Substitute teacher cover

Why must a school pay to keep teachers it hopes never to need?

A school has a fixed budget and a long list of things it cannot afford. Somewhere in that budget is a line for cover: a pool of substitutes retained, or a set of teachers whose timetables are deliberately left with gaps, or a supply agency on standby. On a good week, much of that money buys nothing. The best possible outcome for the cover line is that nobody uses it.

Every other line in the budget is judged by what it produces. This one is judged by a service that, when everything goes well, is never rendered. That looks like waste dressed up as prudence — and a sharp administrator ought to be able to cut it and be right most of the time. So why is cutting it a mistake, and what exactly is the money buying?

Begin with the shape of the demand rather than its size, because that is where the answer hides. Teacher absence is not a steady drain. On a typical day in a large system somewhere around five per cent of teachers are out — illness, family emergency, training, jury service — but that average is assembled from days when almost nobody is absent and days in February when a virus takes eleven at once. The distribution has a tail, and the tail is not rare enough to ignore.

Now ask what the school could do if it planned to the average and a bad day arrived anyway. This is the crucial question, and it is the one that separates a school from an ordinary business. A factory with too many orders lets the queue lengthen. A restaurant makes you wait. Both are buffering demand in time: the work still gets done, just later. A school cannot do this. Period three on Tuesday happens at period three on Tuesday, with thirty children who are physically present and legally required to be supervised. There is no queue. The demand cannot be deferred by even one hour.

That single fact forces the whole design. If demand cannot be moved in time, and it fluctuates, the only remaining place to absorb the fluctuation is capacity — and capacity in this system is people, who must be recruited, cleared, and available before anyone knows they are needed. You cannot hire a substitute at eight in the morning for a call that came at seven forty. So the resource has to exist in advance of the demand it serves, which means it must sometimes exist without any demand at all. The idle time is not a symptom of bad forecasting. It is the physical form the guarantee takes.

Notice what follows about how much to hold. Because you are covering a tail and not an average, the pool has to be sized against the bad days, which means that on ordinary days it is by construction larger than needed. And the returns are steeply diminishing in an instructive way: the first substitute converts the commonest disruption into a normal lesson and is worth a great deal; the sixth is there for a handful of days a year and looks indefensible on any day it is not used. Every argument for cutting cover is therefore correct about the marginal unit and wrong about the distribution it was bought for.

There is one more move, and it is the one that decides real budgets. Cutting cover does not remove the cost; it relocates it. When no substitute is available, the class is split across other rooms, or a colleague loses their planning period, or a teaching assistant supervises without teaching, or the head of department covers instead of running the department. Each of those is a real cost, borne quietly by staff and by the lesson that did not happen, and none of them appears on the line that was cut. That is precisely why the cover line is so easy to cut and so expensive to have cut — the saving is visible and the loss is not.

Two honesties. The five per cent figure varies widely by country, sector and year, and the fraction of absences arising from planned training rather than illness varies more still. And the pool is not the only form redundancy can take: shared cover across a group of schools, a floating teacher on the payroll, and an agency contract are different ways of holding the same slack, with different costs and different reliability on the exact day everyone is ill at once.

The refinement that ties the steps together is that redundancy is what you buy when you cannot buy delay. Systems that can queue their work absorb variability with time and need very little spare capacity. Systems whose demand is fixed to a clock — a school timetable, an operating theatre list, a power grid at seven in the evening — have nowhere to put variability except into standing capacity that is idle whenever the variation is small.

The misconception worth correcting is that the idle capacity is the inefficiency. It is the product. A school that never had an unused substitute-hour would be a school that had left classes uncovered on its bad days, and the correct reading of a lightly-used pool is not that it was too big but that the year was kind. Judging it by its utilisation rate applies the wrong measure entirely — the right measure is the fraction of periods that went uncovered, and that number is only ever visible when the pool is too small.

One boundary. This does not mean any amount of slack is justified. The size of the pool is a real judgement about how far into the tail to insure, and holding cover for a once-in-five-years absence rate is genuinely wasteful. The argument establishes that the optimum is not zero and that the idle time is inherent; it does not tell you where the optimum sits.

• How hospitals size on-call rotas against the same undeferrable-demand problem.
• Whether pooling cover across several schools buys the same reliability for less slack.
• What the true cost of an uncovered period is, and why nobody measures it.

Every term the collection defines is gathered in the glossary.