The evidence
What a study can and cannot tell you
Four things research on this subject genuinely establishes, four it is routinely made to say, and the difference between them.
Claims this part is about, and what each was answered with
New starters take longer to become productive remotely
Empirical→EvidenceThe one claim answered with evidence that measured it
Productivity fell when we went remote
Empirical→DataRemote workers put in longer hours than office workers
Empirical·Nothing; it was never answeredthree claims · three empirical · none of the answered ones was answered in the wrong kind · one was never answered at all
Three empirical claims with three different fates.
There is a real literature on working location and it is better than either side's use of it. Being specific about what it supports is worth more than another round of citation.
Four things it does support
That output did not collapse. Across many large samples and several countries, the catastrophic productivity drop predicted in 2020 did not appear. This is about as well established as anything in the area.
That the effect is not uniform by task. Work requiring sustained individual concentration and work requiring frequent negotiation behave differently. This is consistently found and is the most useful finding for anybody actually designing an arrangement.
That people's stated preferences are strong and stable. Large surveys agree on roughly how much people value flexibility and that the valuation has not fallen. This is a fact about preferences, not about outcomes, and it is solid.
That early-career outcomes look different. The evidence on new entrants is more consistent than most of the rest, and points towards a real cost to isolation at the start of a career.
Four things it is made to support and does not
That offices are more productive, in general. The unit is usually the firm, the comparison usually spans a period in which much else changed, and the people who chose each arrangement differ from each other. Any of those three breaks the inference.
That remote work is just as good. Same objections in the other direction. "No collapse" is not "no difference", and absence of a large detectable effect is not evidence of no effect on the things that were not measured.
That culture requires presence. Almost nothing measures this, because almost nobody has operationalised culture in a way that could be measured. What exists measures reported belonging, which is related and not the same.
That any particular number of days is correct. Nothing anywhere supports three rather than two or four. The number is always a judgement, and dressing it in research is the most common misuse of the literature in this argument.
The table version
| Question | Does the literature answer it? |
|---|---|
| Did output collapse? | Yes — no |
| Does task type matter? | Yes — considerably |
| Do people value flexibility? | Yes — a lot, stably |
| Do new entrants do worse remotely? | Reasonably well — yes, on several measures |
| Are offices more productive? | No |
| Is remote just as good? | No |
| Does culture need presence? | Barely addressed |
| How many days? | Not at all |
How to use it honestly
Cite it for the four things in the top half, and say what the unit is when you do. "There is good evidence that new entrants do worse remotely on time-to-productivity" is a sentence that survives scrutiny, is useful, and is narrower than what people usually say.
And when somebody asks what the literature says about the number of days, the honest answer is the short one: nothing, and anyone telling you otherwise is reading a judgement as a finding.
The question that is never studied
Nothing in the literature addresses the question that these discussions are actually about, which is how much of somebody else's time a benefit is worth.
That is not a gap in the research. It is not the kind of question research answers. Every study in this area can tell you, at best, that an arrangement produces some effect of some size; none of them can tell you whether that effect justifies seven and a half hours a week of a particular person's life.
Which means that even a complete and perfect literature leaves the central question untouched, and the central question is a value question with a cost attached.
What to do with somebody who will not accept this
Some people hold, sincerely, that the matter is empirical all the way down — that given enough data the right arrangement would be calculable.
The useful response is not to argue about it in the abstract. It is to ask what the calculation would take as inputs. Within a few steps the answer requires a weighting: how much is an hour of an employee's evening worth against a marginal unit of collaboration?
That weighting is not in any dataset. Pointing out where it enters the calculation is more convincing than any amount of argument about the nature of values, and it is a question rather than a claim.