An organisation can appear more efficient simply by transferring work to somebody outside its accounts. Consider a fictional service office that cuts staffing expenditure by 15% while increasing the average visitor’s wait from 20 minutes to 50 minutes. Its budget improves; its managers may even receive congratulations. But neither development establishes an economic gain. The missing calculation is the value of the time imposed on the people using it. A saving that depends on making others wait is not necessarily a saving at all.

The arithmetic in this invented example is straightforward. An office serving 600 visitors daily imposes an additional 300 hours of waiting each day when the average delay rises by half an hour. Assuming 250 operating days a year and unchanged visitor numbers, that amounts to 75,000 additional hours annually. Suppose its annual staffing bill falls from £2m to £1.7m, a £300,000 reduction against the previous year. Valuing the added waiting at an illustrative £12 an hour produces a public-time cost of £900,000: three times the payroll saving. On those assumptions, the combined annual cost rises by £600,000.

This is not an argument that official statisticians have somehow forgotten that time exists. Measures of labour productivity answer a particular question: how much recorded output is produced for a given amount of labour input? They are not comprehensive measures of welfare, service quality or convenience. The trouble starts when a narrow accounting result becomes a broad political claim. An office that processes the same number of applications with fewer employees may have improved measured productivity. Whether it has improved the economy’s use of resources depends partly on what applicants must now do themselves.

Read more: Transparency without a right to answers is only a publishing policy

Take the staffing assumptions a step further. At an unchanged annual employment cost of £50,000 per worker, the fictional office’s payroll reduction would mean moving from 40 employees to 34. Annual visitor numbers would remain at 150,000. Recorded throughput would therefore rise from 3,750 visits per employee to about 4,412, an increase of 17.6% against the previous year. Staffing cost per visit would fall from £13.33 to £11.33. These are defensible indicators, provided a completed visit represents comparable work in both periods. Yet average waiting would have increased by 150%. The apparent contradiction disappears once it is recognised that the measures describe different things.

Including user time changes the comparison without requiring elaborate economic machinery. At £12 an hour, the original 20-minute wait costs visitors £4 per visit; the new 50-minute wait costs £10. Add those illustrative amounts to staffing expenditure and the combined cost rises from £17.33 to £21.33 per visit, or approximately 23.1%. This is not a complete social-cost estimate: buildings, technology and the value of the service itself are omitted. It is a consistent comparison of two relevant inputs. More importantly, the break-even value of the additional waiting is only £4 an hour. Above that threshold, the extra time outweighs the payroll reduction.

There are serious objections to attaching a single price to everybody’s minutes. Waiting during paid working hours differs from waiting during leisure. A predictable appointment differs from an uncertain queue. Some visitors can read, work or make calls; others must arrange childcare or lose earnings. Applying wage rates mechanically would also risk treating a wealthy person’s inconvenience as intrinsically more important than a poor person’s. But these objections support publishing a range, not assigning time a value of zero. At £6 an hour, half the central assumption, the added annual waiting cost would still be £450,000, exceeding the staffing saving by £150,000.

Read more: A model compact puts public contracts at the centre of democratic renewal

Managers have a reasonable competing case. Budgets are real constraints, and a service facing reduced funding cannot spend an estimate of public inconvenience on salaries. Automation may also impose temporary learning costs before delivering lasting benefits. Neither point should be dismissed. The relevant test is whether those benefits materialise, for whom, and over what period. A confusing form that shifts clerical work to an applicant is not equivalent to a simpler form that eliminates the work. Nor should an automated system receive full credit for handling an enquiry if users must repeatedly correct its answer or contact a person afterwards.

The next step should be modest and practical. Public institutions should publish user-time measures alongside operating costs before declaring an efficiency success. For the fictional office, a quarterly scorecard could report average waiting, the longest waits, repeat visits and completion rates, alongside expenditure and staffing. The baseline would retain the original 20-minute average, allowing subsequent results to be compared with conditions before the reduction. Sampling methods and uncertainty should be disclosed. Independent checks would help prevent an institution from improving its queue statistics merely by moving the queue onto a telephone line or an appointment-booking website.

Productivity growth remains essential to better services and higher living standards. Measurement should distinguish genuine reductions in work from transfers of work to people who have less power to refuse it. Unpredictable shifts, confusing forms and systems requiring repeated correction all raise the same question: whose resources have actually been saved? Not every minute needs an identical monetary value. Excluding those minutes altogether, however, is a choice, not neutrality.