Filling Out Forms is Corvée Labor

Plus! The Real News Business; Credit, Wealth, and Risk-Adjusted Returns; Less-Than-Truckload; Capex; Bundling and Scale; Diff Jobs

Byrne Hobart Byrne Hobart

In this issue:

audio-thumbnail
The Diff February 24th 2025
0:00
/879.516735

Filling Out Forms is Corvée Labor

Historically, taxes were collected in a few forms: some countries had the kind of economy where there was a tax base that could actually hand over a set quantity of physical coins at some appointed time in order to provide revenue, but in many cases taxes were paid in either goods or labor—a share of the crops, some unpaid labor building a road, etc. The usual term for taxes collected in the form of work is corvée labor, and it's a sufficiently old and persistent practice that the term comes to English by way of French translations of a Roman legal term. Over time, as economies have gotten more market-oriented and the scope of government activity has broadened, we've switched to a system where taxes are collected entirely in currency.

Or did we?

That model certainly dominates discussions of government policy, outside of social issues: how much money gets raised, and from whom? How much gets spent, and on what (or, realistically, on whom)? And what's the optimal gap between the two? The Department of Government Efficiency's tracking tool uses monetary value—efficiency being the extent to which the government spends finite tax dollars on useful things.

But, especially when it comes to government spending, it's often helpful to avoid thinking about dollars and to think about real resources instead. Healthcare spending is not just dollars-in and then dollars-out: providing healthcare means paying for the time of people with specific skills, some of which are valuable and scarce. It requires buildings, capital equipment, and consumables. And doing all of this shapes incentives: in the very long run, the impact of government healthcare policy shows up in people's decisions to go into medicine, pharma research, etc., and for companies in that space to alter their R&D spending. This clarifies issues that would otherwise be hard to talk about: no money is directly lost by anyone when student loans are zeroed out. The cost to society is the incentive created to pursue or provide education in ways that are legible to sources of student loans.[1]

So, one large line item that doesn't show up on the Federal budget, but which is a meaningful cost, is the time spent complying with regulations. Conveniently, the US government actually provides an estimate for this: filling out paperwork takes roughly 11.9 billion hours per year (that page is pretty cryptic, but this post, the source of the link, has more). The average US hourly wage is $36/hour, so assuming the time burden is fairly equal, that implies that government spending is underestimated by $391bn (the good news, if you're worried about the deficit, is that this spending is 100% financed by a time-denominated tax.

But that's probably a low estimate of the cost. There's likely to be a roughly U-shaped relationship between income and the paperwork burden: if you're very well-off, you probably have interestingly complicated taxes, and there's a disproportionate chance that you work in a highly-regulated industry (the most common job in the top 1% isn't management, or law; it's medicine). Even if the average very wealthy person isn't preparing their own 1040, the person they're paying to do this for them probably earns well above the average income, too.

At the other end of the scale, if you're impoverished enough to qualify for welfare programs, you'll be doing a fair amount of paperwork to get it—ChatGPT estimates that the form used to apply for food stamps in New York has almost 400 fields, though that includes the page where people signing up for these benefits can register to vote. There is probably a positive correlation between qualifying for such benefits and being daunted by this form. Since the audience of a newsletter about finance and technology is not a representative sample from which to make judgments about the average person's relationship to the written word, it can be helpful to calibrate by looking at things like these NAEP sample questions. Read the passage, take the test, and look at the distribution of answers—students who are close to finishing high school get a bit more than two thirds of these right, and an application form for government benefits is basically a series of reading comprehension questions.

The 10.9bn-hour figure is limited in other ways: it's a measure of time actively spent on paperwork, but that's a subset of time spent on or because of regulation. Like the difference between CPU time and wall-clock time, there's a gap between the theoretical burden and the practical burden of rules. If you're waiting—for an updated passport, a green card, approval for power transmission lines, whatever—you're probably doing something with your time, but it's not what you'd be doing if you'd gotten approved. The cost of this is hard to calculate, especially since one big cost is the set of worthwhile projects that aren't undertaken at all because time and approval costs are so uncertain. There is also a very real cost to context switching. In some fields like software engineering, going to find the correct bit in developer docs or traversing Stack Overflow to find the solution to a specific problem has been somewhat obviated by tools like Cursor, but this break in focused flow, and the resulting time burden, largely still exists in many other fields (arguably with much larger consequences. Police officers for example spend 3 hours or more per shift filling out forms and paperwork. It's a bit like airport security, where the real cost is not the time you spend in line, but the sleep you forgo to get to the airport on time and then the time you spend at the gate reading in uncomfortable seats while periodically being interrupted by announcements that mostly don't concern you.

It's important to note that these costs are not pure dead weight. They're expenditures, like other government expenditures. If there's a mandate to sort your recycling, the question of whether or not that's a good mandate comes down to the cost of everyone's time and the benefit of sorting the recycling. So this time spending isn't a pure cost. But if the government can spend either your time or your money, and the monetary component is the one that gets highlighted in policy debates, the incentive is to underspend money and overspend time. But since the money mostly comes from taxes which are the result of people trading their time for money, what this really amounts to is an incentive to use taxpayers' time inefficiently.

It's somewhat unfair to contrast this with the private sector, for a few reasons. First, private companies have ruthless selection against unnecessary process, at least outside of monopolies. In some cases, a form is something that stands between a willing customer and a willing seller, so every superfluous field has a real cost, and cleverly autocompleting addresses or providing useful error messages pays a direct, measurable dividend. So even if the median private sector participant is exactly as red tape-happy as the average bureaucrat, the ones you'll actually encounter aren't, because "average" isn't good enough to stay in business. Second, governments specialize in the kinds of things that are hard to privatize because it's hard to measure their outputs. If you're putting together an e-commerce checkout flow, you can estimate whether some extra step reduces fraud enough to offset the lost sales from customers dropping off after that step. But what's the tradeoff between someone getting food stamps if they shouldn't versus someone else not getting them when they should. (This example also illustrates the ambiguous politics of efficiency: conservatives talk a lot more about overregulation, but a streamlined process of applying for benefits leads to a net increase in the availability of those benefits. The Diff has long opposed the model of achieving political consensus by letting one side attempt to do something the other side thinks they'll fail at ($), because social and technological changes alter the feasibility of implementing various policies, making them, in a political sense, a random number generator that retrospectively determines which side wins.) A final reason that the private sector has a leg up on government is that it has a different PR attack surface. In many industries, the dynamic is that smaller companies want less regulation, because they aren't vulnerable to PR blowback; they don't have a reputation to tarnish. This means that they both make the industry look bad and do things that give them a competitive leg up. Governments catch more flak for errors of commission than errors of omission, and that shapes their incentives.

This is something DOGE can actually do. Described in purely neutral, apolitical terms, DOGE is a group of young people with a tech background and a mandate to streamline government processes. This is bland enough a descriptor to apply to the US Digital Service, which was launched under Obama, and which has technically been rebranded as—yes!—DOGE. If DOGE sets spending cuts based on total dollars spent, it's going to quickly run into a problem: a big fraction of what the government does is to collect revenue and redistribute it, either directly or in the form of access to healthcare. In tech accounting terms, a large portion of the budget is GMV rather than revenue. But also, a large portion of that GMV doesn't show up on the books at all, and is immensely aggravating to everyone who interacts with it. Measure government spending in time as well as dollars, and there are cuts that make everyone richer and better-off. Even if some of them do use that free time dividend to argue about issues that are closer to the political efficient frontier.

  1. A student loan borrower is, in real terms, promising to forgo future consumption in exchange for getting access to education, as well as housing and food during the time they're getting that education. The bet both sides are implicitly making is that the incremental consumption they can afford with the skills from their education means that they're better-off, while the rest of society is afforded a more productive workforce. It's not necessarily the case that every degree needs to lead to a better job to be worth doing, of course, but to the extent that we want to subsidize degrees that don't lead to a sufficient improvement in income to offset their cost, or to subsidize degrees for people who don't understand that getting a generous interest deferral option means that the total amount they owe rises over time, then a loan is the wrong structure. They should be replaced with grants, and these grants should be debated in terms of the positive externalities they create. ↩︎

Diff Jobs

Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:

Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.

Find a Role

If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.

Find Talent with Diff Jobs

Elsewhere

The Real News Business

If you think of what the brand name New York Times connotes, you probably think about news and opinion. If there's some major global event, there's a good chance that the Times has a reporter nearby, who has existing contacts with the relevant people, and who can whip up context so Americans know what's going on. And if there's a major issue, the Times will probably weigh in, whether through the editorials it runs or through the stories it chooses to emphasize. But if you read their investor materials, the news part of the business gets less and less relevant. Their most recent quarterly presentation, for example, highlights that over the last year, news-only subscribers went from 26% to 17% of their subscriber base, while single-product subscribers who didn't sign up for news went from 26% to 30%. So a newsroom that spans the globe is still important, but less so compared to the gadgets/games/grub mix that actually motivates subscribers.

Apple is now moving in the same direction: they've had a news product for a while, and now the paid version is going to bundle more recipes and food-related content. One of the paradoxes of the news business is that the bigger the news is, the faster it's commoditized; the NYT got plaudits for publishing the Pentagon Papers, and the Washington Post had a nice scoop with Watergate. But the magnitude of these stories meant that the advantage lasted for a day or so before every other organization for which that kind of story could conceivably be someone's beat had someone reporting on it. And that same dynamic explains why news is generally tough but financial news is basically a different market: if you're reporting information that everyone else will have in the very near future, you're providing, or helping someone generate, alpha. And that's something they'll pay for.

Credit, Wealth, and Risk-Adjusted Returns

A few years ago, The Diff asked: given that risk-adjusted returns over very long periods are similar for equities and credit, why are there so many more equities fortunes ($)? This may have been a case of early-thus-wrong: Bloomberg highlights a dozen and a half billionaires produced by private credit (and meanwhile: when's the last time there was a new high-profile equities-focused manager? Twelve years after Peter Lynch first started managing money, he had a bestselling book, but it's hard to name any high-profile stock-pickers whose asset management career started in the early 2010s or later). It's an interesting illustration of where the alpha is accruing: as the PE business has institutionalized, it's become one of those secular trends that's hard to bet against, but also hard to get excess returns betting on. Whereas providing credit for those deals still creates opportunities for real upside.

Less-Than-Truckload

Amazon (disclosure: long) is possibly getting into the less-than-truckload freight industry. Less-than-truckload is basically anything that is big enough that UPS and Fedex won't ship it for you by default, but smaller than a full shipping container or truck. This makes it a dense logistics network and an excellent read on supply and demand for the long tail of physical goods is enormously valuable. The most bullish thing about this might be how long it's taken for Amazon to actually enter this space. They're a natural participant given the business they're in and the infrastructure they own, and the main reason to stay out is that they had even higher-return opportunities elsewhere.

Capex

A few weeks ago, an OpenAI-led consortium announced plans to spend $500bn on AI infrastructure in the US over the next four years (see this Diff writeup for more). Apple has apparently decided that that's a nice round number to aim for, and also plans $500bn in spending within the US, though not purely focused on AI ($, WSJ). One of the critiques of the OpenAI plan was that neither they nor their disclosed backers didn't have the financial resources to actually make this kind of investment; Apple doesn't have this problem, given that their last four years of operating cash flow come out to a little over $450bn.

This is also a good reminder that buildout races can overshoot, but they don't need many active participants to keep going. Microsoft is less enthusiastic than it used to be about actually building more datacenters (but performatively excited to lease the overbuilt ones later on). But as long as at least one company wants to be #1, and another one is trying to catch up, the race is still on.

Disclosure: Long MSFT.

Bundling and Scale

The story in streaming video over the last few years has been that many companies have a big enough content library and popular enough brand to take a shot at it, and that in most cases they don't quite reach the escape velocity necessary to compete with Netflix, unless their streaming business is bundled with something else. But a pure streaming bundle can work, if it's big enough: the combination of Disney, Hulu, and Max not only has lower churn that the individual services, but also lower churn than Netflix itself ($, WSJ). It's still possible for this to be a bad deal; the ad-free version of this bundle is $30/month, compared to $25/month for Netflix's priciest ad-free plan. And that means the streamers are getting an average of $10/month in subscription revenue, so these lower-churn customers are also lower-revenue ones. The streaming story isn't over, but it's getting closer: a sufficiently generous bundle can compete with a single scaled player, but only if everyone who contributes to the bundle is willing to accept lower revenue per user as a consolation prize.