The Reward for Good Work

In 1911 Frederick Taylor reprinted a warning he had first given engineers eight years earlier. “The natural laziness of men is serious,” he wrote, “but by far the greatest evil from which both workmen and employers are suffering is the systematic soldiering which is almost universal under all of the ordinary schemes of management.” Men on piece rates held their output down on purpose, and Taylor thought they had studied their own interests carefully. A workman who had seen “the price per piece of the work he is doing lowered two or three times as a result of his having worked harder” had learned what extra output bought him, and became, in Taylor’s words, “imbued with a grim determination to have no more cuts if soldiering can prevent it.” Taylor blamed the systems of management in common use.1

Donald Roy found the same knowledge still in use a generation later. He spent most of a year in the mid-1940s as a radial-drill operator in a machine shop, and recorded how the operators capped what they turned in well below what they could have made on the easy jobs, the ones they called gravy. Turn in too much and the rate would be cut.2 Soviet planners met the behavior from the other side. Enterprises were set targets from what they had achieved the year before, so a factory that beat its plan by a wide margin earned a harder plan, and managers learned to hide reserves, understate what their plants could do, and beat the target by a little. Economists who studied the system called the rule the ratchet. Martin Weitzman modeled it in 1980. Once this year’s result feeds next year’s target, the reward for doing better now has to be weighed against the harder target it brings, and the incentive the planners were relying on is blunted.3

Three shops, two economies, one lesson. A capacity a person demonstrates becomes a baseline someone else sets. The lesson reaches well past piece rates. It governs the nurse who covers for the software, the clerk who remembers what the process forgets, and the parent who drives across town because two offices will not share a form. The piece rate made the ratchet visible because it put a price on each turn. Without a price the ratchet turns all the same and leaves nothing in the records.

Can, will, must

The sequence has four steps, and none of them looks like a decision.

A deficiency shows up and someone covers it once. A nurse catches the dose an order set got wrong, a scheduler rebooks a follow-up the software sent to a clinic that closed, and the parent makes the drive with the form. From each of these the institution learns something true: that person can.

Nothing about the deficiency changed, so it recurs, and the person covers it again. Now the covering is a pattern, and patterns get planned around. A staffing model stops leaving time for the error, because in practice the error gets caught. The institution has moved from observing that someone can to assuming that they will.

Then the assumption hardens. Once the plan depends on the cover, withholding it breaks the plan, and breaking the plan counts as failure. The nurse who caught the error as a favor now catches it as a duty, and if she stops, the missed dose is written up under her name. Can has become must. The last step is indispensability. The slack that would have let anyone else do the work has been cut, because the work was being done, and the institution can no longer run without an effort it never entered on its books.

Each step has a local justification. The inference from can to will is made by planning, which has to assume something, and the inference from will to must is made by dependence, which accumulates on its own. By the time anyone asks whether the arrangement is fair, asking has become expensive, because the margin that would have let the institution stop depending on the answer is already spent.

Across the sequence the act never changes, and its moral character reverses. The first time the nurse stays late to catch the error, it is a gift, and she is thanked. By the end it is an obligation, and she would be blamed for skipping it. The hour is identical. What changed is the plan, and the plan was never put to her.

Research on who gets planned around matches the machinists’ experience. Christy Zhou Koval and her colleagues found that people seen as highly self-controlled are handed more work by others, who underestimate what it costs them.4 Linda Babcock and her co-authors found that women are asked more often than men to take on tasks that do nothing for promotion, and are asked more because men and women alike expect them to agree.5 The ratchet turns fastest on whoever is expected to comply, so the expectation doubles as a way of choosing whom to take from. Rank Is the Right to Remain Rigid found the same sorting from the other end, where the most reliable person in the building holds the fewest refusals.

The principle

A person’s capacity to make up for an institution’s deficiency gives the institution no claim on that capacity.6 And when a predictable, correctable deficiency is routinely covered by effort nobody agreed to supply, a smooth result is evidence about the effort and says nothing about the design.7

The objection from resilience engineering

The strongest objection comes from the field that has studied human adaptation most closely, and it lands on this archive as hard as on anyone. Resilience engineering, in the work of Erik Hollnagel, David Woods, Richard Cook and others, starts from the observation that work as done always differs from work as imagined, and that the difference is where things go right. People adjust, improvise and trade one goal against another, and those adjustments are why most shifts end without an incident.8 On this view a hospital whose nurses never adapted would be more dangerous than one whose nurses do, and calling adaptation a subsidy, as essays here have done, mistakes the source of safety for a symptom of failure.9

It is right about the world. The skill of people working in the gap is real and cannot be engineered away, and a principle that forbade institutions to rely on human adaptation would forbid them to operate.

Resilience engineering also supplies the concept that draws the line. Adaptive capacity, in its account, is finite. Woods calls the ability to stretch past the edge of normal operations graceful extensibility, and names one characteristic way adaptive systems fail as decompensation. The system keeps absorbing a growing disturbance, showing little strain, until its capacity to adapt runs out and performance falls away abruptly.10 Cook and Jens Rasmussen described the hospital version in 2005. As units run closer to full capacity the buffers between activities disappear, the system goes solid, and disturbances that slack used to absorb pass straight through.11

Read beside the ratchet, those findings rearrange the objection. The margin resilience engineering prizes is the reserve the ratchet spends. Adaptation used every day to cover a known, recurring defect is adaptation unavailable for the surprise. An institution that plans on its people’s resilience has converted its safety margin into routine operating capacity while still reporting the margin as present. Resilience taken this way is borrowed against the next emergency.

The line, then, falls between two sources of variability. Some comes from the world: the patient who presents atypically, the storm, the case no rule foresaw. Adapting to that is the work, and it is the work resilience engineering describes. Some comes from the institution’s own choices: the software nobody fixed, the staffing set below a load everyone knew about, the two offices that will not share a form. Adapting to that is covering for a decision someone else made and could unmake.

Telling them apart

Any label will be contested case by case, so the distinction needs tests. Four questions do most of the work.

The first is whether the deficiency recurs. Variability that arrives once belongs to the world. The same failure on Tuesday and again on Thursday belongs to a design that has had time to notice.

The second is whether it could be corrected, and by whom. A defect nobody could fix at any reasonable cost is a condition of the work. A defect the institution could fix, at a cost it chooses not to bear, is a transfer, and the person covering it pays the difference. The cost of the fix counts too, including what it would take from the people the institution serves.

The third is who benefits and who carries the risk. When the savings land on the institution’s ledger and the consequence of a missed catch lands on the person who missed it, or on the patient downstream, the adaptation is working for the institution at someone else’s risk.12

The fourth is whether the person could decline, this time, without being blamed for what follows. It is the sharpest of the four, because it detects the moment can became must. Adaptation that is chosen can be withheld at the cost of the outcome. Adaptation that has been taken can be withheld only at the cost of the person, whose refusal is filed as their failure.

Even adaptation to a correctable defect can be asked for legitimately. The difference is whether the institution has said so. An adaptation it names in the staffing model, treats as work, and can manage without on the day the person declines is part of a job. One it only assumes is a levy. The people carrying it also need some authority over the arrangement: a way to report the recurring defect that obliges someone to answer, a way to put the staffing itself on trial, a right to stop covering while the fix is pending. Visibility alone changes little if the people who carry the dependence cannot contest it.13 Where they can, adaptation tends to stay honest, because a defect that keeps being reported by people with standing eventually costs more to keep than to fix.

The planner’s problem

Soviet planning posed a real problem, and the principle has to answer it. A planner must learn what an enterprise can do, and demonstrated performance is the main evidence available. Forbid learning from it and you forbid planning.

The principle asks for nothing of the kind. Learning that someone can is legitimate, and often the only way to find out. Learning turns into taking at the next step, when the observation becomes a standing assumption without the person’s agreement and without payment for what the assumption consumes. The planner can learn from last year’s output and still be bound to fund next year’s.

The ratchet’s history also shows what an institution loses when it skips that step. People who learn that demonstrated capacity becomes obligation stop demonstrating it. Roy’s operators capped their output, Soviet managers kept reserves off the books, and the colleague who learns that finishing early brings more work stops mentioning when they finish. The institution loses the surplus, and the planner loses something worse. The only evidence of what its people could do was evidence it punished them for giving, so the evidence stops coming. When the same withdrawal resurfaced in 2022 as quiet quitting, much of the commentary treated it as a crisis of engagement. Roy’s operators would have recognized it as a reply to the rate.

When the withdrawal is collective, the institution finds out what it had been taking. Work-to-rule, doing exactly what the procedure says and nothing the procedure forgot, runs the ratchet backward. It withholds every adaptation the plan had come to assume, and the operation slows to the speed of its own design. Nothing is sabotaged, and the design is shown as it was built.

What the principle leaves alone

The principle constrains plans and leaves acts alone. The nurse who stays past her shift because a patient she has cared for is dying has done nothing it objects to, and neither has the engineer who works through the night of an outage or the neighbor who organizes the street after a flood. Emergencies outrun every plan. Devotion that is chosen belongs to the person who chooses it.

What the principle forbids is writing the next morning’s plan on the assumption that it will happen again. An institution may hope for its people’s surplus, and it may be grateful for it. It may not book it. A gift differs from a levy in whether the arrangement fails when it is withheld, and one that fails without the gift was running on a levy all along.

Refusing to book the surplus is also the only way to keep it. Extra effort freely given survives only where giving it does not raise the rate, which is what Taylor’s workmen knew and what his managers kept failing to learn. An institution that plans on it teaches people to stop offering it. One that leaves it off the books still has it on the night the surprise comes. The question to put to any smooth result is the one the ratchet makes expensive to ask. If this person stopped tomorrow, whose failure would it be called?