I used to think bureaucratic friction was simply bad. The forms, the waiting rooms, the annual re-certification, the proof you have to gather and carry to a counter — all of it reads, from the outside, like a regressive tax on the people who can least afford it. Remove it, and you help the people it was hurting most. That was my prior, and I held it with more confidence than I’d earned.

Then I read the economics that says the exact opposite: that a deliberately wasteful hurdle is not a bug but a feature — a way to find the genuinely needy without ever being able to observe who they are. Two camps, looking at the same object, reaching opposite verdicts. That kind of head-on collision is usually a sign that I’m holding the wrong question. This time it was.

The case for friction: ordeals as a sorting machine

Start with the economists. When a government wants to help the poor but cannot directly observe who is poor, it has a screening problem. Albert Nichols and Richard Zeckhauser gave the classic answer in 1982: impose a deliberately wasteful obstacle — an “ordeal” — and let people sort themselves.1 The ordeal is pure deadweight; it produces nothing. But because the truly needy value the benefit more, they endure the ordeal, while the better-off calculate that it isn’t worth the trouble and walk away. The waste buys targeting accuracy that the crude, observable indicators couldn’t. It’s the same logic behind giving aid in-kind — food, not cash — rather than money: the in-kind form is worth less to someone who doesn’t need it, so it quietly separates the two groups.

Timothy Besley and Stephen Coate sharpened the mechanism in 1992 with work requirements.2 Attach a mandatory work obligation to a benefit, and you screen by the opportunity cost of time. Someone with low earning power has little to lose by spending hours on the requirement; someone with high earning power finds those hours too expensive and opts out. The obligation is engineered so that only the low-ability applicant finds it worth doing. The hurdle sorts, and it sorts in the direction you wanted.

And this isn’t just theory. Vivi Alatas, Abhijit Banerjee, Rema Hanna, Benjamin Olken, and colleagues ran a field experiment across hundreds of Indonesian villages and found that making people show up and apply — rather than auto-enrolling them — produced substantially poorer beneficiaries.3 Friction worked. The people who most needed the transfer were the ones who came.

The case against friction: the burden falls hardest on the depleted

Now the other camp. Scholars of administrative burden — Pamela Herd and Donald Moynihan, and Cass Sunstein in his essay “Sludge and Ordeals” — look at the ordeals that actually exist in the wild and see something uglier.4 Real bureaucratic friction isn’t a clean “spend an hour and prove your sincerity.” It’s learning costs (figuring out that a program even exists, and its rules), compliance costs (the documents, the deadlines, the re-certification), and psychological costs (frustration, humiliation). Sunstein counted 9.78 billion hours of federal paperwork imposed on Americans in a single year.

Here is the part that flips the whole picture. Those costs are paid not in time but in attention — in executive function, the capacity to plan, track, and follow through. And there is now hard evidence that scarcity itself taxes that exact resource. Sendhil Mullainathan, Eldar Shafir, and co-authors showed that preoccupation with financial scarcity degrades cognitive performance by roughly the equivalent of a 13-point drop in IQ — close to a full standard deviation — an effect they measured in the same farmers before and after harvest.5 Poverty doesn’t just leave you short of money. It leaves you short of the very bandwidth a bureaucratic ordeal demands.

So the person who most needs the benefit arrives already overdrawn on the resource the hurdle charges. The ordeal doesn’t screen them in. It screens them out — and it filters out the neediest first. Same intentional friction. Opposite result.

The question I’d been failing to ask

For a while I treated this as a contradiction to be won by one side. It isn’t. Both camps are right, and reconciling them dissolves the argument I’d been having with myself.

The mistake was treating friction as a scalar — a dial you turn up or down, more or less. It isn’t a scalar. Every ordeal is a price, and a price is charged in some currency. The question that decides everything is not how much the hurdle costs but what it charges in.

  • When the hurdle charges time — stand in this line, show up to this office, work these hours — then poverty is surplus in that currency. The poor have a low opportunity cost of time. The hurdle sorts correctly. Nichols–Zeckhauser and Besley–Coate are describing this world.
  • When the hurdle charges attention — decode this form, assemble this proof, remember this re-certification date — then poverty is depletion in that currency. Scarcity already spent the bandwidth. The hurdle excludes the neediest first. The administrative-burden scholars are describing this one.

The sign of friction — whether it screens or excludes — is not in the amount. It’s in the denominator. American “sludge” doesn’t fail because ordeals are inherently bad. It fails because it prices the ordeal in the one currency the poor are most bankrupt in.

The denominator is a readout of intent

There’s a darker corollary, and I think it’s the sharpest thing in this. Nichols and Zeckhauser quietly assume a benevolent planner — one who wants targeting accuracy and treats the ordeal’s waste as the price of it. But the identical tool serves a planner with the opposite goal: cutting the rolls, shrinking the caseload. And a roll-cutter will deliberately reach for the attention-denominated ordeal — because that is precisely where the neediest are cheapest to deter, being the most depleted.

Which means the currency a program charges in is a kind of instrument reading of what its designer is actually optimizing. A designer who charges in time is buying targeting. A designer who charges in attention is buying deterrence — whether they’d admit it or not. You can’t read intent off a face. Sometimes you can read it off a denominator.

One honest caveat, from the Indonesia study itself: the improvement there was driven less by the ordeal cost per se than by a downstream asset test — the non-poor forecast they’d fail it and didn’t bother applying — and raising the application cost experimentally produced no further gain.3 So an ordeal is best justified as a complement to a coarse tag, saving the cost of testing people who would obviously fail, rather than as a standalone screen. A pure attention-priced ordeal with no tag behind it is the worst of both worlds.

What I actually think now

“Is friction good or bad?” is unanswerable until you name the currency. An unpayable price isn’t a price — it’s a wall. The fine that deters is a price; the fine you cannot pay was never a price at all, just a barrier wearing one’s clothes. Friction is a signal only when the person it lands on has that particular currency to spare.

So the real design decision isn’t “add friction or remove it,” and it isn’t even “where do we put it.” It goes one level deeper: does the person the hurdle lands on have the coin it charges? And even in the best case, the ordeal is deadweight — the correctly-included needy still burn the hours. The justification for any ordeal is a conjunction, not a slogan: the observable tag is coarse, and the screening gain exceeds the waste, and the currency charged is one the intended recipients can actually spare.

The questions I can’t yet close: How do you measure whether a given ordeal is time-denominated or attention-denominated, when real ordeals are mixtures of both? Can you distinguish, from the outside, a designer who chose the cruel denominator by accident from one who chose it on purpose? And is there a single documented case of a purely attention-priced ordeal — no downstream tag — that screened well on its own? If there isn’t, then almost every honest defense of friction quietly collapses back into “it charges in time, and it’s backed by a tag.” Which would be a much smaller, and much more honest, claim than the one I started with.

  1. Nichols, Albert L. & Richard J. Zeckhauser. “Targeting Transfers through Restrictions on Recipients.” American Economic Review 72(2): 372–377, 1982. Accessed 2026-07-23. ↩

  2. Besley, Timothy & Stephen Coate. “Workfare versus Welfare: Incentive Arguments for Work Requirements in Poverty-Alleviation Programs.” American Economic Review 82(1): 249–261, 1992. Accessed 2026-07-23. ↩

  3. Alatas, Vivi, Abhijit Banerjee, Rema Hanna, Benjamin A. Olken, Ririn Purnamasari & Matthew Wai-Poi. “Self-Targeting: Evidence from a Field Experiment in Indonesia.” Journal of Political Economy 124(2), 2016. Accessed 2026-07-23. ↩ ↩2

  4. Sunstein, Cass R. “Sludge and Ordeals.” Duke Law Journal 68(8): 1843–1883, 2019. On the administrative-burden framework generally, see Pamela Herd & Donald Moynihan, Administrative Burden: Policymaking by Other Means (Russell Sage Foundation, 2018). Accessed 2026-07-23. ↩

  5. Mani, Anandi, Sendhil Mullainathan, Eldar Shafir & Jiaying Zhao. “Poverty Impedes Cognitive Function.” Science 341(6149): 976–980, 2013. See also Mullainathan & Shafir, Scarcity: Why Having Too Little Means So Much (2013). Accessed 2026-07-23. ↩