“I’ve got a great idea…”
There’s an old joke about founders: “Big-picture thinker seeks detail-oriented doer to handle implementation.” That is, ideas overpopulate, execution a rare species.
Effective implementation, be it software or policy, is determinative of success. The “doers” have to solve actual problems - technical limitations, resource constraints, maybe marketing, and if it involves a sufficient number of people, inevitably boring organizational management (read: internal politics).
Ever true in government, this is what torpedoes even amazing ideas—failure to navigate the incentives, which we’ll return to shortly.
Data, Data Everywhere, Not a Bit to Sync
Consider data: the federal government relies on gobs and gobs of it from every corner of the country, from individual tax returns to niche corporate reporting in small industries. This data is both a necessary input to government functioning, hugely valuable output for the private sector, and every permutation thereof.
And it’s damn costly.
First off, it’s inefficient. Much of it is duplicative. Overlapping requirements mean the same data is reported to multiple nodes of the Fed. For example, a company hit with a major data breach may have to submit the same information to DHS, SEC, and DOD. Add in state authorities too, each having justifiable reasons to need it.
Pity this poor bank CIO amidst a Code Red hack, getting yelled at by the CEO, trying to make sure they don’t have to reverse a bunch of credit card charges, now also has to not get crosswise with multiple federal agencies. And it’s not like it’s the same form you can just photocopy. Each has a bespoke format and deadlines and so on.
The craziest part is all of this [gestures wildly] is advocated by literally no one.
No less true in more mundane contexts. Remember that old IRS meme?
IRS: You owe us money. It’s called taxes.
Me: How much do I owe?
IRS: You have to figure that out.
Me: Can I just pay what I want?
IRS: No, we know exactly how much you owe, but you have to guess the number too.
Me: What if I guess wrong?
IRS: Jail.
Inconsistent data fields, duplicative reporting, and guess how all these stores of information get along. Completely seamlessly and without friction? No, of course not!
When our neighborhood throws the annual block party, we use a signup sheet to ensure every item is covered and we don’t double up. One does not simply bring pasta salad without checking first—what are we animals?
Okay it’s more complicated than that. Really, to stretch this metaphor to the point of distension, it’s more akin to every house throwing its own block party, at different times, with different themes and rules and different menus but some overlap.
Every office in every agency has a statute (ideally, ignoring for now) that tells it what to do - or more often tells it to figure out what to do - and for the most part didn’t bother to check what laws and regulations applied elsewhere.
At some point a Congressional committee was ginned up about something and determined to handle it. Say, shark fin sales (H.R. 81 - Shark Conservation Act of 2009, passed as Pub. L. 111-348, codified at 16 USC 1801). Of course they don’t actually descend into the waters and monitor all the cartilaginous fish appendages. They tell NOAA to look into it (50 CFR 600). And if asked, and NOAA can’t provide the data showing progress, Congress might not move on just because the implication is gathering this data from private parties is costly.
Now multiply this over hundreds of times over multiple decades, each one deemed important and each one requiring data specific for that purpose.
The point is the solution is obvious: consolidate all the data across agencies, and make it available to them. So why hasn’t it been tried?
Oh it’s been tried.
There Oughta Be a Law
Congress even legislated it. The OPEN Government Data (2018) is one such example. There’s been some progress but it turned out implementation is more difficult than expected. It comes down a few roadblocks:
Everyone agrees the aims are laudable. No one objects. But things like this are simply a low priority, and frankly somewhat boring. From the senior political level, it’s almost never going to be top of their to-do list. Which is to say, they’ll task staff with moving the ball, but not at the expense of other goals. Inevitably this project will come at the expense of something else. At the career staff level, it’s rare this will be the most exciting use of their time.
Data fields a normie would identify as identical, at a granular level exist in slightly different forms across the government. Small differences in format, reporting method, estimation methodologies, even coding language make for large frictions to concatenation. The individual agency incentive is to insist: “you should adopt our format.”
It turns out privacy is a big deal in this country. Any information applicable to an individual or organization has to be considered in the context of how it was provided and what can be imputed when it’s combined with other information.1 The longstanding cultural bias in government is to err on the side of caution. That’s generally a good thing, but it does obstruct otherwise commendable efforts.
I observed this firsthand, especially #3 during my time in government. An effort to share Social Security Administration info on deceased beneficiaries with Treasury’s Do Not Pay list was stymied by among other things major privacy concerns. Not just risk aversion, real legal risks. The Privacy Act of 1974 and CIPSEA in particular. While the public reporting focused on IT interoperability and formatting, which were real, the behind-the-scenes conflicts were all privacy obligations.
Cybersecurity
As you can imagine, it gets harder when you add data to regulations, but the dilemma is structurally identical. As GAO pointed out recently, even in spite of clear Congressional direction on the matter, attempts to streamline compliance in the area of cybersecurity are met with resistance. I don’t dispute GAO’s findings, that fiscal costs are a friction to implementation. But whereas GAO is not in a position to opine on organizational incentives, I submit fiscal costs are merely a proxy for the interagency incentive alignment problem. Yes legacy computer systems are a real and discrete barrier but these are not the truly binding constraint–we know this because even when Congress provides additional funding the outcome is the same.
While colleagues and I were digging into these issues, cybersecurity in particular, this was a persistent issue. Bipartisan consensus is rare, but not on this. This was not an issue of more regulation vs. deregulation. This was widespread understanding that significant resources were going to a totally unproductive use.
And not for nothing, the country’s cybersecurity was (and remains) woefully inadequate, so this was no small thing to solve.
Who Let the DOGE Out?
Whither DOGE and what’ve they got to do with it?
Well, armed with a seeming disrespect for authority and big talk about technical prowess, on this one problem it was a combination particularly fit for purpose:
Data, IT, software, y’know computer stuff…this is supposed to be what they were not only skilled at, but among the best in the world.
Afforded a, how to say this, DGAF remit.
So I was in this one case optimistic about DOGE’s likelihood to meaningfully break through the status quo ante constraints, being armed with this combination of expertise and something resembling political immunity.
It’s possible there was some success in this space. But for now we haven’t seen it.
And yet I still maintain the diagnosis of the problem is accurate, and DOGE was conceptually the best chance to solve it. So how do I reconcile that with their failure?
Simple. DOGE didn’t actually understand the problem they were trying to solve. That and/or for all their superhuman pedantic posturing, they weren’t really as well-equipped as they said.
The author is Nonresident Senior Fellow at the Foundation for American Innovation
It turns out even after “deidentification” (where elements deemed attributable to individuals or small groups are removed, formally Individually Identifiable Data), using just publicly available government data the missing information can be reconstructed in surprising and somewhat upsetting ways. https://www2.census.gov/library/working-papers/2023/adrm/ces/CES-WP-23-63.pdf. This was done well before widespread use of LLMs.