Cash out veterans’ preference in federal hiring, honor their debt
FedRAMP Director Pete Waterman offered a glimpse at federal hiring frustrations earlier this month with this post on LinkedIn:
I’m really sorry to all the folks who applied for the FR Cybersecurity Service only to be told that their application was not considered or reviewed because they weren’t a veteran. We were told we’d have access to the highest qualified candidates based on the specialized experience and job analysis worksheets we created but our HR decided instead to only review veteran applicants.
On the plus side we’ve already decided to make one person that did qualify an offer, and are interviewing more this week. We’ll never know who we missed out on though.
FedRAMP decides which cloud services the federal government is allowed to use. The people you want doing that job are people who have specific technical experiences with cloud services. That applicant pool is not large. Waterman’s team wrote the job analysis themselves to find those people, and got handed a list by HR that was filtered on a criterion they didn’t choose and couldn’t override. The people who decided which applicants were qualified were not the people who understood the work.
This is a story about a badly designed compensation program — one that fails twice over:
- As compensation: Most veterans never use it, it can’t be sized to anyone’s service or sacrifice, and if it follows the general pattern of deferred compensation, it isn’t valued particularly highly by the people it’s intended to help.
- As policy: Its costs are real, large, and counted nowhere, but paid out of state capacity — the government’s ability to put the best available people on the work it needs done.
Veterans’ preference is compensation
Some of the comments on Waterman’s post discussed whether this qualifies as affirmative action — but I don’t think that’s the right framework to analyze this. The Office of Personnel Management’s Vet Guide explains that preference exists to recognize the economic loss suffered by people who served, restore them to a favorable competitive position, and acknowledge the obligation owed to disabled veterans.
The word for something of economic value awarded for service is compensation.
And if we’re compensating veterans, we should be asking whether this is a good way to pay them. That’s a design question, and there is a whole literature on it in the context of military retirements, and the conclusion was that service members don’t value deferred compensation at anything close to what it costs to provide.
And that’s relevant because good compensation is, by and large, compensation people value when they’re choosing whether to take the job. This is how we know the deal is worth it to the person signing up, and it’s what lets the government recruit and retain efficiently — with offers that are compelling to the people it’s trying to reach.
There’s one exception: veterans’ healthcare. We wouldn’t let an 18-year-old trade away future coverage for a disability he or she doesn’t have yet, and we’d be right — but that exception is earned by its tight coupling to service. The need it covers is the one the job created, and the benefit extends more broadly for two reasons: It’s impossible to parse out which health conditions were service-connected and which weren’t, and even where you could, fracturing a person’s care across two systems produces worse medicine. A referral-list advantage 20 years later — one that goes to veterans with short careers and long ones, to those who suffered economic losses by serving and those who didn’t — is much less coupled.
The military retirement experiment
In graduate school, I did a lit review on military compensation, and the finding I remember a decade later was this: Service members discount the future steeply, so deferred compensation is a terrible way to compensate them.
The canonical study is Warner and Pleeter (2001). During the 1990s drawdown, separating service members chose between a lump sum and an annuity. Most took the lump sum, at implied discount rates the Government Accountability Office summarized as at least 18% for more than half of officers and over 90% of enlisted members. At 18%, a dollar in 20 years is worth about 3 cents now. GAO’s conclusion was blunt: The compensation mix is highly inefficient for recruiting and retention, because cash today beats future benefits, especially for a workforce in its 20s.
Then there was vesting: No pension at all unless you hit 20 years, which most enlisted members never did. For the majority, the benefit wasn’t just discounted; it was zero.
The Department of Defense eventually conceded all of it. The Military Compensation and Retirement Modernization Commission reported in 2015, and the Blended Retirement System arrived in 2018 — smaller annuity, immediately-vesting TSP match, mid-career continuation pay. The logic: A benefit most of your people never receive and barely value is a bad deal for everyone, including the government paying for it.
Veterans’ preference is like that — but in some ways, even worse.
Who collects, and how much?
I pulled the numbers from OPM’s EHRI data, going back to 2010, which is when the data begins to include veteran status.
About 560,000 veterans currently hold federal jobs in the workforce EHRI covers. That’s 28% of the federal workforce, which is a lot. But as a share of veterans, it’s only about 6% of the 9.3 million working-age veterans.
But the current count isn’t quite the right measure, because people cycle through employment. Turning the hire history into a lifetime figure means combining a few things — how many veterans are hired each year, how many of them are getting a first federal job rather than returning to one, and how long people stay. It’s an estimate, not a measured count, but across reasonable values it lands at roughly 1 in 5 to 1 in 4 veterans ever holding a federal job.
This is a ceiling: It overstates the help. Many of those veterans would have been hired anyway, on their qualifications — the government was always going to employ a lot of people with military experience, especially at the Department of Defense, by far the biggest employer of veterans. The number whose outcome preference actually changed is some smaller fraction, and nobody knows what it is. So we can say that this helps a minority of veterans, but not exactly how many — and for a compensation program, “we can’t say how many people benefit” is not a small detail.
The military knows how to target compensation. This isn’t that.
Nothing else in military compensation treats service as one uniform thing. Base pay varies by rank and years in service. Reenlistment bonuses vary by specialty and scale with scarcity. Special pays track burdens — hazardous duty, hostile fire, sea time, family separation. VA disability compensation is graduated in 10% increments, because disability isn’t one thing either.
Then, after their service, all that resolution collapses into veterans’ preference. Under the old numerical scoring, it was five points or 10: four years at a stateside desk and three combat deployments earned the same five. Under category rating, which is how most competitive hiring works now, it’s coarser still. Applicants get sorted into quality categories; veterans rank above non-veterans within their category, and a veteran with a compensable service-connected disability of 10% or more floats to the top of the highest category. A 30% disability rating and a 100% rating land in the identical spot1.
1This is an estimate, I reach it two ways that roughly agree. Flow method: (a) about 57,000 veterans are hired each year (the 2022–24 average, excluding transfers); (b) length-of-service records show at least 71% of those hires have zero prior federal service, and about 79% are first-timers once you account for the fact that some of the rest are veterans whose military time — not a prior federal job — shows on the service clock (non-veteran hires, who have no such confound, return at about 13%); (c) I scale up by 1.15 to cover the Postal Service, which does apply preference, plus a few agencies OPM’s public data omits; (d) about 200,000 people become veterans each year. So (57,000 × 0.79 × 1.15) ÷ 200,000 ≈ 26 percent. Stock method: (a) about 713,000 veterans held federal jobs before the 2025 cuts, roughly 8%of the 9.3 million working-age veterans; (b) a veteran is working-age for about 30 years; (c) the average federal job lasts about 11 years. So 8% × (30 ÷ 11) ≈ 21 percent. The one-in-five-to-one-in-four spread comes from the two inputs no data pins down: how many hires are returners, and how long people stay.
There are other problems. It’s only useful if you want to work for the federal government. It’s illiquid; you can’t sell it, bank it, or transfer it. Compare the GI Bill: legible, near-term, transferable to your kids, and heavily used.
And its value to any individual veteran is set by accident. Preference pays out only when it flips a hire, so what it’s worth depends on who else happened to apply to the job — not on service, sacrifice, or need — and it’s worth zero the rest of the time.
The value in any given year also depends on how much hiring the government happens to be doing. In 2024, about 54,000 veterans were newly hired into federal jobs (excluding transfers) — roughly 0.58% of the 9.3 million working-age veterans, or about 1 in 170. In 2025, under the hiring freeze, it was about 21,000 — 0.23%, 1 in 435. The share of working-age veterans hired into a federal job fell by roughly 60% in a single year, meaning that the compensation was cut to less than half its value by a headcount decision that had nothing to do with veterans. Deferred compensation that fluctuates with the payer’s staffing plans is lousy compensation.
Why an HR office reviews only veterans
Here’s the part of the FedRAMP story I initially skipped: For FedRAMP, reviewing only veterans almost certainly wasn’t required. Preference isn’t supposed to mean “only look at veterans.”
But it doesn’t matter whether the law requires it. What matters is the implementation, and risk-averse implementation of hiring rules is reliably dumber than the law requires.
And in this case, it’s actually even worse than that, because these were jobs that could have bypassed competitive hiring entirely by using direct-hire authority. Direct-hire authority is an admission, written into policy, that the normal process degrades hiring badly enough that for work we deem important enough we switch it off. Merit Systems Protection Board data shows the switch: Veterans were half of all competitive-examining hires from 2014 to 2018, but only 22% of hires made under direct-hire authority, where preference doesn’t apply.
And it could have been used for these jobs — but instead, they picked a different alternative hiring authority that still comes with veterans’ preference.
So the real rule is this: You get to hire well if you’re able to route around the standard process.
The energy tax
This is what the routing-around process looks like.
Managers post jobs and cancel them when the referral list has no one who can do the work. They shop for hiring authorities — can we direct-hire this? Does the candidate we want have a disability? Is there an excepted-service path? A detail? A contractor?
And you can see it in the veteran numbers themselves: Veterans’ share of new hires has slid from around 30% in 2010–2015 to about 21% in 2023–24 — consistent with hiring migrating into channels where preference doesn’t bind.
Using these alternative hiring paths is not itself the scandal. The scandal is the energy. Across every agency, every year, an enormous amount of skilled effort goes not into assessing candidates but into finding a legal path to hiring someone qualified. That effort is huge and expensive, and it appears in no budget, no scorecard, no appropriation. It’s paid in burnt manager hours and in applicants who responded to a posting that got cancelled and won’t be back.
Nobody has priced any of this, and I don’t think anybody can. But both the hiring of worse candidates and the effort to route around it are taxes on the government’s capacity to do its job.
Let’s go back again to the FedRAMP job. One small office decides what’s secure enough for every agency to use. When it works, the government gets modern tools; when it doesn’t, agencies wait years or settle for worse, and the systems holding your tax records and health data run on whatever made it through. If you think that work matters, learning that its manager was blocked from hiring the people who had run cloud services should make you livid.
Cash out the preference, honor the debt
My proposal is a trade.
Right now the government pays an unknown, uncounted amount — in hiring dysfunction, match quality, cancelled postings, authority-shopping — to deliver a benefit that most veterans never receive and whose value to the recipients nobody can state.
So do what DOD did when it faced the same math for retirement plans: commission the study, price the benefit, pick a number. Then pay that number to the veterans we’d be taking preference from — all of them, not just the ones who wanted federal jobs — in a form they can use: separation cash, a bigger GI Bill, funded transition and credentialing support.
Targeted problems deserve targeted instruments: If the problem is credential translation, fund credential translation; if it’s disability, target disability — the VA already knows how to size a disability payment. I would rather pay veterans more, in a form they’d choose, than keep paying an unknowable amount for a benefit most of them will never use.
Of course, preference persists precisely because its costs are illegible. It’s a way to honor veterans at no apparent fiscal cost. But it’s actually funded by an unappropriated tax on the government’s ability to do anything, and quietly devalued every year by the workarounds it provokes.
Pay debts in money, not in the ability to hire people who can do the work.
What happens next at FedRAMP
Waterman’s post ends with “fingers crossed we can get that sorted and try again.” So consider what trying again involves.
Someone rewrites the announcement. Maybe they hunt for a different hiring authority. Weeks pass, then months. And here is the thing about the people FedRAMP needs: Engineers who have operated cloud services are among the most employable people in the country. They are not refreshing USAJobs waiting for the job to be relisted. Some of them are gone for good to an employer whose hiring process works the first time.
And the energy gets spent again: the manager hours, the HR back-and-forth, all of it uncounted, none of it appearing in any accounting of what veterans’ preference costs.
Maybe the next round works. I hope it does. But even in the best case, we’ll never know who we missed out on.
Abigail Haddad is a former artificial intelligence/machine learning engineer with the Department of Homeland Security’s AI Corps.
