Displacement Hypothesis

What it is

Every claim that a technology is costing people something has a displacement hypothesis buried inside it. The hours have to come from somewhere. If people are spending four and a half hours a day online, as Ofcom's passively tracked panel reported for United Kingdom adults in May 2025, those hours were previously spent doing other things, and the interesting question is which things. That question sounds obvious and is almost never answered, because answering it needs a particular kind of data.

Most of what circulates on this subject is survey correlation: people are asked how much they use their phone and how lonely they feel, and the two numbers are compared. That instrument has a known fault. Douglas Parry and colleagues, reviewing 106 effect sizes in 2021, found that self-reported digital media use rarely tracks logged device data. A time-use diary is a different instrument. It asks what somebody actually did on a specified day, minute by minute, and it does not ask them to estimate anything about their habits. The American Time Use Survey has been running since 2003, which is why the best evidence on displacement comes from economists and sociologists rather than from psychologists.

Wallsten, writing alone for the National Bureau of Economic Research in 2013, built those diaries for 2003 to 2011 into a person-level dataset of about 124,000 observations and asked what an extra minute of online leisure sits alongside. The answer, from his abstract: 0.29 fewer minutes of all other leisure, with about half of that coming from television and video, 0.05 minutes from offline socialising and 0.04 from relaxing and thinking. Beyond leisure, each minute online sits with 0.27 fewer minutes working, 0.12 fewer sleeping, 0.10 fewer travelling and 0.06 fewer on education. Hobbies gain, slightly.

Three things follow, and the first is the one nobody repeats. Television and film account for 0.12 of the 0.29, so the biggest component of the trade is one mediated experience giving way to another. The displacement argument in circulation is about embodied life being replaced by screen life. The best available estimate says the main transaction is screen for screen.

The second is that face-to-face contact does move, in the direction the hypothesis predicts, at roughly a twentieth of a minute per minute. That is a real number and a small one. Wallsten's own summary of the social side is more careful than any single coefficient, pointing at socialising, parties and cultural events together: a cost of online activity, he writes, is less time spent with other people.

The third is that these are correlations and they are marginal rates. Wallsten says so in his own abstract and introduction: the paper estimates the opportunity cost of online leisure and does not identify what substitutes for what. Online leisure ran at about thirteen minutes a day at the end of his window, so multiplying his coefficients by a modern screen-time total would be arithmetic performed on a quantity the paper never measured.

Alongside the per-minute trade sits a much larger aggregate trend that nobody will attribute. Patrick Sharkey, also writing alone, took the same survey through 2022 and found that average time spent at home among American adults rose by one hour and thirty-nine minutes a day between 2003 and 2022, roughly ten per cent, for every subset of the population and virtually all activities, rising steadily before the pandemic. He then states plainly that the survey has no precise questions tracking time on smartphones or social media, argues in his own voice that the breadth of the trend complicates any internet explanation, and names fear or discomfort in public space, the rise in gun violence and larger homes among the alternatives. That combination, a large measured trend and a researcher refusing to explain it, is the honest shape of this whole question.

In effect

In research: Scott Wallsten's 124,000 diary days

The only direct measurement of the trade is a working paper almost nobody cites. Wallsten's subcategory estimates, all significant at the one per cent level unless noted, run: television and film, non-religious, minus 0.12 minutes; socialising and communicating, minus 0.054; relaxing and thinking, minus 0.037; parties, minus 0.016; attending cultural events and institutions, minus 0.010; listening to the radio, minus 0.0044. Hobbies gain 0.0036, which he notes is less trivial than it sounds, because Americans average about two minutes a day on hobbies. Religious television, smoking, writing and listening to music other than radio show no significant effect. Reading for personal interest is coded separately and does not appear among his subcategory results at all, so the paper supports no claim about reading in either direction. His framing device is Robert Fogel on railroads: the true net benefit of an innovation is what it adds beyond the value of what it displaced.

In everyday life: the hour the television lost

The household version of Wallsten's table is unglamorous. The evening that used to end with a broadcast schedule now ends with a feed, and the diary records a substitution between two sedentary indoor activities rather than between a screen and a conversation. Sharkey's nineteen-year trend is the same observation at national scale: work-related activity done at home rose two percentage points between 2003 and 2019 and remained fourteen points higher in 2022, and education-related activity at home rose twelve points before the pandemic began. People did not stop doing things. They stopped leaving to do them. Sharkey adds a further finding with his own hedge attached, that time at home is associated with lower happiness and less meaning, and says the data cannot establish whether being at home causes that.

In commerce and product design: National Theatre Live and the cinema three kilometres away

The hypothesis is also a commercial question, and the best test of it is British. Hasan Bakhshi and Andrew Whitby took about 16 million transaction records for 54 English performing arts organisations, representing 44 million tickets, from the launch of National Theatre Live in 2009 to late 2013, and coded treatment as distance from a cinema that had shown a broadcast. Theatre visits ran about ten per cent higher within a kilometre of such a cinema, seven per cent at one to two kilometres, three per cent at two to three, and nothing beyond. Their conclusion was that they found no evidence of cannibalisation, with the caution that interpretation requires care, and they report against themselves that the same pattern appeared for non-theatre art forms, which their own hypothesis had not predicted. Matthew Skiles and colleagues found the same shape in a different market: science and engineering conferences that moved online in 2020 drew audiences roughly 40 to 120 per cent larger than their own in-person editions, with travel cost as the proposed mechanism, and attendance from Asia failing to rise because the live sessions ran on a United States clock. Format substitution can add audience rather than take it, and whether it does is a design decision.

What it does not say

It does not say that screen time causes any of the displacement above. Wallsten states in his own words that he does not evaluate which online activities substitute for or complement offline ones. Every rate on this page carries the word correlated, and dropping that word is the single most common misuse of this literature.

It does not license scaling. The coefficients are marginal rates from a window ending in 2011, when online leisure was about thirteen minutes of a roughly five-hour leisure day. Multiplying them by four and a half hours produces a number that exists nowhere in the research.

It does not say the retreat indoors is a screen effect. Sharkey, who measured it, says the survey cannot see a smartphone and names alternatives. Anyone attaching his 99 minutes to screens has removed his limitation from the same paper it appears in.

And it does not resolve in either direction as a general claim. The Facebook deactivation experiment run by Hunt Allcott, Luca Braghieri, Sarah Eichmeyer and Matthew Gentzkow in 2020 found that offline socialising rose when the platform was removed for four weeks, which is direct evidence of displacement at the margin for one platform and one month. Patricia Valkenburg and Jochen Peter, testing displacement against stimulation in 1,210 Dutch adolescents, found stimulation. Rebecca Nowland, Elizabeth Necka and John Cacioppo conclude that the relationship is bidirectional and depends on use. The honest verdict is that it does both, which is why this page is graded mixed and why the evidence that matters is the kind that reports units.

---