Managers who've done it say it's surprisingly hard to link performance and salaries.

For years, federal managers have taken flack for poor leadership and people skills. But now, with laws granting broad new powers to managers at the Homeland Security and the Defense departments, Congress and the Bush administration have put their faith back in bosses. Managers are getting real control over hiring, disciplining and firing employees. At the core of the changes is pay for performance - the belief that federal employees, like their private sector counterparts, should be rewarded, or not, based on how well they do their jobs. And the link between salaries and performance is the manager's evaluation of how well each employee is doing.

The notion that employees will perform better knowing that managers will decide their fate will have profound effects on workplace relationships and on the implied contract between civil servants and government. But most of all, it will place a heavy burden on managers. Bosses will have to get tough. No more giving everyone on the staff better-than-average ratings, no more ignoring people who don't pull their weight. This time there will be no excuses: If staff performance is off, managers will have no one to blame but themselves.

On its face, pay for performance could not seem more logical, and, to be sure, there are strong arguments in its favor. But government managers who've already embarked on performance-based systems and private sector human resources managers readily agree that it's not easy. In fact, it's likely that the transition from the decades-old civil service system - where employees could expect raises based on seniority - to pay for performance will be among the most difficult challenges government managers face in their entire careers.

While performance pay may appear to be an ideal way to reward top workers and weed out poor ones, a system that isn't set up properly can spark divisive competition among colleagues, erode teamwork and alienate employees who view pay for performance as unfair. "The bottom line is that this is a huge task," says Terry Esvelt, senior vice president for employee and business resources at the Bonneville Power Administration. Bonneville Power was one of the first government organizations to address the problem of evaluating performance. "It's hard when you get down to thinking about how to define results," Esvelt adds.

The government is diving into the world of performance-based pay in part to mimic private sector practices it perceives as effective. But experts in the corporate world warn of difficulties ahead. "Most well-run organizations try to assure that people doing the good jobs are the ones getting the promotions. Still, expect a lot of frustration and failure along the way," says Pete Smith, president of the Private Sector Council, a Washington group that attempts to pass on private sector best practices to the government.

Many human resources experts argue that performance-based pay is exactly the wrong way to go. Because of the competition that results from performance-based systems, the typical result is that "your people are going to be unhappy," says Stanford University business school professor Jeffrey Pfeffer. And there are broader risks too, he says. If incentives are geared toward individual, rather than organizational, achievement, the result can be backstabbing and vicious competition among employees.

At the same time, supervisors in charge of judging employee performance have a natural tendency to favor people like themselves, according to Pfeffer. This might result in bias against women or minorities. And then there's the money. These performance-based systems are expensive to design, set up and implement. "There will be huge administrative costs for I don't know what result," Pfeffer says.

For now, federal managers have little choice. Washington is moving ahead with pay for performance, and a federal manager's best bet is to study best practices and to work to get it right the first time. The General Accounting Office, which oversees executive branch management for Congress, has developed a checklist to test whether a performance-based pay system passes muster. To make the grade, it should have at minimum:

• A clear link between individual performance goals and the agency's objectives.
• Methods of gauging meaningful distinctions about the performance of different employees.
• A final review of all rating decisions by an impartial arbiter.
• A grievance process to address employee concerns about ratings.
• Annual publication of overall performance ratings and pay decisions.
• Periodic employee surveys to measure satisfaction with the system.

"Most federal agencies are a long way from meeting this test," GAO Director for Strategic Issues Christopher Mihm told the House Government Reform Committee in April (GAO-03-612T).

But his office is working for an A. Since the passage of the 1980 GAO Personnel Act, the congressional watchdog has had the authority to implement wider salary ranges and pay-for-performance programs. But after taking over as agency head in 1999, Comptroller General David Walker found that the average rating for a GAO analyst was high - 4.6 on a 5.0 scale. He immediately assigned one of GAO's human resources executives, Susan Kladiva, to revamp the system. "It had lost any relevance or ability to make distinctions about performance," she says.

Indeed, a perennial problem with performance appraisal systems is a tendency for managers to inflate grades over time. Giving someone a low grade or poor evaluation never is easy, and most managers' natural tendency is not to do it. But the result of bloated grades at GAO, says Kladiva, was widespread staff frustration. "Our employees would say, 'I got a 4.5 rating. On half the elements, I was perfect; on the other half, I exceeded expectations, but I didn't get the big increase. . . . They didn't have the perspective to know that a 4.5 was below average," she says.

Walker immediately ordered managers to get tougher and to set 3 as the standard rating for an employee meeting but not exceeding all expectations. In 2000, the average rating fell to 4.19. The number of perfect ratings fell from 18 percent to 2 percent, while ratings of 4.6 or above fell from 50 percent to 11 percent as GAO began to see a meaningful dispersion in the ratings. In the meantime, Kladiva worked with an outside consulting firm and agency employees to better align the bases for employee ratings with the agency's mission. Ultimately, GAO decided to discard process-oriented rating areas such as data gathering and documentation, data analysis, and job planning in favor of outcome-oriented competencies such as thinking critically, collaborating with others and achieving results.

GAO set measurable goals in each area. To counter ratings inflation, these targets were aligned with four rating levels: role model, exceeds expectations, meets expectations, and below expectations. The meets expectations rating became a 1.5; exceeds expectations a 3; and role model a 5. In 2002, the average analyst rating was 2.19 - none had an average of 4.7 or above.

To win over wary employees, Walker made sure they had opportunities to comment throughout the development of the pay system, and managers now focus on regular performance feedback discussions with employees. At the end of each year, employees write self-assessments that kick off the grading process. Supervisors follow up with their evaluations, which are then reviewed by a member of the Senior Executive Service.

Employees are ranked against their peers. In 2002, the top 8 percent to 12 percent in each team were placed in the highest pay category and were eligible for a merit raise of 4 percent or more. Employees ranked in the second, third and fourth pay categories received smaller merit raises. Those in the fifth category received no merit raise, and those who fell below expectations in three grading categories were given a limited amount of time to either improve or lose their jobs. All GAO employees continue to receive the annual governmentwide pay increase, though the agency is seeking congressional approval to absorb those funds into its pay-for-performance system. "We want the primary factor in pay to come down to the skills, knowledge and performance of our people rather than just the passage of time," says Walker.

GAO isn't the only agency that has moved forward aggressively with pay for performance. In the 1990s, both the Internal Revenue Service and the Federal Aviation Administration won freedom from the constraints of civil service law and adopted pay for performance. Several smaller agencies, many at the Defense Department, were granted flexibilities under the Office of Personnel Management's personnel demonstration project program and initiated performance-based pay.

Then there are the agencies that have moved ahead even while still operating under the old civil service rules, among them Bonneville Power and the Court Services and Offender Supervision Agency, a relatively new federal agency responsible for overseeing parolees and probationers in the District of Columbia. The Web site of the Partnership for Public Service, a nonprofit group in Washington that encourages young people to go into government work, recently established a "Solutions Center," profiling some of these programs. The pay-for-performance programs are so new, however, that there is no concrete data to indicate whether employees have embraced them, or whether they have helped boost hiring and retention of top workers.

Still, some personnel experts in government are eager to move ahead. In 2002, Doris Hausser, the chief policy adviser at OPM and the agency's chief human capital officer, wrote "A Fresh Start for Federal Pay: The Case for Modernization," a report promoting pay for performance. Hausser argues that changes in the nature of federal employment over the last 50 years - during which the nature of government work has grown much more complex - make it critical that federal managers have greater power to hire and promote top workers and to cut loose poor ones. Instead of an army of clerks, the government now needs highly skilled technicians, engineers and scientists, she says.

"Our competitors, when they offer positions to people with these kinds of skills, are able to offer a value proposition: If you work here and perform admirably, there are these kinds of immediate rewards," she says. By contrast, in the federal government, "We can offer only long-term promises along with very steady and predictable work that fits with a 30-year career model....We are sending the message that time [on the job] is more important than performance, and as a result, we will attract people who are risk averse."

Competing with the private sector and adopting the hard-nosed competitive personnel policies of corporate America informs much of the reasoning behind the move toward pay for performance. In the intense competition of the marketplace, companies must hire and retain their best people, and cut out dead wood quickly to offer the best products at the best prices, the theory goes. Government should do the same to get the biggest bang for the taxpayer's buck.

The problem is that even in the private sector, basing pay on performance has proved to be a tough nut to crack. "In terms of merit pay programs in the private sector, they've had at best mixed results," says Jason Shaw, a professor of management at the University of Kentucky who has written extensively on the topic. "I would hesitate to say that these programs have been even modestly successful."

In a broad study in which he and some colleagues evaluated more than 40 prior analyses of merit pay, Shaw found that there was evidence showing that workers boosted their work output when driven by financial incentives, but there was little evidence that the quality of this work had improved.

Two main problems arose in the private sector examples and both are relevant to government managers as they attempt to implement performance pay. First, pay-for-performance systems tended to fail to motivate workers when the merit pay offered was minimal. In an era of tight budgets, that should weigh heavily on the minds of managers. By Shaw's estimation, a merit raise needs to be at least 7 percent of an employee's salary before the employee will find it meaningful or will boost his or her future efforts as a result of the raise.

To cut costs, and to constrain the impact of grade inflation, some agencies have adopted variations of a "forced distribution system" in which only a predetermined percentage of employees are eligible for the largest pay increases, a certain percentage are designated as poor performers, and the rest fall somewhere in between. This helps agencies stay within budget constraints, but it also can alienate employees who narrowly miss a larger raise, or any raise at all.

In addition, when these pay distributions are done at the work group level and not agencywide, they could penalize people in especially strong units and too richly reward those in weaker units.

The experience of human resources managers at Bonneville Power provides a cautionary tale about the perverse effects of tight budgets on performance-based pay. Bonneville Power is an Energy Department agency that markets wholesale electrical power and operates and markets transmission services in the Pacific Northwest. It relies on its businesses for funding. In the late 1990s, the agency put in place a pay-for-performance system for its executives. Detailed performance contracts were drawn up to gauge how managers were contributing to the agency's bottom line, and how customers and employees felt about the managers' service and leadership.

"Most managers like the clarity of what is expected of them" under the contract system, says Bonneville's Esvelt. "You know exactly at the beginning of the year what you will be responsible for." But when the economy went sour in 2001 and the agency's revenues dipped, the money for merit raises and performance bonuses dried up. Until those revenues turn up again, the program has been suspended indefinitely. "We had a hugely popular system in place," says Esvelt, "but now we're not so popular because it's been taken away." Inevitably, there will be many times when federal agencies' budgets are short through no fault of their employees. Finding a way to continue to reward top workers in such a climate is a challenge for which there may be no easy solution.

It's also much more difficult to objectively assess the performance of the highly skilled workers government now employs than the low-level clerks it used to, says management professor Shaw. It's relatively easy to count the number of documents a clerk processes or the number of machines a mechanic fixes, and reward that employee accordingly. But when a job is more complex, the appraisal is more difficult. How do you gauge the work of a cancer researcher at the National Institutes of Health? Or the conclusions drawn by a GAO analyst auditing a federal program? Sometimes, these tasks are easy, but more often, evaluations are subjective. If too many employees feel that they are being cheated out of merit raises, then they won't buy into the system, Shaw says. Certain agencies have found reasonable ways of dealing with the problem. Consider the Court Services and Offender Supervision Agency for the District of Columbia. In the 1990s, Washington's city government was in financial turmoil, and as the Partnership for Public Service noted in a report on the agency, city services were failing. In 1995, a convicted murderer on parole killed a young woman. A subsequent report by the city's inspector general revealed serious operational and management flaws in the district's probation and parole systems. Soon after the report came out, Congress passed legislation to relieve the district of several costly criminal justice functions. As part of that legislation, the agency was created to manage the city's probation and parole services.

From its launch, the agency has worked within the existing civil service structure to set up a performance management system aimed at reducing recidivism among parolees. The agency's human resources department convened focus groups of supervision officers to identify the performance areas likely to help offenders make a successful transition back into society. If the agency's supervision officers perform well in these areas, the theory goes, the agency will achieve its goal of reducing recidivism.

Supervision officers now are graded annually on their proficiency in improving the initial assessment of an offender's needs and the risks posed by that person to the public; providing closer supervision of offenders; assuring offenders have access to treatment for drug or health problems and support services; creating community partnerships; and providing accurate and timely information about offenders to the courts.

To successfully meet performance targets, for example, supervision officers must complete an initial assessment of offenders within 20 days of receiving the case. Officers are not judged on the rate of recidivism among the offenders under their supervision, however, because the limited number of offenders assigned to each officer may not be a representative sample of the whole.

Though merely anecdotal at this point, employee reaction to the evaluation system has been good, says Joyce McGinnis, a management analyst at the agency. But McGinnis notes that the Court Services and Offender Supervision Agency has an advantage in evaluating performance because "what we do is really well-defined and discrete. We don't have 800 different types of programs that we are trying to manage."

It's easy for an agency to skimp on developing clear performance measures. GAO chastised the FAA in February 2003 (GAO-03-156) after finding that two-thirds of the employees interviewed said they felt the agency's pay-for-performance system was unfair. That dissatisfaction appears to have had dramatic repercussions. For example, since 1995, when legislation was passed giving FAA new personnel authorities, the percentage of employees joining unions has increased from 63 percent to 80 percent.

According to the report, FAA human resources managers said they were under "significant pressure" to implement the new pay system quickly and that they should have "spent more time to develop baseline data and performance measures." GAO auditors reported that because employee performance measures were not aligned with agency goals, assessment of the reforms' success would be difficult.

Mary Ellen Dix, acting assistant administrator for human resource management at the FAA, says the agency responded to the GAO report in 2003 by developing objective criteria to measure performance and linking it to overall agency goals. In December, the agency signed an agreement with air traffic controllers, extending the number of agency employees under a pay-for-performance plan to three-quarters of the agency's workers. But Dix warns other agencies now embarking on pay for performance "to expect that it will take some time" to develop good systems. Human resources managers should put a "primary emphasis on communication," she says. "The more reassured employees feel that we are doing everything we can, that we value their input, and are willing to incorporate changes based on it," the more they will feel ownership of the process, says Dix.

Indeed, adds the Private Sector Council's Smith, "You're not going to find any organization that has the answers on a piece of paper someplace." He adds that managers must not forget that "pay accounts for about 1 percent of [employee] motivation." As long as pay is competitive, he says, employees care far more about the work "environment, the value of the mission, the feedback they receive from supervisors and the recognition of their fellow employees, clients or citizens. That's where you get job satisfaction."