The Line

Here is the operation the four cuts share — what makes them one problem and not four. Each forces a continuous, many-dimensional reality through a discrete legal category: how economically integrated a worker is, collapsed to employee-or-not; how much of the continuum is work, collapsed to counted-or-not; what an hour is worth, collapsed to a single dollar figure; how certain we are, collapsed to proven-or-not. That operation has a name in information theory — quantization, the encoding of a continuous signal into a finite alphabet — and §7 shows it carries an unavoidable cost. The four are not quite uniform: three quantize the continuum into a decision — in or out, counted or not, proven or not — while valuation quantizes it into a measurement, a dollar figure, and that lone difference is why self-reference surfaces only at the price (Cut III), where the rate is read off a total that already contains it. First, though, the structure. The cuts compose into a pipeline, and the order is not incidental. Coverage is a gate: a covered employee, of a covered enterprise, outside any exemption — a switch that, if open, zeroes everything after it. Compensability defines the domain: which slices of the continuum enter the count as hours. Valuation prices each unit: the regular rate. Proof allocates the residual doubt about all three. Schematically, exposure is a product — coverage, times the compensable hours, times the rate, scaled by the penalties — and because it is a product, error does not add. It multiplies.

This is the structural fact the technicalities obscure. A twenty-percent dispute about hours and a fifteen-percent dispute about the rate do not make a thirty-five-percent quarrel; multiplied, they open a band of exposure roughly twice as wide at the top as at the bottom. The gate, being a zero-or-one, can extinguish a vast sum or unlock it on a single ruling. And the proof cut decides who stands inside that compounded band. Run the pipeline below: set the disputes, open or close the gate, and watch a modest disagreement at each stage become an immodest range of dollars — the range within which every case is actually settled.

In §2 each cut became a quantizer of the continuum; here is what quantizing costs, and why no cut can pay nothing. Treat a legal rule as an estimator — a procedure that takes the facts of a case and returns the outcome the law is after. Statistical learning theory decomposes any estimator's expected error into three parts, and they are exactly our three corners. Approximation error — how far the best rule of its kind can sit from the truth — is infidelity, and it is large for a coarse rule that cannot bend to the cases. Estimation error — how much the rule's output swings with the particular facts a tribunal happens to find — is indeterminacy, and it is large for a rich rule read from sparse evidence. And complexity — how much the rule must say to specify itself — is the cost of administering it. The corners are not three analogies bolted together; they are the three terms of one decomposition.

What forbids having all three small at once is a bound, not a mood. To shrink approximation error you must enrich the rule — move it toward a standard, give it more degrees of freedom to fit the cases. But a richer rule has more ways to go wrong on limited evidence, so its estimation error grows; structural-risk theory makes the growth precise — estimation error rises with complexity and falls with the quantity of reliable fact. So fidelity demands complexity, and complexity, unless paid for in evidence, buys indeterminacy. The only escape is to spend the third resource outright: a detailed code drives both errors down by enumerating the answer for every case — effectively supplying its own evidence — and pays for it in sheer complexity, the administrability cost. Determinate, faithful, administrable: hold any two, surrender the third. That is the trilemma, and it is now an identity with a bound behind it, not a slogan.

Two quantities therefore govern where the optimal cut falls, and the second is the one the triangle hides. The first is the heterogeneity of the domain — how varied the cases are — and a varied domain punishes a rule's bias, pushing the optimum toward a standard. The second is the evidentiary budget — how much reliable fact the system can actually ascertain and process — and this is nothing other than what the proof cut of §6 governs. When facts are cheap and certain, a standard's estimation error is tamed and the law can afford fidelity; when proof is dear, a rule's determinacy is worth more than a standard the tribunal cannot reliably apply. This is why the case for a compensability standard strengthened exactly as timekeeping technology improved: Troester reasoned that the old administrative-difficulty excuse for the de minimis rule no longer holds once employers can capture the time. The proof cut does not merely allocate residual doubt downstream — the cost of proof feeds back to set the optimal shape of every other cut. Move both sliders below, the variety of the world and the evidence the system can bear, and watch the loss-minimising cut respond.

The triangle is traversed, not occupied

The trilemma is not a one-time choice; rules and standards migrate. A bright-line rule, applied to the cases it betrays, accretes exceptions to cure the unfairness — and a rule riddled with exceptions is a standard. A standard, applied case after case, hardens into presumptions and safe harbors as courts hunt for predictability — and a standard reduced to presumptions is a rule. The field drifts, perpetually, between the corners; reform is the deliberate relocation of a doctrine across the triangle. Watch one such move below: California's AB5 took worker classification from the open Borello standard and dragged it toward the categorical ABC rule — buying determinacy and administrability, paying in fidelity, on a bet that the domain was uniform enough to bear it. The doctrines scatter across the space because the space has no stable centre to settle in.

A cut is drawn against a party who will rearrange the facts to its advantage, which makes the setting of any line a mechanism-design problem: the regulator must choose the quantizer knowing it will be optimised against. The sharpest consequence falls on bright lines. Wherever crossing the line yields a discrete benefit — a salary at the threshold buys the exemption, an hour under forty avoids the premium, a contract just clearing prong B escapes employee status — the regulated party crowds the favourable edge. The employer does not pay what the job is worth or schedule the hours the work needs; it pays exactly the line and schedules exactly to it. Economists call the pile-up bunching, and it is measurable: a spike sits in the salary distribution precisely at the exemption threshold, with a hole carved just beneath. Raise the stakes below and watch the spike grow — and switch to a standard to watch it vanish.

Two consequences fall out. First, a rule's true bias exceeds its boundary error: beyond the cases it misclassifies, it distorts the cases it does not, paying a crowd of workers exactly the line rather than their worth — and a standard, having no line to crowd, escapes this particular cost even as it incurs the other. Second, the line is where the law is made and unmade. The cases that litigate are the close ones — clear cases settle, as Priest and Klein observed — so the disputes that reach courts cluster at the very boundary the cut cannot get right, and the doctrine is forever built, and forever destabilised, at its own edge. The regulator's question is therefore not "where is the correct line" but "where is the line that survives the response it provokes" — which is why reform is less a discovery than a move in a game that does not end.

Assemble the architecture. Wage-and-hour law is the mandatory completion of a contract that is constitutively incomplete, executed as a pipeline of four quantizers — gate, domain, price, residue — each paying the same rate–distortion price, with error compounding down the line and the proof cut charging the residue to whoever could have recorded it and did not. That is the structure. Two things follow, one strategic and one distributive.

The strategic one makes the pipeline operational. A case's settlement value is, to a first approximation, the expected judgment — the probability of liability times the damages if liable — adjusted for each side's litigation cost and appetite for risk. Lay that over the cuts and every term has an address. The probability is fixed almost entirely at the binary cuts, the gate and the categorical compensability question, which a case wins or loses outright. The expected damages are the midpoint of the compounded band: hours times rate times the penalties. The risk the parties price is the band's width. And the proof cut moves two terms at once — missing records both raise the chance of establishing the violation and slide the conditional damages toward the top of the band — which is why the burden-shift is worth more than it appears, and why the presence or absence of timekeeping, more than any single legal question, sets where a case settles. Willfulness then scales the whole figure twice, doubling the award and adding the third year, a combined multiplier near three on everything beneath it. The map from doctrine to dollars is not loose; it is this product, and a litigator who knows which cut moves which factor knows where the money is.

The distributive one is the deeper, and it begins by noticing what the optimisation in §7 quietly assumed. To speak of an optimal cut — the one that minimises expected loss — is to presuppose a loss function: a settled account of the target the rule should track, against which approximation error is even defined. But the target of wage-and-hour law is not settled. Is it to pay workers for time actually worked, to shift surplus toward labour, to halt a race to the bottom, or to deter employer opportunism? These aims do not coincide, and they price the same error differently — so the loss-minimising cut is optimal only relative to a chosen objective, and choosing the objective is itself the distributive act. The machinery of §7 is real, but it runs only after someone has fixed the loss function, and fixing the loss function is not a technical step; it is the whole political question, entered as a premise. With that in view, the location of each cut is plainly not neutral. A determinate bright line favours neither labour nor capital as such; it favours whoever benefits from where it is drawn — a daily-overtime rule transfers to workers, a high salary threshold that exempts more of them transfers away. An open standard does not split the difference; it relocates the contest to adjudication, advantaging the repeat player and the side that can bear delay and litigation risk. A detailed code advantages those who can afford the experts to read it. There is no neutral place to cut a continuum, because every cut both prices labour and allocates the risk of the pricing — and those are the very stakes the parties were fighting over.

This is why the lines never hold still. The independent-contractor rule, the salary threshold, the conditions on the fluctuating workweek — each flips with the administration, not because the law is unsettled in some remediable way, but because each is a contested position on a trilemma whose every point allocates between capital and labor. The technicalities you began with — the fractions, the exemptions, the counted and uncounted minutes — are the surface of a distributive struggle conducted in the grammar of line-drawing. The companion essay showed that self-reference forces every system rich enough to describe itself to surrender one of universality, negation, and consistency. This one closes on its sibling: the wage relation forces every legal order that would price labor to surrender one of determinacy, fidelity, and administrability — and that surrender is never abstract, because it always names a party to bear it.