The White House proposal to place more emphasis on performance in federal pay has revived longstanding questions of whether employee performance evaluations are fair and reliable enough to base more of an employee’s pay on them.

Regarding fairness, federal employee unions are contending that using ratings to determine pay leaves employees vulnerable to favoritism and discrimination. That contention helped sink the last major pay for performance system attempted for rank and file employees, the national security personnel system installed at DoD during the Bush administration. It operated only for several years, and only for a segment of the workforce, before being repealed early in the Obama administration.

Unions say that agencies should make more use of the authority they already have to move employees up the steps of a pay grade for good performance.

However, many individual employees favor pay for performance, including some of those who were under the NSPS system. Such systems have operated for years with little controversy in “demonstration projects and other pockets of the government. And a lack of rewards for good performance and lack of consequences for poor performance are among the biggest complaints on the annual federal employee survey.

Reliability meanwhile has been an issue in prominent pay for performance system already in place, in the SES, has been criticized for many years because about half of execs annually are rated at the top of their five levels, with almost all of the rest rated just one level below–and almost none rated at the lowest level. While forced ratings patterns are barred, that pattern has struck many experts inside and outside the government as top-heavy.

A 2016 GAO report found much the same pattern for employees below the executive level. In five-level systems, 38.6 percent were rated as outstanding and 35.1 percent as exceeding fully successful; in four-level systems (where there is no “minimally successful” level) the figures were 35.7 and 52.3 percent. In three-level systems, 33.7 percent were rated as outstanding and 66.1 percent as fully successful, and in two-level pass-fail systems, 99.9 percent passed.

GAO did not state an opinion regarding whether the ratings overall are too high, although it did stress that ratings should make “meaningful distinctions” between levels of performance. It also noted that ratings patterns below the executive level vary considerably by occupation; in the SES system–where the occupation is essentially the same for all–they meanwhile vary widely by agency.