Why GPRA Can’t Be Shored Up

The Man Who Believed the Assignment

I.

Daniel Smith came to Washington in the summer of 1996 with two suitcases and a conviction.

The conviction was this: that the government was better than anyone knew, and that the reason nobody knew was that the government could not explain itself.

He was twenty-three. He had a degree in public administration from a state school his parents had not heard of until he enrolled, and he had read The Federalist Papers the way other young men read box scores. He believed — sincerely, unfashionably — that the American state was one of the great cooperative achievements of the species, and that it was being slandered daily by people who had never seen it work.

He got an internship at the National Institutes of Health. He was assigned to an office in Building 31 with a window that looked out on a parking lot.

And then he saw it work.

II.

The Human Genome Project was six years old that summer and still, in most rooms in America, a rumor. Daniel walked past the labs on his way to the copier. He did not understand what he was seeing. He understood only that people were doing it at ten o'clock at night, in worn cardigans, for salaries that would have embarrassed a second-year associate at any firm in the District.

There was a woman named Ruth on the fourth floor who had spent nineteen years on the pathophysiology of a single class of immune cell. Nineteen years. She had a child in high school who had never known a year in which her mother was not working on the same problem. Daniel asked her once, in the elevator, whether it ever felt long.

"It's the only thing I want to do," Ruth said. She seemed puzzled by the question. "Why would I want to do something else?"

There was no patent at the end of Ruth's work. There was no exit, no acquisition, no vesting cliff. There was a paper, and then another paper, and somewhere out past the horizon of her career there was a possibility — not a promise, a possibility — that some child not yet born would not die of something.

The best minds in the country were in these buildings. Not all of them. But enough that Daniel walked to the Metro in the evenings feeling something close to vertigo at the gap between what he had seen and what the country believed.

Because the country believed the government was a joke. The country believed it was fraud and waste and abuse, a phrase Daniel came to hate for its lazy trimeter. And the country believed this, in part, because when Congress asked the government what it had accomplished, the government could not produce an answer a human being could read.

He saw the budget documents. He saw the Congressional Justification that year — nine hundred pages of appropriation language, obligation tables, object class breakdowns. Salaries and expenses. Travel and transportation of persons. Rental payments to GSA. It was all true. Every number in it was true. And it said nothing at all.

Nowhere in nine hundred pages did it say: Ruth has been at this for nineteen years and she is close.

III.

That was the summer Congress was talking about the Government Performance and Results Act.

Daniel read it in a photocopied packet, sitting on the floor of a group house in Adams Morgan with a beer going warm beside him, and he felt the specific electricity of encountering your own inarticulate belief in someone else's finished sentences.

The Act said: agencies shall state what they are trying to do. They shall say how they will know if they are doing it. They shall report what it cost, and what it achieved, and they shall do this in language that a member of Congress — a busy, non-expert, well-meaning member of Congress — can actually use.

Yes, Daniel thought. Yes. That is exactly it.

The problem had never been that the work was bad. The problem was that the work was illegible. GPRA was a translation project. Somebody had to sit between Ruth and the Committee and build the sentence that connected them. Somebody had to be able to say: this much money, this goal, this progress, and here is the evidence.

He decided that night, on the floor, with the warm beer, that he would be that somebody.

He would be a budget officer.

It was, he understood, an unglamorous ambition. His roommates were interviewing at law firms. But it seemed to him that the leverage was here — that a thousand press releases about government efficiency were worth less than one honest schedule that showed what the money bought. The idealists all wanted to be the scientist or the senator. Nobody wanted to be the person who made the two of them comprehensible to each other.

Fine. He would do it. Somebody had to.

IV.

The people doing the science are not the people who can design mutually exclusive cost categories.

The interagency training was held in a windowless room at OMB with bad coffee and a facilitator from a consulting firm.

Daniel took notes the way he had taken notes in college — verbatim, in a black composition book, because he did not yet know which parts would matter.

The facilitator explained the requirement. The agency would define strategic goals. Beneath the goals, objectives. Beneath the objectives, performance goals with measurable targets. And then — this was the part Daniel underlined twice — the agency would present its budget request against that structure. Every dollar of the appropriation, distributed across the goals.

Someone raised a hand. What about the overhead?

"Allocated," said the facilitator. "The circular is explicit. Resources should be fully costed, with centrally funded administrative services and support costs allocated to each program."

Daniel wrote it down and felt, again, the electricity.

Fully costed. That was what would force agencies to explain ALL that they do and make it consistent and defendable. If the rent and the lawyers and the IT and the Director's suite all had to land somewhere — if there was no junk drawer, no residual, no "other" — then the schedule would be true, and if it was true it would be useful, and if it was useful then somewhere on Capitol Hill a staffer would open it and understand, for the first time, what Ruth's nineteen years had cost the taxpayer and what they had bought.

Every dollar lands once. Nothing omitted. Nothing counted twice.

He drove home that night thinking it was the most beautiful assignment anyone had ever given him. He thought of ways to defend that nineteen years with nothing to show yet except the advancement of human knowledge about how our bodies work – is the very point. Who else would fund it – he was ready to write the true story and defend it.

V.

He was, by then, twenty-six, and he had a real job: budget analyst, then budget officer, at an agency small enough that he was one of four people who understood how the money actually moved.

He understood how the money actually moved. This is worth dwelling on, because it is the thing that everyone outside a budget office fails to grasp about the people inside one.

Daniel knew that the appropriation came in three accounts. He knew which account paid for the building and which one couldn't. He knew that the Office of the Chief Financial Officer was funded out of a working capital fund that billed the programs, and that the billing formula was a fiction inherited from 1987 and never revisited, and that this meant one division was subsidizing another by about four hundred thousand dollars a year, and that nobody had ever asked. He knew where the money was. He knew what it was doing. He could have drawn you the map on a napkin.

What he did not know — what he could not know, because it had not been decided — was what the goals were.

So he went to ask.

VI.

The first meeting was with the program directors. Twelve of them around a conference table, and Daniel at the end with his composition book and a one-page handout explaining the requirement. Ruth was there.

He explained the assignment. He explained that he needed a goal structure — a set of categories that covered everything the agency did, so that every dollar could be assigned to one and only one of them.

He explained exhaustive and he explained mutually exclusive, and he watched twelve faces go politely, opaquely blank.

Then a man named Whitcomb, who ran the largest division, said: "With respect, this isn't a question for you. If congress wants our goals, that is for us."

Daniel said he wasn't sure he followed.

"The goals," Whitcomb said, "are what we're trying to achieve. You don't know what we're trying to achieve. You know what things cost. Those are different." There was a murmur of agreement around the table, the particular warm murmur of a room discovering it has a shared grievance. "This is a program question. We should be the ones defining this."

And Daniel — this is what he would think about for years afterward, at four in the morning, when the thing that woke him was not the fraud but his own manners — Daniel said: That's fair, but the goals have to cover the whole budget  - all of it covered so I need to help with that part.

And the managers said, we will tell you what we want to do, you worry about the budget part after. That is your job.

So they had the next meeting without Daniel, and the next, and the next.

Of course the scientists should say what the science was for. Of course he, a twenty-six-year-old with a composition book, should not be the one deciding whether the agency's purpose was to advance understanding or to deliver interventions. He had come to Washington precisely because he believed the people doing the work were better than the people describing it.

So he handed them the assignment. He said: you write the goals. Take a month. I'll build the numbers around whatever you give me.

They took four months.

VII.

What came back was seven pages.

Daniel read it three times at his desk with the door closed.

Goal 1: Advance the frontiers of knowledge to improve human health and well-being.

Goal 2: Foster excellence, innovation, and collaboration across the research enterprise.

Goal 3: Ensure that the benefits of discovery are translated equitably to all communities.

Goal 4: Cultivate a workforce and infrastructure worthy of the mission.

Goal 5: Educate scientists of tomorrow from communities underrepresented in science.

And beneath each one, four to seven objectives, each of which was a sentence of the same kind: Strengthen. Enhance. Promote. Support. Leverage. Catalyze.

He sat with it for a long time.

It was not bad writing. It was moving.

But he could not put a dollar in it. He looked at Goal 5 – that one he recognized, one scientist he knew wanted to have a program to attract and mentor female and minority scientists – she wanted this to happen but it has not happened. No such program exists.

Then he thought of all the money the agency spends on vaccines and vaccine distribution – all the logistics money spent – it is hard to say logistics money spent on getting vaccines distributed - advances the fronteirs of knowledge. If you’re being very honest it isn’t directly covered in any of the five goals. So where, Daniel wondered, does he put all the money spent on the logistics of the vaccine program?

Next Daniel took a single grant — a five-year award to a lab studying protein folding — and he tried to place it. Did it advance the frontiers of knowledge? Yes. Did it foster excellence and collaboration? It was a multi-institution consortium, so yes, obviously. Did it translate benefits equitably? The consortium had a minority-serving institution as a partner, so yes. Did it cultivate the workforce? It funded four postdocs. Yes.

One grant. Four goals. All four true.

He tried it the other way. He took Goal 3 — ensure that the benefits of discovery are translated equitably — and he tried to find the money in it. Which appropriation line was that? Which division? Which contract, which FTE, which square foot of leased space? It was not a thing the agency bought. It was a quality the agency hoped its purchases would have.

You cannot cost an adverb.

And then he went looking for the rest of it — the parts that were not glorious. The general counsel. The FOIA office. The mail room. The nineteen people in the Office of the Chief Information Officer. The rent. The rent was eleven million dollars.

There was no goal for rent.

Of course there wasn't. Nobody writes a strategic plan that says Goal 5: Pay for the building. The program directors had written down the things they were proud of. Naturally. Every person in that room had answered the question they were actually asked, which was not how does the money divide but what are we for.

And the answer to what are we for is never mutually exclusive, and it is never exhaustive, and it does not have a line for rent.

He had asked twelve honest people to do a job, and they had done a different job, beautifully, and it was not the job.

The thing that was most disappointing to Daniel is what he wanted to do – he wanted to do what the assignment called for – make what he saw being done every day – clear. Tell the story with the American people’s taxpayer dollars directly behind the work he saw.

He wanted to say that altruistic brilliant people are getting paid lower than industry salaries to do the long range, not for profit science, that could cure disease and help us understand how our own bodies are built. The kind of science no profit driven company would do, but benefits all of us. He wanted this to tell the story; but advancing frontiers, fostering excellence, and ensuring benefits – these goals do not tell the story he saw, and he knew it.

VIII.

He went back to them.

He was careful. He brought a draft. He said: I've tried to build the schedule against your structure and I'm running into a problem, so let me show you.

He showed them the protein-folding grant landing in four goals at once.

He said: if I put it in all four, the total is four times the appropriation. If I split it, I need a basis for the split, and I don't have one, and neither do you, because there isn't one. If I put it in one, I have to pick which — and any pick I make is a statement about your science that I am not qualified to make and you have not authorized.

He showed them the eleven million dollars of rent with nowhere to go.

Whitcomb looked at the page for a while.

Then he said: "So put it in Goal 1."

Daniel said: all of it?

"The rent supports the research. The research is Goal 1." Whitcomb shrugged, not unkindly. "Daniel, you're overthinking this. It's a budget document. Nobody reads it."

IX.

He escalated.

He wrote a memo to the CFO. It was four pages, it was scrupulously polite, and its argument was simple: the schedule we are about to submit cannot be made accurate, and I would like the record to reflect that I said so.

He proposed alternatives. He proposed building the goal structure from the appropriation up — starting with the accounts, the divisions, the actual cost-bearing categories, and naming them in mission language rather than starting with mission language and hunting for costs. He proposed a residual line, honestly labeled, for the things that were not goals. He proposed, if all else failed, publishing the schedule with a stated methodology and a stated uncertainty, so that at least the reader would know what they were holding.

The CFO read it. The CFO was not a bad man. The CFO said: "I hear you. That's a real concern. Let's make sure we hear from the program side too."

And Daniel understood, standing in that doorway, that he had just been converted from an officer into an opinion.

His objection — which was a factual claim about arithmetic — had been received as a point of view, to be weighed against other points of view, and the other points of view belonged to people who did not have to sign the page, but had more authority inside the organization.

X.

The examiner's call came in September.

Daniel had submitted the budget materials without the goal schedule and with a two-page note explaining why. He had rehearsed the conversation in his car. He was going to explain the double-counting. He was going to explain the rent. He had brought the protein-folding grant as an example, because it was concrete and undeniable, and he believed — he genuinely believed, at thirty-one years old, having by then been in Washington for eight years — that if he could just get someone at OMB to look at the arithmetic, they would see it.

The examiner listened. She was not hostile. She sounded tired.

When he finished she said: "Okay. But we can't accept the justification without the schedule."

Daniel said: I understand. That's why I'm calling. I'm telling you the schedule can't be made accurate.

"Right," she said. "But it has to foot."

There was a pause on the line in which Daniel felt something in his chest reorganize itself permanently.

He said, slowly: I'm telling you the numbers won't be true.

"And I'm telling you," she said, "that the columns have to sum to the appropriation. Look — I don't make the circular. If it doesn't come in with the schedule we'll have to hold the submission, and if we hold the submission you're looking at a stop-work. Do you want me to send you an example? Interior does a nice one."

She was a person with forty agencies and a deadline, holding the only instrument anyone had given her, which measured one thing.

Whether it added up.

XI.

He built it.

That is the part he never found a way to say out loud, in all the years afterward, without his voice doing something he didn't like. Not they made me. Not I had no choice. There is always a choice; the choice was to resign, and he did not resign, because he had a mortgage and a daughter and because — and this is the true reason, the one he only admitted to himself much later — because he still believed the system could be fixed from inside it by a person who stayed.

He sat down on a Sunday in October with a spreadsheet.

Goal 1 got the research divisions. Goal 2 got the collaboration office, which existed, and the training grants, which sort of fit. Goal 3 — equitable translation — had no budget line anywhere in the agency, so he took twelve percent of the extramural awards, because twelve percent of the awards had a community partner named in the abstract, and he wrote a footnote explaining the twelve percent that he knew no one would read.

The rent went to Goal 1.

The general counsel went to Goal 1.

The Chief Information Officer went to Goal 4, cultivate a workforce and infrastructure worthy of the mission, because the word infrastructure was in it.

The vaccine money went nowhere so he left it off. The rent belonged nowhere, so he hid it under research. Lab research belonged in three places so he put it in three, and the triple count nearly offset all the lines that went nowhere.

The schedule footed because his mistakes cancelled and the leftover figure that needed to get on the chart to get it to match the full budget – he put it in the aspirational goal that doesn’t actually exist at all.

He sat with that for a while. The only test anyone had ever applied to this document — the only property OMB would hold up a budget over — was a property that two offsetting fabrications could satisfy exactly as well as the truth.

The research if triple counted, almost made up for the missing vaccine money that was left off completely.

He moved four hundred thousand dollars from one column to another at eleven at night because the columns had to sum to the appropriation and they were four hundred thousand dollars short, and there was no principled place for it, and so he put it where it would make the sum come out.

Then he printed it.

Then he signed it.

The schedule footed. Every column summed. Every dollar landed exactly once. It was, in the only sense that anyone had ever tested, perfect.

He drove home and sat in the driveway with the engine off for twenty minutes.

XII.

They loved it.

He had expected, at worst, indifference. What he got was praise.

The schedule went into the Congressional Justification. It was clean. It was clear. It had five goals and a total that matched the appropriation to the dollar, and a staffer on the Hill said — this got repeated back to Daniel three times, by three different people, each of whom thought they were paying him a compliment — that it was the first time anyone had shown them what the agency actually spent its money on.

The Deputy Director put it in a speech.

The schedule appeared in the annual report. Then in the strategic plan. Then, reformatted, on a poster in the lobby: a pie chart, five wedges, in the agency's colors.

Daniel stood in front of the poster one morning holding a coffee and looked at the twelve percent.

XIII.

And then he emailed the CFO and told him – I am not doing it again next year, and the CFO said, I understand now we have the splits, I will do the table next year and I will sign it.

The first audit finding came fourteen months later. It was mild, almost apologetic in its language: the agency's cost allocations to strategic objectives were not supported by contemporaneous documentation.

Which was true. There was no documentation. The documentation was Daniel, on a Sunday in October, with a spreadsheet.

He was called into a meeting to explain the remediation plan. He did not write the remediation plan. He came to understand, over about six weeks, that the remediation plan had been written by a vendor.

The vendor proposed a time and attendance system.

It cost — the first year, with implementation — a little over two million dollars. It would allow every employee of the agency to allocate their hours, in increments of a quarter hour, against the strategic objectives. This would produce, the vendor's slide said, an auditable evidentiary basis for the agency's performance-based cost allocations.

Daniel said, in the meeting, that this would not make the numbers true. It would make them documented. Those were different things. You could not fix a broken category structure by asking two thousand people to code their hours into it; you would only be distributing the error across two thousand people and calling the result data.

He was told that the finding had to be closed.

XIV.

They bought the system.

And Ruth — nineteen years on a class of immune cell, a daughter now in college, still in the worn cardigan, still there at ten o'clock at night — Ruth got an email explaining that beginning the first of the month, she would be required to allocate her time, in quarter-hour increments, across five strategic objectives.

She came to Daniel's office. She had printed the email. She was not angry. She was something worse; she was bewildered.

"I don't know what this is asking me," she said. She put the page on his desk and pointed at Goal 3. "Ensure that the benefits of discovery are translated equitably. Daniel, I look at cells. I've looked at the same cells since 1988. Is that Goal 1 or Goal 3?"

And Daniel — the budget officer, the one who understood how the money actually moved, the man who had come to Washington to build the sentence connecting Ruth to the Committee — Daniel looked at the woman who had given nineteen years of one life to one problem, and had to tell her that the honest answer was that it did not matter, that she should pick one, that no one would ever check, and that the number this produced would be printed in a document, and audited, and put on a poster in the lobby.

She said: "But that isn't true."

He said: "No."

She looked at him for a moment.

Then she went back upstairs and coded her hours, because she was a good employee, and because she did not want to make trouble, and because she assumed — the way every honest person assumes, when confronted with an absurd requirement from an institution they respect — that somebody, somewhere, must have a reason.

Two thousand people at that agency coded their hours that year. Then two thousand did it again the next.

The lie did not stay on Daniel's desk. A lie that is documented must be sustained, and to sustain it you need everyone. You need Ruth. You need the postdocs. You need, eventually, every grantee, and every subrecipient of every grantee, and the case manager in a treatment clinic in Ohio typing a woman's answers into a tablet, a meals on wheels volunteer in Mississippi were given the reporting codes to pick from, then given training, then told they can use up to 20 percent of their grant money to report back against the categories.

The fiction had been pushed down. It was resident, now, in two million keyboards. And every one of the people typing assumed the person above them had a reason.

XV.

He kept fighting.

He wrote to the Committee. Twice with his name on it, and then — after the second letter produced a call to his CFO and a conversation about lanes — anonymously, four more times, on a personal laptop from a coffee shop in Silver Spring, laying out the arithmetic in language he tried to make impossible to misread.

He never got a reply. Years later, he saw a committee report complaining that agency budget justifications were drowning in pleonasm and devoid of useful information, and he sat at his kitchen table and laughed until his wife came in to see what was wrong.

He testified once, briefly, to an interagency working group. He was thanked for his input.

He went to conferences. He found, to his surprise, that there were forty or fifty people in the country who understood exactly what he was talking about, and that all of them were budget officers, and that all of them had signed the same page.

They would find each other at the coffee urn, and one of them would say something like so how's your twelve percent, and everybody would laugh, and it was the loneliest laughter Daniel had ever been part of.

XVI.

And then OMB reversed, or rather codified what had already happened inside agencies.

That was 2011.

The word came down in a revision to the circular. There was no announcement, no hearing, no vote. There was a new edition of a document, and in the new edition, the section that had contained the requirement — resources should be fully costed, with centrally funded administrative services and support costs allocated to each program — did not contain it anymore. The section number was reassigned to something else. The phrase performance budget simply stopped appearing.

In its place: a website. Goals, objectives, milestones, quarterly reviews, goal leaders, priority goals, a dashboard with green and yellow and red.

No dollar column.

And the authority to define the goals and the measures — the assignment Daniel had gone to Washington to carry, the one he had handed to the program directors in a conference room in 1999 because they told him it wasn't a finance question — that authority was now formally, theirs. It was in the statute. Goal leaders. Program managers. The people who did the work would say what the work was for. They don’t know the budget. They expressly do not want to know. So the goals would never tie to costs, never not duplicate costs or cover everything the agency does, and they could officially include something a manager wants to do but spends no money against.

The doctors picked the goals. The goals had no dollars. The dollars stayed in the appropriation accounts, where they had always been, where Daniel could have drawn you the map on a napkin at any point in the entire eighteen years if anyone had given him the authority to do it.

And the schedule he had forged — the one that footed, the one on the poster — was no longer required.

Nobody told him that. He found out by reading the circular.

Nobody rescinded the time and attendance system either. Along with the parallel financial system that codes expenditures to the same 5 categories,  across all of HHS, they cost eight hundred thousand dollars a year to maintain. Two thousand people still coded their hours into five strategic objectives that no longer had to reconcile to anything at all.

XVII.

Last spring Daniel was asked to prepare a document for a new agency initiative, and he opened a legal pad — he still uses legal pads — and he started, out of an old habit of thoroughness, by listing everything he would have to produce that year. Submissions he builds with input and work hours that extend to every employee in the organization.

The strategic plan. The annual performance plan. The annual performance report. The Congressional Justification. The Agency Financial Report. The Summary of Performance and Financial Information. The Agency Priority Goals, quarterly. The Cross-Agency Priority Goal contributions, quarterly. The Budget In Brief. The Federal Program Inventory submission. The Evidence Act learning agenda. The Evidence Act capacity assessment. The Annual Evaluation Plan. The Enterprise Risk Management profile. The internal control assurance statement. The improper payments risk assessment. The improper payments sampling and estimation plan. The DATA Act submission, monthly. The FFATA subaward reporting. The Single Audit corrective action plans. The IG's top management challenges response. The GAO recommendation status update. The records management self-assessment. The Section 508 accessibility certification. The Federal Managers' Financial Integrity Act statement. The FISMA report. The Privacy Act SORN reviews. The Paperwork Reduction Act clearances. The training certifications — cybersecurity, ethics, records, harassment, insider threat, each annual, each tracked, each auditable.

And then the ones that were nobody's law and everybody's requirement. The performance scorecard one administration had wanted, in green and yellow and red, still generated, still reviewed. The progress on audit findings. The equity assessments a second administration had required, still on the books, still collected from every grantee. The plain-language rewrites. The customer experience surveys. The one-page summaries with the pictures. The table that says you spend money on diversity and inclusion. The new one that certifies you do not.

He wrote them all down. It took two pages.

He looked at the list and understood, with a clarity that had nothing to do with anger, that not one of these had ever been removed. Every administration had added its vocabulary. None had subtracted one. And every one of them was still enforceable — an Inspector General could still write a finding against any line on that page — which meant that he would continue, producing every single one, for an administration that had left, for a purpose that had lapsed, for a reader who did not exist.

He thought about the fact that no one had ever added it up. Not OMB. Not GAO. Not CBO. In thirty years, nobody had been asked what the machine cost, because the machine was made of accountability, and it is very difficult to get anyone to audit an audit.

He thought about Ruth, who had retired in 2019, and who in her last decade had spent — he had actually calculated this once, and then wished he hadn't — something like nine full working weeks of her life entering her hours into categories he had invented on a Sunday in October to make a column sum.

Nine weeks. Out of a life spent looking at one class of cell, because it was the only thing she wanted to do, because why would she want to do something else.

He thought: I came here to make the government legible.

He thought: I have spent twenty-nine years making it illegible, at enormous expense, and every single step of it was compliance.

He sat for a long time in an office with a window that looked out on a parking lot.

Then he wrote, at the bottom of the second page, in the same handwriting as everything else:

Should I go?

And under that, after a while:

If I go, who tells them?

XVIII.

The other budget officers left.

It never looked like an exodus, because it never happened faster than one at a time. A retirement here. A move to a contractor there, for double the salary. A woman he had known for eleven years, who had built her agency's cost model from the accounts up and could tell you to the dollar what the mail room cost, took a job at a think tank and did not say goodbye to anyone.

The new ones were bright, they were credentialed, many of them had master's degrees. They arrived on their first day and were handed the file.

Fifteen years of time and attendance data. Two thousand employees, quarter-hour increments, coded against five strategic objectives, going back further than any of them had been alive in this business. Cost allocation models with documented methodologies. Crosswalks. Reconciliations. Audit trails with minimal findings, because the systems had been bought.

The new budget officer would sit down with all of it and try to make it make sense.

And it would not make sense. Of course it would not. Goal 3 had twelve percent of the extramural awards in it because on a Sunday in October, in a year before they were born, a man had counted the abstracts that contained the phrase community partner. Goal 5 held a plug. The rent was research. The vaccine logistics were nowhere at all, and the triple-counted lab money had covered the hole so neatly that no one had ever looked for it.

The new budget officer would stare at this and feel a specific, familiar shame — the shame of the intelligent newcomer who cannot follow the reasoning and assumes the failure is his own.

I'm new. It's complicated. There must be a rationale I haven't found yet.

And then, because he was diligent, he would go looking for the rationale. And he would find the data. Fifteen years of it. Thousands of people, hundreds of thousands of entries, every hour accounted for, every quarter-hour coded.

And he would think: well, this is documented.

The data was all that survived. And data, to a young person who has been taught that data is the antidote to fiction, looks like a foundation.

He watched them arrive and trust it. One after another. They cited the twelve percent in their own analyses. They built on it. One of them presented a variance report to the Deputy Director explaining a shift in the Goal 3 allocation, and the explanation was thoughtful, and internally rigorous, and entirely about a number that had come out of nowhere in 1999.

Daniel said none of this is real many times. And he was, told back, no one reads it, but they require it.

So the fiction did not merely persist. It matured into evidence. And the newest, smartest, most rigorous people in the building were now its defenders — because they had checked it against the data, and the data agreed with it, because he had built the data to agree with it.

XIX.

Daniel is a fictitious character and the categories doctors picked here are fictitious though the point and errors they demonstrate are true, but the accomplishments listed below, that NIH achieved in the timeframe of this narrative – those accomplishments are real, and you probably haven’t heard of all of them because they never got written up in a report that sounded like the one Daniel dreamed of.

And here is what happened, out the window, at NIH.

Between the summer Daniel arrived and the spring he sat down with the legal pad, the age-adjusted cancer death rate in the United States fell by about a third.

 

A measured rate, against a peak in 1991. Something on the order of four and a half million people did not die of cancer who would have died at the old rate. They are not identifiable. They do not know who they are. They are simply — statistically, indisputably — here, walking around, in a country that spent those same decades being told the government could not accomplish anything.

Five-year survival across all cancers went from sixty-three percent to seventy. For children, from fifty-eight percent in the 1970s to eighty-five.

Invasive cervical cancer among women in their early twenties — the first cohort to receive the HPV vaccine as adolescents — fell sixty-five percent in seven years. A vaccine against a virus, preventing a cancer, in a generation of women who will never know they were spared, because you cannot feel the absence of a tumor, and NIH made the vaccine possible, and few people know and it.

HIV became a chronic condition. A twenty-year-old starting antiretroviral therapy in 2010 could expect roughly a decade more of life than one starting in 1996. Some of them are grandparents now.

And in December of 2023, the Food and Drug Administration approved the first CRISPR gene therapy in human history, and what it cured was sickle cell disease — a disease that for a hundred years had been a sentence, borne overwhelmingly by Black Americans, treated for most of that century with palliation and apology.

At NIH they edit the patient's own cells. They give them back. The cells do not sickle.

The tool that did it came out of basic research into how bacteria defend themselves against viruses.

Nobody in the private sector funded the sickle cell research that led to the cure. Nobody could have. There was no business case. There was no market. There was a woman in a worn cardigan, somewhere, at ten o'clock at night, following a question because it was interesting, funded by a public that never met her.

That is what a national research budget is. It is a country buying, on behalf of people not yet born, a set of answers that no one can predict and no one can price.

And Ruth is in that history. Not with her name on it — that is not how it works. Her name is on forty papers, and the papers are cited, and the citations are cited, and somewhere four steps downstream is a mechanism that became a target and a target that became a drug. Daniel looked it up once. It took him an afternoon. He is not a scientist and he could not follow all of it, but he followed enough to know that a woman on the fourth floor of a building with a bad parking lot spent nineteen years — then twenty-nine, then thirty-four — on one class of immune cell, and that people are alive because she did.

Nine of those weeks she spent entering her hours into a category he invented to make a column sum.

And here is the thing Daniel cannot get past, that he turns over at four in the morning, that he has now spent twenty-nine years failing to explain to anyone with the power to fix it:

None of that is in the report.

Not one line of it. Not the four and a half million. Not the sixty-five percent. Not the children going from fifty-eight to eighty-five. Not the cure — the actual, literal, first-in-human-history cure — for a disease that had never had one.

The report says: Goal 1: Advance the frontiers of knowledge to improve human health and well-being. $1.6 billion.

The report says: Goal 3: Ensure that the benefits of discovery are translated equitably. Twelve percent.

He came to Washington because he believed the government was better than anyone knew, and that the reason nobody knew was that the government could not explain itself.

He was right.

He was right about all of it, and he still is, and that is the unbearable part — that the diagnosis was correct and the cure was worse than the disease, and that he has spent his working life building the machine that made the thing he loved less legible than it was when he found it.

The 1996 Congressional Justification told you nothing, and it took nine hundred pages to do it.

The current one tells you nothing, and it takes twice as many, plus a website, plus a dashboard, plus two million people entering quarter-hours into categories that reconcile to nothing, plus a compliance apparatus whose total cost no one has ever calculated — because the apparatus is made of accountability, and nobody audits an audit.

They cured sickle cell disease.

He sat for a long time in an office with a window that looked out on a parking lot.

Then he wrote, at the bottom of the second page, in the same handwriting as everything else:

Should I have walked out all those years ago?

And under that, after a while:

If I go, who tells them?

He did not have an answer to the second question.

He has not left. The thing he came to fix is still broken. He fears he broke it further. It is still worth fixing.

He is still writing letters and witnessing.

 

 

This is a fictional story told to illuminate something that is shockingly complex and difficult to make people understand. So, is it really like this?  Is it government wide?  Here is what examiners found:

 

In its major review of GPRA implementation, GAO concluded:

1.       Only one of the six agencies whose reports it closely reviewed clearly linked its costs to the achievement of performance goals or objectives.

2.      GAO found managers struggled with the system.

GAO surveyed federal managers across government.

Managers reported continuing difficulty with:

  • setting outcome-oriented goals,

  • measuring performance,

  • collecting useful performance data,

  • developing meaningful outcome measures,

  • especially in research agencies, where outcomes are inherently difficult to quantify.

3.      GAO found that

·         fewer than half of federal managers received relevant training in results-oriented management.

·         and

·         managers lacked the authority, training, and incentives needed to implement GPRA effectively.

That's government-wide.

4.      GAO also found that agencies were linking funding requests to higher, more general levels of expected performance rather than to specific performance goals.

In other words, the connections were vague, not precise.

 

Although Daniel is fictional, the management dilemma he encounters was not unique to one agency. GAO found that managers across government struggled to establish meaningful outcome measures, collect useful performance data, and link costs to performance goals. In one government-wide review, only one of six agencies clearly linked costs to the achievement of performance goals or objectives.

GPRA et al  - - Why It Doesn’t work now and how it could

 

Why GPRA Can't Be Fixed the Way Everyone Keeps Trying to Fix It

Everyone agrees federal performance reporting doesn't work. GAO says so every year. Congress says so at hearings. The agencies say so privately. What nobody has explained is why thirty years of fixes haven't fixed it.

Here is why.

In 1993 Congress asked agencies for what is, unmistakably, a financial report — the full cost of what each agency produces, every dollar landing against something. In 2010 the assignment was formally handed to program managers: scientists, physicians, engineers, social workers. People who came to government precisely because the work there isn't ruled by money, and who take a certain pride in never thinking about it.

That's the cause. It is not the finding.

The finding is what happens next. When you give a misunderstood assignment to a thousand specialists, you do not get one error repeated a thousand times. You get something closer to three hundred different errors, each repeated a few times, none of them the same shape as any other.

The scientist who thinks the assignment is to show off his best work names his favorite activity and quietly omits half his program's budget. Another believes three outcomes from one dollar is good stewardship, and is proud of it. A third assumes the data will eventually be used against him and reports defensively. A social worker writes that her program clears minds to promote spiritual growth. Fish and Wildlife names the steps of its own pipeline — Listing, Consultation, Conservation, Recovery — so that consultation's cost sits inside planning's cost sits inside recovery's. A budget officer, cornered, spends a million dollars on a time-and-attendance system to force employees to code into categories that cannot hold a dollar, so that she can certify something.

Those are not six versions of one problem. They are six different problems. Omission, double-counting, defensive reporting, uncostable abstraction, nested taxonomy, and expensive machinery built to launder nonsense. Each requires a different correction. No two of them respond to the same remedy.

And every reform since 1993 has been a single remedy.

That is the whole trouble with explaining this. Show a reader one failure and he will collapse the mess into that failure — he'll hear "duplicate reporting" and propose deduplication, and he will be addressing perhaps two percent of the problem while believing he has addressed all of it. Which is worse than doing nothing, because it spends money and closes the question.

So this paper does something that will feel excessive. It accumulates. It shows failure after failure, agency after agency, because the irreducibility is the finding, and there is no way to demonstrate irreducibility except by piling it up until the pattern is undeniable — the pattern being that there is no pattern.

What comes out the other side is not a database of unreliable numbers. It is an artifact with the shape of data and none of the content: it foots, it audits, it is carried to the dollar, and it contains no information whatsoever. Congress and GAO, by reading it seriously and demanding better versions of it, have spent three decades conferring on it a credibility it has never earned.

The fix is not to train the managers. You cannot train five hundred program managers into a skill that is rare among finance professionals. But there is a fix, and it is smaller than anyone expects, and this paper ends with it.

 

 

GPRA asks for a full-cost segment report; that is a financial task- the government's own accounting standards board said so in 1995 and required it in 1998; it was handed to mission staff; mission staff produce category sets that cannot carry cost; everything downstream is noise laundered into false precision.

The state of GPRA inside agencies is not "unreliable data." It is that there is no dataset. There are 500+ agencies perhaps 25 percent of valid data× hundreds of idiosyncratic misunderstandings × thousands of T&A codes × two million certifiers × every grantee in America — producing an artifact that has the shape of data (it foots, it audits, it's carried to the dollar) and contains no information whatsoever. And Congress and GAO, by reading it seriously and demanding better versions, keep conferring on it a credibility it has never earned.

The fix isn't "train the managers." You can't train 534 program managers into a skill set that's rare even among finance people.

You take twenty unusually disciplined people, and they build the product/service structure for all 500+ reporting agencies from budgets, division names, and admin expenses in a few months — then help managers name outcome measures that hang correctly off product lines that actually cost out. Twenty right people, not five hundred idiosyncratically brilliant ones in their diverse specialties.

 

 

 

 

Letter One — What Happened

The report Congress asked for is a financial report, and the government has no one whose job it is to write it.

Why the Government Performance and Results Act and its 2010 modernization have never produced what Congress asked for — and what would.

Members of Congress:

You have asked, reasonably, a question that deserves a real answer: performance-based budgeting works in the private sector, so why has thirty years of it failed inside the federal government?

I have spent a career inside that failure — on the staff of five cabinet-level agencies, trying each time to make these reports mean something. The honest answer is not that government is lazy, or that the agencies are hiding the ball. The answer is structural, and it can only be understood in layers, so I am going to send it to you in layers. This first letter establishes the one thing everything else rests on:

The report you asked for is a financial report. And the government has no one whose actual job it is to write it.

That sentence sounds impossible. Let me show you why it is true.

What the report actually is

Picture a company about to go public. For the first time in its life, it must produce audited financial statements for investors — and for the first time, it must show its costs the way investors require: the full cost of each thing it makes, spread so that every dollar the company spends lands against something it sells.

Before the IPO, the company could tell itself a comfortable story. "This product costs us eight dollars to make, and we sell it for thirty." On one level that is true — eight dollars of materials and direct labor went into the unit. But it is the most dangerous kind of true, because it quietly leaves things out. It does not carry the general counsel's salary. It does not carry the rent, the CFO's own office, the executives at the top, the shared systems, the security desk in the lobby. Those costs are real, and they were spent to make and sell that product too — but they don't sit on the factory floor where anyone can see them.

When the company goes public, it is no longer allowed to tell the comfortable story. It has to spread the entire budget across the product lines — including the general counsel, the rent, the leadership, every shared cost — until nothing is left unassigned. And when it does, the eight-dollar product is revealed to cost fifteen. It still sells for thirty; it is still a good product. But the return an investor was about to bank on was inflated by everything the comfortable story left out. Investors deserve to know the fully weighted cost of what they are buying into. That is the entire reason the discipline exists, and the reason getting it wrong carries real legal risk: a company that reports the eight-dollar figure when the truth is fifteen has misled the people whose money it took.

Now — that is precisely what Congress wanted from federal agencies. One consistent, fully loaded answer to "what does this government good or service actually cost?" — an answer that could not be $15 when the agency needed more money to make more stuff, but then $8 when the question was about cutting the budget.

And you wanted agencies to have to state plainly what it is they do.

Those are good and legitimate things to want.

They are also, unmistakably, a financial assignment — even the, what do you produce question, is a financial assignment, because finance knows the whole company better than any other single worker, and they know how expenses are tracked, and are unable to be tracked, giving them the best view to make sure all production is covered in the answer of, what do you produce.

In a large publicly traded company it is unthinkable to give such an assignment to anyone other than the CFO.

Hold onto that, because it is the hinge of everything: the thing you asked for is a financial document.

Three differences stand between the government and that IPO report. The first two you may partly know. The third, almost no one in Congress knows, and it is the one that matters most.

First: the Heads of the Agencies and the entire top political layer seldom have ever worked in the ranks of the agency.

In the company, a chief executive owns the enterprise, knows its products, and can be held to the number. In a federal agency, the equivalent of the CEO is a presidential appointee who sometimes arrives with no familiarity whatsoever with the agency he now runs. Rick Perry has said he did not understand, when he asked to become Secretary of Energy, that the Department is responsible for the nation's nuclear weapons complex. Pete Buttigieg arrived at the department of Transportation without direct experience.

The President appoints and congress confirms Secretaries with no experience working inside the agencies – that is not without a cost.

Second: there is little activity on the revenue side of the government ledgers.

This one is bigger than it sounds, and it reaches all the way into the culture. A company lives on the tension between what it earns and what it spends; that tension is the engine that makes cost discipline matter. Program managers are more aware of company organizational and cost structures in private companies than most government program managers are.

A federal agency, for the most part, has one source of revenue: its appropriation. It is essentially a single entry. (Yes, there are revolving funds and fee-for-service exceptions — but they are exceptions.)

That single fact gives the CFO in a COMPANY power.

The culture in the government is not driven by finance in excellent and not so great ways.

A great many of the government's finest people — scientists, physicians, engineers, researchers — came to government precisely because it is a place where the work is not ruled by money. They can do pure science without being driven by revenue, without being bothered by it, in many cases without ever having to think about it at all. That is a feature; it is part of why the government can do things no company would.

But it means the people doing the mission are, by disposition and often by design, the people least fluent in cost, and least fluent in organizational structure, or administrative functions. Remember that when we get to who was eventually handed this financial assignment.

Third — and this is the one Congress does not know: a government CFO is not a private-sector CFO. The job you are imagining as CFO is actually two different jobs inside government agencies, held by two different people.

When you picture performance-based budgeting working in a company, you picture the CFO doing it — because in a company the CFO owns both the money coming in, the money spent, and the structure. Inside the federal government, that single role is split in two, and the half you need most for this assignment, is not the half you would expect, nor is it the half you handed the assignment to.

The federal CFO is, very often, an accountant elevated from running the agency's systems of expenditure and, not its budget system. This person's world is the internal accounting ledger — what the government calls AMSCO-type lines, categories of spending that Congress seldom sees and does not appropriate against. The CFO tracks obligations and unliquidated obligations, performs reconciliations, and carries the financial-statement audit. This is real, skilled, essential work. But notice what it is: it is expenditure and audit work, on a chart of accounts that is not the budget you fund. The CFO frequently holds a CPA and rose through financial systems, not through programs. In smaller agencies especially, the CFO can be less familiar with the agency's own divisions, its organizational structure, and the actual products the agency produces than you would ever guess. (In my own career about 70 percent of the CFOs I worked with had accounting and ledger background while about 30 percent had budget and program – these are NOT the same systems or accounts.)

The person who does know those things is the budget officer. The budget officer works in the categories Congress actually appropriates and caps against. The budget officer has to keep the organizational chart current, and pushes money appropriated from congress to divisions and object classes — they produce budget forms, the strategic plans, and the performance materials. The budget officer is the one who understands the programs, who gets pulled into performance, who can see how a division's money maps to what the division does. Sometimes a budget officer is promoted into the CFO chair — but more often the CFO seat is filled by a CPA from the systems-and-audit side, and the two functions drift apart.

So when you imagine a competent CFO producing your IPO-style report, understand what the government actually offers for that task: a CFO fluent in expenditures and audits but often distant from programs and structure, and a budget officer fluent in programs and structure and the appropriation — two people, holding between them the two halves that a private CFO holds in one pair of hands. And the half your report most needs — the half that knows the divisions, the programs, and the budget structure well enough to build categories a dollar can fit inside — lives with the budget officer, generally, the lower-status of the two.

What Congress did with the pen

When GPRA was written in 1993, the exercise was tied to the budget's own structure — built on program activities, where a costable number can actually live. It landed in finance, assigned to the CFO – generally above the Budget Officer. Not perfectly placed, but often the budget officer did play a role.

Then, in many agencies as the CFO or budget officer tried to teach managers what they needed to do to craft good performance measures, program managers pushed back hard - - imagine a low status CFO or budget officer telling a world renowned bee scientist, or rare disease specialist or a dr Anthony Fauci – level researcher, they must produce performance categories and when the categories overlapped and had nothing to do with how money was allocated the finance people push back on the program managers.

Perhaps they write a measure that could capture all money spent without double counting, the cancer researchers said – excuse me you pay the bills and cost things out – I tell you what I do – you don’t tell me. Enough renowned program managers complained hard enough that eventually…

in the 2010 GPRA Modernization Act, the pen moved in writing the way it already had inside agencies — and it moved the wrong way. In practice the assignment was already taken away from the finance side and handed to the program managers. The modernization act codified it into law.

The program managers did not want to be dictated to by what they regarded as accounting clerks telling them what their performance measures should be. We set the measures, the argument went; let the budget staff cost them out afterward.

Sit with who those program managers are. They are, very often, exactly the mission experts I described earlier — the ones who came to government to do the work and not to be ruled by money, who have no training in finance and, in many cases, no wish for any. They are excellent at what they do and most of them never think about finance and it is a source of pride that they don’t.

And the pen for a financial assignment was placed in their hands, with the budget officer — the one person structurally equipped to produce the thing Congress asked for — sent to the back of the room to "cost out" whatever the mission side chose to name.

That was the start of the failure. Everything that has gone wrong since — the reports you rightly find to be alphabet soup, the numbers that will not reconcile, the categories that fit nothing — descends from this single inversion: a financial report, whose only correct author was the budget officer, was assigned instead to the people who came to government specifically to get away from finance.

 

 

In the previous letter we established that the assignment for determining product lines (goods and services) and the goals and objectives for those activities that would capture all the money the agency spends without double counting anything, was, in the government, not given to finance nor was it given to people who knew enough about the full organization or the expenditures of it  - the way it is in publicly traded companies.

Now we must take a look at what giving, what is first and foremost a budget driven task to people with little to no knowledge of the budget, did.

This will take more than one letter to explain because when you give an assignment to over 1000 people who do not understand the nature of the assignment, you do not get one outcome repeated 1000 times, you get about 300 different outcomes repeated a few times.   So if you want to know the state of things now, understanding only one scenario will dangerously flatten the truth into a single thing that can be fixed a single way, and believing that would create more waste.

 

Imagine Proctor and Gamble being asked to provide the full weighted cost of all its products for the first time ever – imagine it never did it before but it must do it now, today.

So instead of starting with finance you go to product line managers and ask them to imagine what impact they have on the world and what outcome goal they have. 

The align product manager (probiotics) says they allow people to vibrate at a higher rate by promoting health and wellness.

The clear blue easy manager says they ensure survival of humanity.

The polident guy says we prevent dentures from slipping,

 and on and on.

Then they set their less lofty more grounded annual performance targets and twenty different managers have something that uses the word “clean” in it, a version of making things cleaner.

Finally you involve the finance people – you give them the performance data – just a database not coded by product or division – the data just shows that someone recorded that something is cleaner – what is cleaner?  Dentures?  Dishes? Hair? An esophagus?  When you ask the align manager why she didn’t specify her target better, so it can’t be confused with Dawn Dish soap’s cleanliness goal, she says, I didn’t even know we made dish soap.


 

The state of GPRA in agencies now is not uniform. You gave a financial assignment to a thousand people who don't do finance — and misunderstanding isn't uniform. It's idiosyncratic. So you don't get one wrong answer you could correct for. You get:

  • The scientist who thinks the assignment is show off your best work, so he names his favorite activity and simply omits half his program's money.

  • The one who thinks three outcomes from one dollar is stewardship — and is proud of it. (show them the triple-count; the COO says "we heard your objection, let the meeting continue.")

  • The paranoid one who thinks this data will be used against us, and reports defensively.

  • The social worker who writes "we clear minds to promote spiritual growth."

  • The FWS people who name pipeline steps — Listing, Consultation, Conservation, Recovery — a taxonomy of verbs, not products, so consultation's cost is inside planning's cost is inside recovery's.

  • The budget officer who, cornered, buys a million-dollar T&A system to force employees to code into the nonsense so she can certify something.

  • And thousands more.


 

Why It Works There and Not Here

Letter Two — How the mess got made: watch it happen in a simplified company

This letter shows what can happen when the work is done by good people who do not know the assignment is financial — and cannot be made to see it. The caution is that this is a gross simplification – this is one of a thousand different ways a person who does not understand the assignment can try to do it.

 

A bakery

You own a bakery AND Anna does all the books for it. You are the CEO, you know the books and the products. One day your investors ask for an accounting: what did you spend on flour, and what did the bread bring in? What did the cakes cost, and what did they sell for? And What about catering give me the cost and revenue for that too?

You answer it in an afternoon. Flour, butter, the ovens, the two people up front — you have the receipts. Bread here, cakes there, catering in its own column. Every dollar you spent lands in exactly one place, and every dollar you earned lands next to the thing that earned it. A stranger could read it and see that catering is carrying the shop and the cakes barely break even.

The categories followed the money. There is a receipt for the flour and an invoice for the catering job. The boxes you were asked to fill were the same shape as the way the cash actually moved. That is the entire reason it worked.

Now a second letter comes. The investors have been reading a management book and they notice two things about your answers – anna’s salary is not included in any of the product lines, neither is rent. So, the numbers you sent them aren’t the full cost of the whole operation spread over the product lines.   

Also, "Bread, cakes, catering" tells them what you sell, but not what good you do.

Redo the report — total dollars, nothing double-counted — but give let  the people who actually make the products, tell us what they do and the good they do in the world.

So you ask your baker to think of categories that describe what good he does in the world, and whatever he says is what you will have Anna cost out and report. He gives you a list and you give it to Anna to put costs against his categories.

 

Anna picks up her pen, and cannot move. Because his categories are Nourishment. Community. Craft.

You have a receipt for flour.  The flour is Nourishment — it feeds people. But your sourdough is why the neighborhood calls this place theirs, so the flour is Community too. And it is the thing you have spent twenty years learning to handle, so it is Craft more than anything in the building. One sack of flour. Three categories. All three true. And the form says pick one, or split it — and there is no honest split, because the sack did not arrive divided. It is not sixty percent Nourishment. It is not any percent of anything. It is flour.

The first report was easy and the second is impossible, and the difference has nothing to do with difficulty. The first categories were built on how the money moves. The second were built by someone imagining what the bakery is about without any regard to whether the imaginings can be tied to the budget. One follows the receipts; the other follows a feeling. And you cannot cost a feeling, because the register never once recorded a sale of "Community."

That is the federal performance report. "Bread, cakes, catering" is the budget — divisions, programs, object classes, the way the money actually moves.

"Nourishment, Community, Craft" is the strategic-goal framework — a set of beautiful words describing what the agency is about, divorced from any knowledge that the assignment needed to be tied to budgets.

Had a person who understood the books and that the assignment was financial been pushed to produce outcome measures that cost out all expenses, she could have written: we strengthen our community by providing bread, by providing deserts, and by providing catered foods; but unlike in publicly traded companies who are good at this, part one of this assignment was taken away from anyone with knowledge of the budget structure.

Who wrote the three words — and who was overruled

Who chose "Nourishment, Community, Craft"? Not the investors — they only said "tell us the good you do." Picture the categories chosen by your head baker. He has been baking forty years. So he picked the words he loves — Craft, Nourishment — and, being decent rather than vain, he turned to the woman who runs the front of house and said, warmly, you pick one too. She said Community. He nodded. It sounded right.

Two good people, describing the work they love, in the language they love. And in that warm, collegial moment they built a report that can never be costed — because their three words all describe the same sack of flour, and neither could see it, because they have never once prepared financial reports or thought about them.

Since this bakery is really a stand in for federal agencies, in many agencies, “Anna pushed back, trying to educate the ‘bakers’ that all expenses had to be accounted for within the categories and items can’t be double counted.

Anna is right. And Anna is overruled.

Not gently. The head baker has forty years in; Anna does the books in a back office. When she says the categories won't cost out, what he hears is that the bean-counter does not understand baking.

You've never made a loaf in your life, he says. How are you going to tell me what my work is worth? In this building the baker is the artist and Anna is the help. When he gives his little speeches, he thanks "the people in the back who keep the lights on." That is her whole standing.

So the pen is taken from the one person who could have made the categories costable and handed to the one person who could not — because he has the standing, and the investors, when they visit, want to talk to him. They have never met Anna. Anna is moved out of the meeting where the categories are set, and invited back for one thing only: to take the words the baker chose and make them add up.

That is what happened in the federal government.  

The proof that it is comprehension, not evasion

For years, the people who received these reports assumed the agencies were being evasive — dodging, padding, hiding.

In the 1990s, a GAO analyst named John Kamensky surveyed more than a hundred federal components about their plans and measures. The responses came back positive (that head beaker loves the three words they came up with), but the reporting was plainly overblown, and GAO's internal reaction initially was that the agencies had lied on the survey.

Kamensky corrected his colleagues, and his correction is the truest sentence anyone has said about this program. It was not that they lied, he said. It was much worse — they "do not know the difference between a strategic plan and an annual work plan, or the difference between an output and an outcome."

Sit with that, because if you want to know the state of things you need to keep imagining who you gave what assignment to. The hazardous waste program manager inside FAA, the invasive species expert that analyzes plants that could wipe out corn crops in Idaho, the manager that is overseeing the switch from radar to satellite technology to land planes across the world, all told to imagine outcomes, and measures and when what they pick excludes more than half of the money spent in their programs, and a budget person tries to educate them about what the assignment is, they pull rank and will not have a finance clerk dictate performance targets to them.

 

 

 

Who Holds the Pen

Every number in this document is drawn from the Office of Personnel Management's own FY 2025 Congressional Budget Justification.

Before you read this as a member of Congress, read it as someone else

You have a role. You know it well. For the length of this document, set it down.

For the next few minutes you are not a legislator. You are the budget officer at the Office of Personnel Management — the civil servant whose job is to take the money Congress appropriated and show, on the form Congress requires, what that money bought.

Do not play your own part. Play hers. Everything that follows depends on your sitting where she sits and feeling what she feels, because the thing almost no one in Washington understands about these reports can only be understood from that chair. It cannot be explained. It has to be attempted.

So: you have the pen. Let us begin.

Part one — the assignment you cannot complete

On your desk are two things.

On your left: the money. It is clean. It always has been. OPM's appropriation arrives divided the way Congress divides all money — by fund and by object class — and it maps, to the dollar, onto the offices that spend it. You can read it at a glance:

Division (what it does)

Approx. budget

Retirement Services — processes federal retirement claims

~$155M

Healthcare & Insurance — runs FEHB, PSHB, FEDVIP, life insurance

~$73M

HR Solutions — sells HR services to other agencies (earns fee revenue)

~$310M (revolving)

Office of the CIO — runs OPM's information technology

~$277M

Workforce Policy & Innovation — issues federal HR regulations

~$68M

Merit System Accountability — audits agencies' HR

~$18M

Suitability (SuitEA) — security and suitability determinations

~$12M

Office of the CFO — runs OPM's finances

~$20M

Every one of those numbers has a receipt behind it. This is the easy report. A stranger could read it and know what OPM is.

On your right: the four goals. Congress, under the Government Performance and Results Act and its 2010 modernization, requires you to take that same money and lay it out not by division but by strategic goal. Here are OPM's four, exactly as published:

  1. Position the federal government as a model employer — measured by a government-wide employee satisfaction index.

  2. Transform OPM's organizational capacity and capability.

  3. Provide an exceptional customer experience.

  4. Provide innovative, data-driven solutions to agencies.

Your assignment is simple to state: divide all of the money on the left into those four goals on the right. Every dollar must land in exactly one goal. The four goals must add up to the whole appropriation.

Pick up the pen. Start with your biggest program.

Retirement Services. $155 million. Which goal does it fund?

Is processing a widow's annuity "positioning the government as a model employer"? Partly, maybe. Is it "customer experience"? The retirees would say so. Is it "data-driven solutions"? There's an IT system involved. Is it "organizational capacity"? In some sense.

It is a little of all four and cleanly none. There is no goal named administer retirement. Your single largest, most concrete, most costable program — the thing that most is OPM — has nowhere honest to land.

And it only gets worse from there. HR Solutions, $310 million — a business that earns revenue selling services to other agencies. Which of the four aspirational goals does a fee-for-service business belong under? The CIO's $277 million serves retirement and healthcare and policy and everything else — it cannot go under one goal without lying about the other three.

Now feel what your hand is being asked to do. You have two, and only two, honest options, and the form allows neither:

  • Go back to the real categories — report by division, the way the money actually moves. Forbidden. The form demands the four goals.

  • Build a fixed, written rule that assigns each division a fractional share of each goal, and defend that fiction forever. But there is no defensible rule, because there is no fact underneath it — Retirement Services is not 40% "model employer" and 25% "customer experience." Those percentages don't exist to be discovered.

So you do the third thing. The only thing left. You type a number into each goal box that looks plausible and makes the totals foot. And because a few divisions genuinely fit nowhere, you invent a fifth box — call it "additional mission support activities" — and you sweep the homeless money into it.

You have just built OPM's real report. That fifth box is real. It is the largest single line in the agency.

Part two — now you are king for a day

Stay in the budget officer's chair. But now imagine you have something she has never had: authority.

Here is how this works everywhere it actually works. In a publicly traded company, the person who builds this exact schedule is the CFO. When a company tells its investors what its bread line costs and what its catering line earns, the report foots to the penny — and it foots because the CFO builds the categories from the budget and has the authority to make the divisions conform to them. The product lines are defined so that cost can attach. Every dollar has one home. Nothing is left out. It works this way, and it works no other way. No company has ever produced a credible segment report by asking each department head to name the outcome he's proudest of and then trying to make the money match.

(One translation note for government: inside a federal agency the right person is usually not the CFO but the budget officer. The government CFO is an accountant who works the internal-expenditure ledger; it is the budget officer who works the object-class structure that Congress actually appropriates and caps against. That distinction is exactly why this keeps failing — even when the job was handed to "finance," it often went to the wrong seat within finance. Give it to the person who reads the appropriation.)

So watch how easy — and how correct — this becomes when a budget officer does it from the budget categories, the same way every publicly traded company does. Let's do it here, now, together:

Goal A — Administer the nation's largest retirement and insurance benefit system efficiently and with high customer satisfaction. (Captures Retirement Services + the retirement side of Healthcare & Insurance — every dollar, once.)

Goal B — Deliver reliable, high-value health, dental, vision, and life insurance to the federal community. (Captures Healthcare & Insurance program operations, including the new Postal program.)

Goal C — Set and steward government-wide human capital policy that upholds merit-system principles. (Captures Workforce Policy + Merit System Accountability. Note the scope: what OPM's policy shop actually controls — not the workplace conditions at agencies OPM doesn't run.)

Goal D — Provide agencies cost-effective, high-quality HR shared services on a self-sustaining basis. (Captures HR Solutions — reported like the fee-earning business it is, on its own revenue-versus-cost statement, never blended into the appropriated goals.)

Goal E — Safeguard the integrity of the federal workforce through timely suitability and credentialing determinations. (Captures SuitEA.)

Goal F — Equip the government with authoritative human capital data and analytics. (Captures the data and analytics division.)

Goal G — Manage OPM's own enterprise responsibly (IT, finance, facilities, leadership). (Captures the CIO, CFO, and overhead — either shown transparently as what it costs to run OPM, or allocated down onto Goals A–F on a written, auditable basis. Either is honest. What is not honest is calling it "Goal 5" and pretending it's a mission.)

Sum Goals A, B, C, E, F, and the overhead in G, and you reach the discretionary appropriation — to the dollar, FTE for FTE. Goal D reconciles separately to its own revenue. There is no residual, because when you start from the ledger there is nothing left over. The junk drawer had to exist in the old report only because the goals were built from aspirations that the money didn't fit. Build the goals from the money, and the drawer disappears.

 

Part three — now take off the king's hat

Here is the budget officer's actual life. She does not set the goals. The goals were set — by the mission side, in a room she was not in — and they are the four you already met: a government-wide satisfaction index, "organizational capacity," "customer experience," "data-driven solutions." She cannot change a word of them. Her job, and the CFO's, is only this: make the money foot into those four boxes, and make it survive an audit, every year, forever.

So sit back down in her chair, with no authority, and solve the problem.

OPM's outside auditors, KPMG, reported for FY2025 that the agency "utilizes numerous legacy feeder systems that are not integrated to its core financial system" — a significant  deficiency in internal control over financial reporting, with a recommendation that pointed directly at the "allocation of personnel resources." In plain terms: the audit found that OPM cannot cleanly tie its people's cost to what it reports.

There is no evidence that the people of OPM work in the categories listed as goals in its budget. We could have told you that just by looking at its division names in the budget and then reading the goals.

Now read OPM's response, written into its own budget as a formal strategic objective — and note who owns it. Owner: the Chief Financial Officer. Deputy owner: the Associate CFO for Budget and Performance. The objective reads:

"Implement a process and structure that enhances labor code management and reporting across the agency" — and "leverage enhanced reporting and external studies to support future budget justifications."

. The audit says the money doesn't tie together cleanly. So the CFO commits, as a goal with her name on it, to build a labor-coding structure whose stated purpose is to feed the budget justification:

Let me make it simple your goals you swear to congress you allocate ALL your money against read like this: Position the federal government as a model employer.

I can’t imagine any person in OPM doing that – I mean maybe you could argue that administering a survey about job satisfaction indirectly works some toward that – but it’s a stretch. 

But these are your goals – they are hard to change – impossible for you to change them you have no authority.  So you have to allocate costs against goals that have little to do with the work your agency actually does – so what do you do – what the auditor told you to do – and they probably recommended it because it is what every other agency does with their own mismatched goals – make time and attendance categories that match the goals and make every employee pick only from them – then certify that that is how they spent their time.

 

 

So let me ask you to change hats one more time. Take off the budget officer's hat.

You process retirement applications. That is your job — you did eleven of them today; you fixed a survivor-benefit error. Concrete, real, nameable work.

It is the end of the pay period, and every hour must be coded in the new labor management system the audit recommended. You look at your dropdown. There is no code that says processed retirement applications — because "administer retirement" is not one of the four goals. Your choices are "model employer," "organizational capacity," "customer experience," "data-driven solutions." None of them is what you did. You must pick one anyway. You must pick one for all eighty hours. And then you must put your name to it and certify that it is true.

You pick customer experience and data driven solutions – you did not interact with a single customer and data driven solutions is meant to capture the time of the budget office, but what else can you pick?

And having decided — having typed something you know is not what you did, and signed it — how do you feel about your job?

Now understand that the person in that chair, coding time into bogus categories is not always a clerk. At NIH it is a physician who understands a disease mechanism almost no one else alive understands, stopping every two weeks to decide whether her hours were "model employer" or "organizational capacity." At DARPA it is an engineer picking the code that clears. And here is the cruelty at the center of the whole apparatus: the more integrity a person has, the worse this is for them. The person who doesn't care picks a code in four seconds and forgets it. The person who does care — who wants the record to be true — is handed no honest option and made to enter a falsehood under her own name, every two weeks, for as long as she stays. You have built a machine that asks your most conscientious people to do the one thing they cannot get numb to. So they leave. And the system records it as ordinary attrition — never knowing it just lost exactly the people it most needed, and because almost no one understands it, if asked people who leave say – I hated all the red tape.


 

 

 

The Bee Scientist

An inside account of why the federal government has spent thirty years and something like a hundred billion dollars a year producing numbers that mean nothing — and why almost no one who touches them can see it.

There is an acre of restored wet meadow somewhere in the National Wildlife Refuge System. It was restored once, with one pot of money, by one crew, on one Tuesday. The pollinator biologist counts it under "acres of pollinator habitat restored," and he is right — the meadow feeds his bees. The migratory bird biologist counts the same acre under "acres of migratory habitat conserved," and he is also right — the meadow stages his birds. One acre. Two measures. Two true claims on a single indivisible thing.

Now watch what happens downstairs, at the desk of the budget analyst who has to attach cost to outcomes. The acre cost ten thousand dollars. If she books ten to pollinators and ten to birds, she has spent twenty thousand dollars the agency never had; the page doesn't foot; that's an audit finding. If she splits it five and five, she's asserting the acre was half a bee acre and half a bird acre — a number she pulled from the air, which the bee man would reject because the meadow is fully his and the bird man would reject because it's fully his, and which no transaction in the ledger supports. That's a finding too. There is no split that is both arithmetically honest and biologically honest, because the acre is arithmetically one thing and biologically two things, and the form demands it be exactly one.

And if you are a budget person you might say oh lord just split it 50/50 – I need to tell you that the program managers are asked questions by auditors and each want credit for 100 percent. They think that the same acre producing two outcomes is a triumph and outside of a budget form it is, but congress passed a law, and inside that law and that reporting the budget must drive the measures – a one to one ratio. The full budget totaling the cost of all the outcome measures added together the two must be equal.

And after budget officers spent from 1993 to 2010 pushing back – trying to tell program managers that their measures can’t double count money, and they must include all activities and all money spent in their programs. Then in 2010 the program managers officially won the argument they’d been winning inside agencies for years.

Congress, in the 2010 GPRMA (update to GPRA) officially gave bee scientists and engineers and social workers the authority to name their own measures.

Feel the click of that. At the Fish and Wildlife Service the same acre accomplished two things — and the sentence that is a triumph in the mission office is a catastrophe in the budget office, and they are the same five words, and after 2010 the bee scientist and the bird scientist both claimed 100 percent of the same acre of land and told the budget officer to cost it out but give them both 100 percent, in a form that must equal 100 percent of the budget and not a penny more, but no worries, because so many of the activities and expenses were excluded from the measures that the problem in the aggregate was never that the measures when costed and added together would go over the whole budget, no matter how many times things were over counted.

 

 

 

Read the categories cold

You don't have to take my word for any of this. The categories give themselves away. Here are the real ones the Fish and Wildlife Service uses for its endangered-species work:

Listing. Planning and Consultation. Conservation and Restoration. Recovery.

Read them as a set and they betray what they are.

Compare them to what they should be. A real product-based, full-cost-of-service structure names what an agency delivers — and every category is a distinct output, and together they exhaust the work, so every dollar has exactly one home and nothing is left out. These do neither.

Look at "Consultation" sitting there as a top-level category. Why is it there? Because somebody knew there were consultation contracts, consultation activity, a visible thing people do with money attached. But consultation isn't an output. It's a step — you consult in order to plan, in order to conserve, in order to recover. It's a verb in the middle of producing something else. They elevated a process step to a product line because it was legible to them as a thing that happens, not because it's a separable end-product with its own cost.

That one word tells the whole story. These aren't products. They're activities — things people do, named in the vocabulary of the people who do them, flowing into one another, overlapping by construction. Which is exactly what you get when the mission side writes the list: mission people describe their work — we list, we consult, we restore — and out comes a taxonomy of verbs that can never carry separable cost, because consultation's cost is part of planning's cost is part of recovery's cost. One pipeline, described as four products, by people who did not know that "product" was the assignment.

The category set has holes and overlaps at the same time, because it was never built to the standard that would make cost-of-service mean anything: every output covered, each one distinct. The agency doesn't know that standard exists. They think listing the activities is the assignment.

 

Underneath the measures sits a management-information system with a thousand categories. Four hundred are in use. Six hundred are zombies — codes that should have been retired years ago and never were, dead enough to be meaningless, alive enough to be picked. And every one of two million federal employees, every two weeks, has to charge their time to one of them.

So think about the man working at the airport in Atlanta who, needing a code that will clear, charges his hours to next-generation radar-to-GPS research. He is not doing GPS research. He is doing whatever he actually does. But now a line item somewhere in the federal accounts records that the government spent his salary advancing satellite navigation science. He doesn’t think he lied. He picked a code that worked, because the true code doesn't exist and the coding is the interruption, not the job. Multiply that by two million people, every pay period, for thirty years.

This is not a double-count. A double-count is orderly — you at least know the acre went two places. This is a random-count. The number at the top of the page isn't inflated in a direction you could correct for. It's noise wearing decimals. And no scientist reading the literature, no biologist restoring a wet meadow, no airport worker in Atlanta is producing signal that got blurred. There is no signal. It was garbage at the keystroke, and everything built on top of it inherits the garbage and then launders it into false precision — the $27,064,000 that Congress will fence off "not to exceed" for Listing, carried to the dollar, mostly composed of misfiled hours and an electric bill nobody could find a home for.

And the largest part isn't even the mission work. In every place I worked — and I worked in more of them than most people ever see — the ratio held: for every hour of the mission work the categories are built to name, there are three to five hours of work they cannot name. Paying the electric bill. Writing the RFP for janitorial services. Listing a vacancy. Running safety training. Which code did you want for that? The electric bill isn't a fraction of Recovery; it's the substrate under Recovery and Listing and Consultation and the visitor center, all at once — and there's five times more of it than of the thing the form knows how to ask about. The tail is bigger than the tooth, and the tail has no home, so it gets scattered across mission boxes at random, and the mission boxes were fictions to begin with.

Why the right people were in the room and it failed anyway

When Congress first demanded this — GPRA in 1993, then the Modernization Act, then the DATA Act, a Sarbanes-Oxley for government, pass after pass — the job landed, at first, exactly where it should: on the Chief Financial Officers. The finance shops. The people who actually understood cost. And they tried to do it the only way it can be done — from the ledger up, which means walking into the room where the agency's goals and objectives and performance measures get set and saying: your outputs have to be defined so that cost can attach to them.

And they were thrown out of that room.

Removed. The goals-and-measures meeting is the mission side's room — high-status, purpose-defining — and the bill-payer does not get a vote on what the agency is for. So the Budget Officer was relegated to the one room he was allowed in: reporting. Handed the outputs after the fact, after the bee scientist and the bird man had already traded their measures across the table, and told: now cost these. Invited to perform the impossible half of the job, barred from the room where the impossibility was manufactured.

This is the thing almost no one sees, and it is the whole thing. In the private sector, cost-based budgeting works because finance has authority — the chart of accounts is imposed downward, the divisions must conform, and behind it all stand an auditor who won't sign and an SEC that will prosecute. The outputs partition because the cost system was allowed to define them.

In government, the status structure is upside down. Finance is the servant; mission is the master. So the one function with the competence to build the number has none of the power to impose the discipline the number requires — and you cannot make your master conform to your spreadsheet. The good CFOs, the ones who saw exactly why it couldn't tie out, either burned out eating audit findings for a fiction they were handed, or left. The system experienced their leaving as ordinary attrition. It was actually the loss of the only people who understood.

 

 

 


 

Thousands of agencies gave a financial report to people who feel above finance, many have a disdain for it, almost none understand it. The truth of the sate of performance measurements inside most agencies is far far worse than you think it is.

It is far worse than I can convey here. Most agencies have produced a word salad of time and attendance codes – hundreds and thousands of them – training for staff – all to force data to correspond to categories that realistically do not represent what the agency does  - or double counts – or excludes, and that is the point. It is not one problem. It is all of the above. It is a scientist saying let’s highlight this activity, thinking the assignment is to show off your best, and the conspiracy theorist saying this data is going to be used against us, and the social worker saying we clear minds to promote spiritual growth, and a budget officer saying I am not going to certify this unless I can show hours employees actually spent working on this, then buying a million dollar time and attendance system to force employees to code time into nonsense, and it is thousands of more scenarios, that created thousands of performance categories that we now force every meals on wheels, and hospital, and research lab to certify time into.

 

It isn't one failure mode reproduced 1,000 times. It's ~300 different failure modes, produced by 1,000 people who each misunderstood the assignment in their own way. You gave a financial assignment to a thousand people who don't do finance — and misunderstanding isn't uniform. It's idiosyncratic. So you don't get one wrong answer you could correct for. You get:

  • The scientist who thinks the assignment is show off your best work, so he names his favorite activity and simply omits half his program's money.

  • The one who thinks three outcomes from one dollar is stewardship — and is proud of it. (You showed them the triple-count; the COO said "we heard your objection, let the meeting continue.")

  • The paranoid one who thinks this data will be used against us, and reports defensively.

  • The social worker who writes "we clear minds to promote spiritual growth."

  • The FWS people who name pipeline steps — Listing, Consultation, Conservation, Recovery — a taxonomy of verbs, not products, so consultation's cost is inside planning's cost is inside recovery's.

  • The budget officer who, cornered, buys a million-dollar T&A system to force employees to code into the nonsense so she can certify something.

  • And thousands more.

Each of those is a different defect requiring a different correction, and they're all live simultaneously, in the same agency, in the same document. That's why "just fix the double-counting" or "just integrate the systems" is not merely inadequate — it's dangerous, because it implies a single lever.

Understanding one scenario will dangerously flatten the truth into a single scenario that can be fixed a single way, and believing that would create more waste.

 

To spread costs against the actual products and services government produces can be done but it crosses financial systems, requires methodologies to spread overhead. It requires discipline, and a deep understanding of the assignment.

The right group of twenty intelligent, clear eyed people could look at a budget, division names, and administrative expenses and come up with solid financially driven product and service areas in as little as six months – for all 500 plus reporting agencies. They can be reviewed and agreed on by experts, agencies and congress. Then after that, if given the authority by congress, the same group, could help managers name outcome measures that are written appropriately (that cost correctly).

Make no mistake this is a hard assignment for most people. It requires a specific set of skills that are unusual, a disciplined way of thinking, checking, and keeping the financial task in mind. For the right team of people it is quite possible. It is not possible however to train 534 different program managers, most brilliant in their field, to acquire that skill set.

 

 

 

 

 

 

 

 

 

 

No One Will Believe It Is This Easy

A dialogue with the objections

The hardest thing about this fix is not implementing it. It is getting anyone to accept that it could possibly be that simple.

Because here is what it is:

Throw out the performance measures. All of them. Throw out the time-and-attendance coding built to feed them, and the parallel financial systems built to reconcile to them, and the dashboards and the scorecards and the quarter-hour certifications. Then let a small team with the right skills tell the agencies who they are — from the ledger — and hand it to them.

Thirty years. A hundred billion dollars a year, by some counts, and nobody has ever added it up. Two million people coding hours into categories that reconcile to nothing. And the answer is: delete it, and let twenty people who know how to do this do it properly.

Nobody believes that. The disbelief is the obstacle, not the work. So let us take the objections in the order they actually arrive.

"What team? Who could possibly do this?"

This is the first objection and it is the right one, so let us answer it properly.

There is an old saying: show me your budget and I will tell you who you are.

That saying is usually deployed as a moral point — you spend on what you value. But it belies something else, something literal and technical and much more useful: there exists a person who can actually do that. Who can be handed a budget, a schedule of divisions, and a list of administrative expenses, and produce from it a true account of what an organization does, what it produces, and what each of those things costs — fully loaded, every dollar landed once, nothing omitted and nothing double-counted.

That skill set exists. It is uncommon, but it is not mysterious, and it is not new.

It is the skill of the people who implement Sarbanes-Oxley. Who cut an enterprise's ledger into segments and product lines that carry full cost. Who write the methodologies that spread overhead and shared services down onto the things an organization actually makes. Who do this under audit, under penalty, for every publicly traded company in America, every year, and get it right.

It is not the skill set of an epistemologist. It is not the skill set of a cancer researcher. And it should not be.

We do not want the woman who has spent nineteen years on one class of immune cell to become fluent in cost allocation. We want her looking at the cells. That is what we are paying her for. That is what the country needs from her. Asking her to build a mutually exclusive cost partition is not merely a waste of her — it is a category error about what she is for.

So: let the people with the skill do the thing they have the skill for. Let them look at the budget and show you what the agencies spend their time and money on. Let them do it well, the way they do it in the private sector, where it works — and where it works no other way.

Then hand that to the agencies.

That is the answer to "what team." The team is the one that has been doing exactly this, correctly, in another building, the entire time.

"But the books say this cannot be done."

They do. And they are the most serious obstacle in the room, because they are not stupid books and the people who cite them are not stupid people. So let us take them at their strongest, one at a time, and then take them apart.

Wilson — Bureaucracy (1989)

The claim. Agencies divide into four types by what can be observed. Production agencies have observable outputs and observable outcomes. Procedural agencies have observable outputs but not outcomes. Craft agencies have observable outcomes but not outputs. Coping agencies have neither. Measurement regimes designed for production agencies are actively destructive when applied to the others — and NIH is a craft agency at best. Do not attempt to measure it.

This is the intellectual license behind thirty years of not trying.

The refutation. It conflates two entirely different problems and hides the conflation inside a single word.

You cannot cure sickle cell disease and simultaneously claim the outcome is unobservable. Cancer mortality down a third. Pediatric five-year survival from fifty-eight percent to eighty-five. HPV cohort cervical cancer down sixty-five percent. HIV to a chronic condition. The first CRISPR cure in human history. That is the most observable outcome record of any enterprise in the country.

What is not observable is the attribution — you cannot trace this dollar to that cure, because the causal chain is thirty years long and runs through forty labs and a mechanism nobody was looking for. Fine. True. But "unobservable outcome" and "unattributable outcome" are not the same problem, and Wilson's taxonomy smears them together and then concludes: don't try.

The correct reading is the opposite. The agency-level outcome is spectacularly observable. It is the sub-agency allocation that is fictional — and it is fictional precisely because someone tried to build it from aspiration instead of from the ledger.

And the outputs? The outputs are trivially observable. Intramural research operations. Extramural competitive awards by institute. The Clinical Center. Research infrastructure. Training awards. Countermeasure logistics. Every one has an invoice. Every one exhausts to the appropriation. It took a paragraph.

NIH is an output agency more than most, not less. Its outputs are more numerous, more auditable, and more precisely countable than nearly anything else in government. Wilson put it in the "don't try" bin, and thirty years of people who never opened a budget document have been citing him as permission.

 

Let me show you this because Wilson is not the only one saying this and most look at NIH as the premier example of the agency that has no outcomes – and that would only be said by intelligent people who do not have the particular “show me your budget” skill set. So, let me show you, how and why NIH is not the most difficult case to write good outcome measures for – if you have the right skill set it is the easiest.

 

On NIH

The ways and means: intramural research on campus; extramural competitive awards to universities and hospitals; the clinical center; research infrastructure; training; global research. Every one has receipts. Every one exhausts to the appropriation. None of them requires a single scientist to declare an outcome.

But the this is bigger than the taxonomy:

Mission failure is the product. Not a risk to be managed, not a variance to be explained — the deliverable. A research agency that never falsifies its own hypotheses is not doing research; it's doing marketing. DARPA is the same.

The outcome of research is knowledge, and a falsified hypothesis is knowledge.

If NIH funds 80 percent dead end research that is a fine thing.

Eighty percent dead ends is not a failure rate. It is the yield, and it is the yield the taxpayer is buying. Every one of those dead ends is a road no one has to drive down again — a false assumption removed from the field, permanently, for everyone, forever. That is a public good with essentially no market value, which is exactly why no company produces it and exactly why a public agency must.

And it's measurable. It's not soft. It's countable:

  • Hypotheses tested, resolved, closed.

  • Falsified: N. Warranted further investigation: M.

  • Cost per hypothesis resolved.

  • Replication rate. Publication rate. Downstream citation into successor work.

Every one of those has receipts. Every one costs out against the ways and means. And the ratio is reportable with pride: we tested a thousand assumptions, we killed eight hundred of them, and that is the product. Congress funded the elimination of eight hundred wrong roads and the identification of two hundred worth walking.

Now look at what that fixes at once.

It destroys the incentive to propose only safe science, because the safe portfolio has a bad number. A 95 percent confirmation rate is now evidence you weren't doing research — you were doing development, and a company would have paid for it. The falsification rate becomes the signature of an agency doing the thing only a public agency can do.

It gives Ruth an honest line. Nineteen years. She did not "advance frontiers." She closed questions. Some of them closed as no. Forty papers, cited, cited again, four steps downstream a mechanism became a target became a drug. The nos were load-bearing.

And it explains the cures without lying about attribution. You don't claim the sickle cell cure as an outcome target — nobody proposed it. You claim what actually happened: we bought a thirty-year search of a space nobody else would search, we retired most of it as wrong, and out of the fraction that survived came the first CRISPR cure in human history, from basic research into how bacteria defend themselves against viruses. No one could have priced that. That's the argument for the appropriation.

So the two-layer structure isn't "costable products below, aspirations above." It's:

Ways and means, fully costed, every dollar landed once — intramural, extramural awards, clinical center, infrastructure, training, logistics.

Outcomes, honestly stated — hypotheses resolved, and the proportion resolved as no, reported as the achievement it is.

And then the record — the mortality curves, the survival rates, the cure — as what a country gets when it funds a search with a high and honorable failure rate for thirty years.

That is a performance report. It foots. It's auditable. It's true. And it says the thing Daniel went to Washington to say, which is that the most valuable thing this government buys is the systematic, expensive, public elimination of being wrong — and that no one else on earth will buy it.

 

 

 

 

 

Moynihan — the performance-information paradox

The claim. Agencies produce enormous quantities of performance data and almost none of it gets used in actual decisions. Production scales; use does not. Therefore the demand for performance information is symbolic — about legitimacy and compliance — rather than instrumental. The problem is in the incentives and the culture of use.

The refutation. This is the sharpest error of the four, because it is an inference that only holds if you never look at the data.

The performance data is not unused because managers lack incentives to use it. It is unused because it is garbage, and you can tell in five minutes. It is word salad. It is a scientist's favorite activity with half his program's money omitted. It is the same acre of restored meadow claimed at one hundred percent by the bee man and one hundred percent by the bird man. It is an airport worker in Atlanta charging his hours to satellite navigation research because the true code does not exist and the coding is an interruption, not the job. It is a rent line hidden under research and a twelve percent that came out of nowhere on a Sunday in October.

You cannot conclude anything about the consumers from the non-consumption of a product that is alphabet soup other than that the consumer is smart enough to recognize the emperor has no close and turn their heads.

Moynihan measured the uptake and drew a conclusion about the appetite. The correct finding was upstream, in the construction, and he did not go there. Nobody went there — because going there requires reading a budget, and the people writing about performance measurement are not budget people. That is the entire reason this literature is wrong.

If the information were good, it would be used. He cannot demonstrate otherwise, because the experiment has never been run.

Radin — Challenging the Performance Movement (2006)

The claim. The performance movement imports assumptions that do not hold across government: that goals can be specified in advance; that outcomes are attributable to agency action; that one set of measures can serve managers, Congress, and the public at once. Add federalism — where the actual work is done by states and grantees — and long-horizon agencies where results arrive decades late, and the enterprise is incoherent at the root.

The refutation. Goals can be specified. They can be specified perfectly well.

Administer a funding program to states that has the best chance of succeeding at the outcomes we can realistically affect. There. Specifiable, costable, honest about the limits of attribution, and it names a thing the agency actually buys.

But notice what had to happen for that sentence to exist: it was written from the money up. It began with what the agency spends against and worked outward to what that spending is for. It did not begin with a feeling about the frontier and then hunt for costs to hang on it.

So Radin is not describing an impossibility. She is describing what happens when you assign the writing of a financial category structure to people who have never seen a ledger and were never meant to. The assumptions fail because the pen is in the wrong hand. Put it in the right hand and they hold.

Muller — The Tyranny of Metrics (2018)

The claim. Metrics displace judgment and induce gaming, across medicine, education, policing, the military. Measure what is measurable and you will corrupt what matters.

The refutation. Correct about bad metrics, and irrelevant to the ledger.

The rent is eleven million dollars. There is no judgment to displace and nothing to game. Measuring it does not corrupt anything. The extramural award is a real award with a real dollar amount to a real institution. You cannot game an invoice.

Knowledge is always power.

What Muller is actually describing — accurately, and this is the useful part — is what happens when you take a fictional category structure and then force two million people to certify their quarter-hours into it. That is a tyranny of metrics. But the tyranny came from the fiction, not from the counting. The cure for a bad measure is a true one, not the abandonment of measurement.

And note the irony at the center of it: Muller is the strongest possible argument for this reform, and he is universally cited against it.

The public was never told the truth about what government does. It was told an adverb.

"The agencies will never accept structures handed to them."

They already have. The budget comes in object classes funded against divisions and charts of accounts – measures will be authored matching the existing structure by people who know how to do it, It is that simple.

So: the structures are built, published, reviewed by the agencies and by experts and by Congress — and then handed down. You are funded against these lines. You will report against these lines.

And one boundary, in writing, before the first agency ever sees the team: they define the cost partition. They never define the science. No practitioner tells a laureate what to study. That line is absolute — because the day it blurs, the first hearing becomes a consultant versus a Nobel laureate, and the reform dies in a single exchange.

"You are taking something away from the scientists."

You are giving them something back.

They were handed a financial assignment they never asked for, were never equipped for, and have been failing at under audit for thirty years — while being made to certify quarter-hours into categories that describe nothing they ever did. That is not honor. That is a humiliation administered to the most conscientious people in the building, every two weeks, for as long as they stay.

And here is the cruelty of it: the more integrity a person has, the worse it is. The one who doesn't care picks a code in four seconds and forgets it. The one who does care — who wants the record to be true — is handed no honest option and made to enter a falsehood under her own name. So she leaves. And the system records it as ordinary attrition, never knowing what it just lost.

Take it off them.

Then let them do the thing they actually came to do. Once the costable structure exists, the agency uses it as a launch pad — and tells Congress, in narrative, what it hopes to achieve over the next five and twenty years.

The frontier. The moonshot. The thirty-year search. Let them write it. Nobody ever wanted to stop them. What was wrong was making the aspiration double as the ledger, which broke both of them at once.

Cost the ways and means honestly. Then tell the truth about the ambition. Two documents. Both true. Neither pretending to be the other.

"It cannot be this simple."

It is not simple. It is hard — it requires a specific and unusual discipline, and it will take real work by real practitioners across hundreds of reporting entities.

But it is not complicated. And it is not unprecedented. And it is not unsolved.

We already know how to make an enterprise legible. We do it to every public company in America, under penalty, every single year.

We took the reporting requirement and discarded every mechanism that makes the reporting true — the officer who signs, the auditor who tests truth rather than arithmetic, the consequence.

An OMB examiner once told a budget officer that his schedule had to foot. He told her the numbers would not be true. She said: and I'm telling you the columns have to sum to the appropriation.

No auditor in America would accept footing as evidence of anything. Footing is the trivial test. And that is the only test the federal government has ever applied to a performance report — a test that two offsetting fabrications satisfy exactly as well as the truth.

Show me your budget and I will tell you who you are.

There are people who can actually do that. Let them.

 

 

Previous
Previous

They Did Not Shrink the Government

Next
Next

The Steinway and the meat inspector