS data. During the past decade, OPM has worked with BLS to piece together a methodology to transform the data to a format that can be used.
OPM's approach is convoluted, based on several improvised analytical steps, and resembles a classic Rube Goldberg contraption. During the George W. Bush administration, the Pay Agent reported "major methodological concerns," but no significant changes have been proposed.
There are a number of technical concerns with the methodology, but undoubtedly the most significant is that it fails to focus on the labor markets relevant to federal staffing. That is the textbook starting point for salary planning. The BLS data -- as well as the Census Bureau data the think tanks use -- includes information from the smallest mom-and-pop businesses, as well as industries that clearly are not relevant to federal pay-setting.
No other employer would consider using BLS data for this purpose. They universally compile market pay data from one or more of the many salary surveys conducted annually across the country.
This problem will not be resolved until the government adopts a salary planning strategy similar to other large employers. Everyone seems to agree the federal salary system should be market-sensitive. Credible, easily understood market data are readily available for virtually every occupation.
Overcoming Past Mistakes
Not long ago, the then-president of the AFGE told me, "The union has to learn to live with merit pay." The pay experiment at the Navy's China Lake lab was then 15 years old; the National Institute of Standards and Technology pay system was well-established, and a number of agencies were planning performance pay "demo" projects. Several successful pay systems were rolled out throughout the 1990s. Successful systems also can be found in state and local governments here and abroad.
But resistance began to build with the termination of the Homeland Security Department's MaxHR and the Pentagon's National Security Personnel System. Those stories were, of course, complicated by the Bush administration attempts to limit the role of collective bargaining. Systems attempted at GAO and the Securities and Exchange Commission also ran into problems, though for different reasons.
Before starting new pay initiatives, it would be valuable to understand and learn from those mistakes. The GAO and SEC problems could have -- should have been -- avoided. The size and diversity of the operations covered by MaxHR and NSPS, however, suggest the picture is more complex; there were pockets in those agencies where the new pay systems were accepted.
A common thread is that government needs to strengthen the systems used to manage performance. OPM Director John Berry recently has been highly critical. His remarks have been on target; so has his vision. But the "fix" must address both the systems and the commitment of managers to develop and demonstrate requisite skills. Managers and employees alike need to make this a priority.
Howard Risher is an independent compensation and performance management consultant. He was the managing consultant for the studies leading to the 1990 Federal Employees Pay Comparability Act. He is the author or co-author of five books, including Planning Wage and Salary Programs (WorldatWork Press, 2009).
April 14, 2011