Marginal Cost Estimation

The process of estimating the additional cost of producing one more unit of a good or service at the current level of output.

Source: Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes. 4th ed. Oxford University Press; 2015.

The approach traces which resources would actually be consumed by one additional unit of activity, rather than dividing a total by a volume. That means identifying the consumables used, any additional staff time incurred, and whether the unit would cross a capacity threshold requiring a further block of resource. Where it would not, the marginal cost is close to the variable cost alone; where it would, the whole additional block enters the estimate and the figure jumps. Direct observation of what changes when volume moves is the most reliable method, which is why marginal costs are better estimated from a service that has actually expanded or contracted than from an accounting model.

Source: Drummond et al. 2015

Because almost every decision concerns a change in activity rather than the whole service, and the average contains fixed costs that will not move with that change. A saving projected from average cost overstates what reducing activity releases, frequently by a large factor, and a cost projected from average cost overstates what additional activity requires. Decisions taken on averages therefore misjudge both directions, and the error is systematic rather than random. The error is also asymmetric in its consequences, since an overstated saving fails to materialise and is discovered late, while an overstated cost may prevent a worthwhile expansion that is never attempted.

Source: Horngren, Datar & Rajan 2015

Capacity arrives in indivisible blocks, so marginal cost is not constant but jumps at thresholds, which means a single figure describes only a band of activity. Case mix varies, so the marginal patient is not the average patient and consumes a different bundle. Much staff time is not recorded against individual patients, so the additional effort an extra case generates has to be estimated rather than observed. Each of these makes the quantity harder to establish than the concept suggests. Bulk purchasing adds a further complication, since unit prices for consumables fall as volume rises, so even the genuinely variable element is not constant across the range.

Source: Drummond et al. 2015

Over a short period almost nothing beyond consumables varies, so marginal cost is low and reducing activity releases very little. Over a longer period staffing, contracts and eventually facilities become variable, so marginal cost rises towards the average and reductions release more. A marginal cost estimate is therefore incomplete without stating the period over which it holds, and the same decision can be affordable in the long run and not in the short. This is why a saving that appears unachievable within a financial year may be entirely achievable over three, and why short-run and long-run analyses of the same proposal can reach opposite conclusions without either being wrong.

Source: Horngren, Datar & Rajan 2015

The activity range over which the estimate applies and the level at which the next capacity threshold falls, since the figure does not hold beyond it. The components included, distinguishing consumables from staff time and from any capacity element. The time horizon assumed. And whether the estimate describes an increase or a decrease in activity, since the two are frequently asymmetric, with capacity added on the way up and not removed on the way down. Reporting the method used to derive the estimate matters as much as the figure, since an estimate built from observed change carries very different weight from one derived by classifying account lines.

Source: Drummond et al. 2015