The Half-Empty Centre

AThe forecast that cities would empty is older than the technology now said to be emptying them. In 1973 an American engineer costed the fuel a firm would save if its clerical staff stayed at home two days a week, and concluded that the office district was an accident of what the telephone could not yet do; the paper was reprinted for thirty years and its central prediction did not arrive. What changed after 2020 was not the argument but the number of people running the experiment at the same moment. Within weeks the share of paid working days performed at home across the wealthier economies rose from roughly 5 per cent to something above 50, and settled, three years later, near 27. Whether that settlement has hollowed out the great commercial centres, or merely shifted their custom a few kilometres outward, is the question the surveys of the past five years were built to answer.

BPriya Ramanathan, an urban economist at the Calvert Institute, has spent those years counting transactions rather than opinions. Between 2019 and 2024 her group tracked anonymised card payments and mobile-device counts across 14 metropolitan regions on three continents. In the central business districts, weekday footfall fell by 38 per cent, a figure that has been quoted in almost every newspaper account of the subject. What those accounts leave out is the second column of her table. Across the residential districts lying between four and fifteen kilometres from those same centres, weekday spending rose by 21 per cent over the same period, and the two movements very nearly cancel, since the total sum spent inside each metropolitan boundary changed by under 3 per cent. The city, on her reading, did not shrink. Its custom moved outward by a few stops on the tram, and the loss recorded downtown is the visible half of a transfer.

CThat account makes a prediction, and Ramanathan has tested it. If the mechanism is the abolished commute, the redistribution should appear only where a large share of local jobs can be done at a kitchen table, and should be absent where they cannot. In the eight regions of her sample in which desk-based occupations exceeded 45 per cent of employment, central footfall fell by between 31 and 44 per cent. In the six where such work accounted for under 30 per cent, the fall was 4 per cent, within the error of her method. The asymmetry is what a mechanism looks like when it is real; but she is careful to describe the finding as a boundary on one process rather than a forecast for any particular city, since a place may lose its centre for reasons that have nothing to do with where its residents type.

DAnselm Prieto, a housing researcher at the Vallier Centre, rejects the inference, on the ground that a pattern measured over five years is being asked to describe a decision households take about once in a decade. His group followed change-of-address filings for 240,000 households that left a central district after 2020; 62 per cent of them settled within 40 kilometres of the address they had given up, which he reads as evidence that people bought a shorter journey to the same office rather than an escape from the city. The distances households move, he argues, are set by where an employer might one day insist they appear, and firms have not yet settled that question. He concedes two things. His filings come from a single national register with three regions well covered and the rest thinly, and the register records where a household's post is sent, which for the wealthiest movers may be neither of the two homes they use.

EThe argument has not stayed among researchers. A budget rule adopted in 2021 ties several municipal transport authorities to their fare income, so a district that loses its morning crowd loses the service that once delivered it, and a 2023 review recommended converting empty offices into flats in eleven cities without costing the plumbing that such conversions demand. Ramanathan points out, awkwardly for her own position, that nothing in her transaction data bears on whether a given office building can become housing, since the obstacle there is the depth of the floor plate rather than the level of demand. Prieto replies that the two cannot be kept apart, because councils read footfall figures as evidence about viability and vote accordingly. Both regret the same habit: the subject is argued through a handful of photogenic city centres, and almost never through counts of what happened across ordinary districts in ordinary years.

FA settlement is emerging in which the two accounts are given different work rather than ranked against each other. Over a period of months the daily geography of spending does move, and Ramanathan's columns catch it; over a period of years the map of where people live is made by housing costs, schools and the length of a lease, and Prieto's filings catch that instead. What neither can yet do is say which of the two will have written the map by 2040, because the household decisions that matter have not been taken and the firms whose policies would shape them keep revising. What both reject is the assumption that produced the original forecast, namely that a technology which removes a journey must also remove the reason the journey existed.