The 17% and the $1bn count different people

Three numbers about the same Thursday arrived between 20 and 26 October, and by the weekend they had been welded into a single sentence: GTA6 will empty American offices on 19 November, time-off requests are up 17 per cent, and the country loses a billion dollars. The first belongs to a GameSquare survey of 1,000 full-time employed Americans who describe themselves as gamers. The second belongs to BambooHR, which counted requests inside its own customer base and was reported by Money.com. The third belongs to José García Montalvo, an economist at Pompeu Fabra University in Barcelona, whose estimate was published in 2025 and reported again this week by IBTimes and local television stations. Separately, each is a defensible piece of work. Added together they describe a workforce that does not exist.

Start with what BambooHR actually said, because it is the strongest number of the three and the company is the most careful about it. Its figure is a band, 9 to 17 per cent, measured against the same mid-November window in recent years — English-language relays render the baseline as the third Thursday of November, which is the right comparison to make, since 19 November 2026 is itself the third Thursday. So the weekday matches on both sides and this is not a raw-day artefact. The company also declined to call the rise a GTA6 effect: it described the pattern as a real and repeatable signal, and said the data can only establish that two things coincided in time. A second, smaller early-access release is scheduled for the same date. Every headline written from that report converted the coincidence into a cause; the report did not.

The band itself went missing in the retelling. Headlines quote 17. The measurement is the range, and the top of a range is a maximum rather than a reading — if the effect were one thing measured one way there would be one number, so the spread is doing work that no outlet explains. One German outlet prints a per-day version of the same dataset, with 13 per cent on the 19th, 11 per cent on the 20th and 9 per cent across the weekend, and adds that a general rise of two to four per cent in requests should be subtracted from those figures; a Chinese relay of the same release states that no Friday data exists at all. Both cannot be right. What survives the contradiction is the band, and the band is unadjusted. Net of ordinary growth, the lift attributable to anything happening on 19 November is smaller than every version of the headline.

The billion travels with none of that caution and it is not from the same study. Montalvo’s model takes 1.5 million absent workers and an average annual output of $70,000 per American worker, and applies a short-term productivity dip of two per cent. Run the arithmetic on the two inputs the coverage names and it does not land on a billion: 1.5 million people times $70,000 is $105bn of annual output, two per cent of which is $2.1bn, while a single lost working day for that headcount at 250 working days a year is about $420m. The published figure sits between those two answers, closer to one per cent of annual output, or 2.4 working days. Fortune has separately reported that no published economic study supports the billion. An estimate whose input figures do not reproduce its output cannot be reconciled with a request counter from a different company.

The deeper problem is that the two datasets measure opposite things. A leave request is planned, approved and budgeted; the output is postponed, not destroyed, and in most payroll systems the cost was already priced into the employment contract. The number that would correspond to a real loss is unapproved absence — and BambooHR’s own survey says about half of the people intending to take 19 November off plan to call in sick rather than book leave, precisely because it is easier. Those people never file the request the platform measured. So the 9 to 17 per cent is not the visible tip of the absence wave; it is a different column, and adding it to a sick-day estimate double-counts nothing because the two sets barely overlap. GameSquare’s sample makes the third denominator: 10 per cent plan to skip the day outright and 30 per cent are considering it, of whom roughly 35 to 36 per cent would call in sick, about 11 per cent of the surveyed group. That is 40 per cent of gamers against 18 per cent of all full-time employees in BambooHR’s 1,500-person survey, and the 22-point gap is the gamer filter rather than a disagreement about demand. The 1.5 million at the centre of the billion is under one per cent of the roughly 160 million people employed in the United States — an order of magnitude below either survey’s own extrapolation.

Nothing here changes a purchase. The one date that binds a buyer is 12 November, when pre-loading opens and, on PlayStation, the refund right closes; the review embargo was reported for 18 November, so the crowd taking the 19th off will be playing a game whose reviews are a day old. What would settle the argument is absence data rather than request data, and that only exists after the date. BambooHR has not published the size of its panel, the number of years behind the band, or whether sick days are included in the requests it counted, and until it does, the honest description of this week’s coverage is a request count, a survey of gamers and a 2025 projection, printed as one measurement of one event.