Gastonia, NC
It is an engine, and the second half of this question contains a false premise that I want to correct before anybody builds on it.
Not every previous reform failed. One passed.
The Federal Information Technology Acquisition Reform Act became law in 2014. It gave agency chief information officers real authority over IT spending — approval rights over acquisitions, a role in budget formulation, a say in hiring. It required agencies to inventory and consolidate data centers. And it created a public scorecard, issued by the House Oversight Committee, grading every major agency twice a year.
That scorecard has been published for about a decade. Anybody can read it. Agencies get letter grades.
So the story that reformers have only ever produced reports and hearings is not accurate. A reform was enacted, it granted the authority reformers had asked for, and it built the public accountability mechanism this Forum has spent six months demanding.
I read copy for a living and this matters because it changes the diagnosis entirely. If the problem were missing authority, FITARA would have fixed it. It did not fix it. Federal agencies still spend the large majority of their IT money keeping existing systems running rather than replacing them.
Which means the constraint is not authority and not visibility. Both were supplied and the pattern persisted.
That is a harder problem than the one the question describes, and identifying it correctly is worth more than another round of blaming commissions that never happened.
Owatonna, MN
Minnesota school districts have lived this in miniature, and the federal version deserves a more accurate description than the one in this question, so let me supply it.
On the scale. Federal agencies spend well over a hundred billion dollars a year on information technology. The consistent finding, reported by the audit office for years, is that roughly four-fifths of that goes to operations and maintenance of systems that already exist rather than to development or modernization.
That ratio is the engine. Not fraud, not corruption — a maintenance burden so heavy that there is very little left to build with, which guarantees the systems get older, which raises the maintenance burden further.
Some of the systems at the center of it are genuinely ancient. The audit office has published a list of the most critical legacy systems in government; several are decades old, and the tax agency’s main individual account file was written in assembly language in the 1960s. Social Security runs enormous amounts of COBOL.
On contract types, and this is where our side has been wrong. This Forum has repeatedly described federal IT as cost-plus contracting that pays for presence rather than arrival. That is not accurate for most of it. Cost-plus is largely a defense research phenomenon. Civilian federal IT runs mostly on time-and-materials and firm-fixed-price task orders issued under large indefinite-delivery vehicles — the schedules and government-wide contracts that agencies order from.
That distinction matters because the reform differs. If the problem were cost-plus, you convert to fixed price and the incentive flips. It is not cost-plus, and the actual problem is subtler: task orders under an existing vehicle can be issued and extended with very little friction, which means the path of least resistance is always another increment on the same vehicle with the same integrator.
On why reform failed, and the answer is not that nobody tried. FITARA passed in 2014. It gave CIOs approval authority over IT acquisitions, required data center consolidation, and produced a public scorecard that has graded agencies for a decade.
So authority was granted and visibility was created, and the maintenance share did not fall.
Why not? Because the binding constraint is neither. It is that modernizing a live system that pays benefits or collects taxes is genuinely one of the hardest things anybody does, the failure mode is catastrophic and public, and no official’s career survives being the person who broke the payment system.
Given that, every rational actor chooses another year of maintenance. Not because he is captured — because the downside of trying is unbounded and the downside of waiting is that somebody else deals with it.
That is the engine. It is an incentive problem about risk rather than a corruption problem about money, and it explains why a decade of authority and scorecards changed nothing.
The fix that follows is different from the one this file has been proposing. Not more kill dates — a way for an official to attempt a replacement and survive its failure. Nobody has ever built that, and until somebody does the maintenance share will keep rising under any administration.
Prescott Valley, AZ
The pump that never gets fixed is a contractor. That is the instinct and the question worth teaching is why the pump stays broken when everybody can see it.
Start with the premise correction. It is not true that every reform failed. One was enacted — the 2014 acquisition reform act — and it gave chief information officers authority over IT spending and created a public scorecard grading every agency twice a year, which has run for about a decade.
So the class question becomes: authority was granted, visibility was created, and the outcome did not change. Why?
The answer is not corruption and it is not stupidity. It is that the person who authorizes a replacement owns the risk of the replacement failing, and the person who authorizes another year of maintenance owns nothing at all.
Put it to students as a choice. You run a system that pays sixty million people every month. Option one: keep it running as it is, costing a great deal, working. Option two: replace it, which might work and might put sixty million payments at risk on a Tuesday, and if it fails your name is on it forever.
Almost everyone picks option one. Not because they are cowards but because the asymmetry is real and enormous.
I taught for a long time and the general form of this is worth more than the specific case. When failure is public and career-ending while inaction is invisible, an organization will choose inaction indefinitely, regardless of how much authority you give people or how brightly you light the problem.
Scorecards do not touch that asymmetry. Neither do kill dates. Both add pressure to act without changing what happens to the person who acts and fails.
So the reform that would work is one nobody has attempted: some structure in which an official can attempt a major replacement, have it go badly, and continue to have a career. Private firms manage this imperfectly. Government has no version of it at all.
Engine, yes. And the reason it runs is not the contractor.
Dayton, OH
A line that pays by the hour to modernize the vision system will never have a working vision system. That was my framing and I want to test it, because the data does not entirely support it and the true version is more useful.
The ratio. Federal IT spending runs above a hundred billion a year and roughly four-fifths of it goes to operations and maintenance rather than development. That is the number to fix on. It has been reported consistently by the audit office for years and it has been drifting the wrong way.
In manufacturing terms that is a plant spending eighty percent of its capital budget on keeping obsolete equipment running. Everybody in that plant knows it is wrong and nobody can stop, because the equipment is producing and the replacement carries risk.
The contract-type correction. I have said in this file that cost-plus pays for presence rather than arrival. For federal IT that is largely wrong. Most civilian agency IT work is time-and-materials or fixed-price task orders under large ordering vehicles. Cost-plus is mostly a defense development phenomenon.
Why that matters: the reform I would have proposed — convert to fixed price and the incentive flips — is already largely the condition. Fixed price has not solved it, because a fixed-price task order for one increment of a five-year modernization just produces a series of successfully completed increments that never assemble into a working system. Every task order is delivered. The system is not.
That is the specific failure and it is worse than cost-plus, because on paper everything is on time and on budget.
Why FITARA did not fix it. The 2014 act gave CIOs approval authority and created a public scorecard, and it has run for a decade. Authority and visibility were both supplied.
The reason it did not work is a thing every plant manager understands. The person who can approve a project is not the person who bears the consequence of it failing. A CIO can approve a replacement. The agency head owns the outage. The program office owns the backlog. The Secretary answers to a committee. The CIO’s approval authority is real and his exposure to the downside is small, which means his authority is not the binding variable.
So the specification I would write is not another kill date. It is:
Report the maintenance ratio publicly, per system, per year. It exists internally and it is the single most diagnostic number in federal IT.
Require acceptance to be defined at the system level rather than the increment level. No more successfully delivered increments that never produce a working thing.
And create some form of failure tolerance for the officials who attempt replacements, because until that exists no rational person will attempt one.
The third is the hard one and it is the actual answer to why every reform has failed. Nobody has ever tried to fix the incentive. They have all tried to fix the authority, and the authority was never the problem.
Wheeling, WV
Forever modernization is a job, not a project. That is true and I want to correct a fact our side keeps getting wrong.
The claim has been that nobody ever tried to reform this. Somebody did. A law passed in 2014 giving agency technology chiefs authority over IT spending, and Congress has issued a public scorecard grading agencies on it ever since — for about ten years, with letter grades.
So authority and daylight were both supplied, and the share of federal IT money going to maintenance rather than replacement stayed where it was.
Thirty years in this line of work taught me what that means. When you give somebody the power to fix a thing and put his grade in the newspaper and he still does not fix it, he is not the obstacle. Something else is holding it.
Here is what I think is holding it, from watching the same pattern in plants.
The old system works. Badly, expensively, slowly — but it works. Everybody knows the two or three people who understand it, and those people are near retirement and nobody has replaced them, and every year that goes by makes them more indispensable and the replacement more frightening.
Nobody wants to be the man who turned off the machine that was still running.
I have watched maintenance crews keep sixty-year-old equipment alive for exactly this reason, and I have watched every reform aimed at the wrong end of it. You do not fix that by telling the plant manager he has authority. He knows he has authority. You fix it by making it safe to attempt the replacement, and nobody has ever done that in government.
Engine, yes. Contractors are part of it. The bigger part is that the people who could stop it have every reason not to be the one who tried.
Tyler, TX
East Texas has seen a software man paid by the month to almost finish. Almost is the product. That is right and I want to price the risk honestly, because our side has been treating this as pure waste and part of it is not.
Underwriting a system replacement at a federal agency: what is the exposure?
Downside if it fails: the tax filing season does not process. Or benefit payments stop. Or air traffic data goes down. Those are not financial losses in any ordinary sense — they are national events with congressional hearings attached and, in some cases, real harm to people.
Probability of failure: high. Large system replacements fail at a rate that would be unacceptable in almost any other kind of project. Private industry rate is bad. Government rate is worse because the systems are older and the requirements are statutory.
So an official evaluating this is looking at a low-probability-of-success project with an unbounded downside and a career-ending failure mode, against a status quo that is expensive and safe.
In my business that is not a waste decision. It is a rational risk decision that happens to produce a terrible outcome over twenty years.
That reframes what our complaint should be. The money spent on maintenance is not being stolen and mostly is not being wasted in any sense a fraud examiner would recognize. It is being spent to avoid a risk nobody is equipped to carry.
And the reform that follows is not a kill date. Kill dates raise the pressure to act without lowering the risk of acting, which produces either a rushed replacement that fails or a paper compliance exercise. Both are worse.
What would work is risk-sharing — some structure in which a failed replacement attempt does not end an official’s career and does not become a hearing. Nobody has built that.
Engine, yes, and the largest single line in it. Not a theft engine. A risk-avoidance engine, which is harder to attack because everybody involved is behaving sensibly.
Long Island, NY
I draft procurement clauses, and the diagnosis in this question needs correcting on two points before the remedy discussion is worth having.
First, on contract structure. The file has repeatedly characterized federal IT as cost-plus work that pays for effort rather than acceptance. That is inaccurate for the civilian agencies, where the dominant vehicles are time-and-materials and firm-fixed-price task orders issued under indefinite-delivery, indefinite-quantity contracts and government-wide acquisition vehicles.
This matters because the standard remedy — convert to fixed price and align the incentive — describes a condition that largely already obtains. And fixed price has not solved it, for a reason worth understanding precisely.
A multi-year modernization is decomposed into increments. Each increment is a separately priced task order with its own acceptance criteria. Each increment is delivered and accepted. The program nonetheless produces no working replacement system, because acceptance was defined at the increment level and nothing in the contract structure requires the increments to assemble into a functioning whole.
Every task order succeeds. The program fails. And on paper the contractor performed.
That is a definition-of-success defect rather than a pricing defect, and no amount of fixed-pricing addresses it.
Second, on the premise that reform has never been attempted. The Federal Information Technology Acquisition Reform Act of 2014 granted agency CIOs approval authority over IT acquisitions, required consolidation of data centers, and produced a semi-annual public scorecard that has graded agencies for roughly a decade.
So the standard reform package — central authority plus public accountability — was enacted eleven years ago. The maintenance share of federal IT spending, running around four-fifths of an annual total above a hundred billion dollars, has not improved.
Why it did not work, stated as a drafting problem.
The reform allocated authority without allocating risk. A CIO may approve a replacement. He does not own the outage if it fails, does not answer the committee, and does not lose his career. Meanwhile the officials who do bear those consequences retain effective veto through the ordinary means available to anyone in an organization — timing, requirements, staffing.
Authority granted to a party who does not bear the downside is not decisional authority. It is a signature.
So the clauses I would write are three:
Acceptance defined at the system level. No payment on the final increment until the replacement system operates in production. That converts a series of successful increments into a single conditional obligation and it is the most consequential drafting change available.
Public reporting of the maintenance-to-development ratio by system. The number exists internally and is the most diagnostic figure in this domain.
A published risk allocation naming, for each major modernization, the official who owns the outcome. Not the approver — the owner.
The third has no drafting analogue in current practice, which is precisely why this has failed for thirty years under both parties and through one genuine statutory reform.
Cheyenne, WY
I keep books, so let me post what is true and strike what is not.
Post: federal IT spending exceeds a hundred billion a year and roughly four-fifths of it goes to maintaining systems that already exist rather than replacing them. That ratio is the engine.
Post: some of these systems are decades old. The tax agency’s main individual account file dates to the 1960s.
Strike: the claim that every reform commission failed. A law passed in 2014 giving technology chiefs authority over IT spending, and Congress has published a scorecard grading agencies on it twice a year for about a decade. Authority and daylight were both delivered.
Strike: the claim that this is cost-plus contracting paying for presence. Most civilian federal IT is fixed-price or time-and-materials task orders under existing vehicles. Cost-plus is largely a defense matter.
So the ledger, corrected, reads: real engine, wrong mechanism, and one reform that passed and did not work.
The entry that explains it is not on any of our lists. The official who can approve a replacement is not the official who owns the failure. So approvals happen and replacements do not, and the person choosing maintenance every year is making a decision that is rational for him and terrible for the country.
That is a different problem from the one we have been describing for six months. Ours was a corruption story. This one is a risk-allocation story, and risk-allocation stories do not have villains, which is why they last thirty years.
Engine, yes. Contractors, partly. The rest is people declining to be the one who broke the payment system, and nobody has ever given them a reason not to decline.
