Thread by @ModeledBehavior

I want to do a quick tweet storm about an interesting new paper out on WFH at an IT services company that found productivity declined. bfi.uchicago.edu/wp-content/upl…

Before I dig in, I want to emphasize something really important about some kinds of technological change like remote work and how it will impact productivity in the long-run: it works through selection.

During the pandemic, everyone who could worked remotely. According to Gallup, over half of workers were remote full-time early in the pandemic. But that’s not how all long-run technology adoption works.

In my work I estimate about 20% will be fully remote long-run. This means there are clearly lots of tasks and industries that will be done in person. Selection means that those for whom it works embrace, and those for whom it doesn’t, won’t.

In short, remote work is not universal and technological change is not randomly assigned. Selection -> productivity effect is bigger in the long than short-run. As a result, some companies will find productivity goes down. Not surprising.

That said, I am not convinced this study convincingly shows a negative productivity effect. I’ll dig in to why now.

First, note the study does not show that output went down. Output was unchanged. Instead, hours went up. However, I think there are serious measurement issues with both of their measures of hours.

Their primary measure is based on a productivity tracker. However, if you look at table 1, they find that workers do 25 hours of work a week with this measure. That’s a lot of off-computer work that is not being captured.

These results are consistent with off computer work (like stopping by a co-workers office), instead being done on computer (sending a slack message). Or reading a document on the computer (measured) instead of printed out (unmeasured).

The second measure of hours they have uses Outlook data and other MS teams data. Here we see a longer workday as well, including more after hours work. However, this data can’t tell breaks and focus time, both are empty just calendar time.

The results are all consistent with various off-computer activities being more on-computer during WFH, and people taking more breaks during the day and making up for it by working somewhat later. The latter could also be pandemic related.

I wouldn’t dismiss the study entirely, they have a ton of data and it’s very interesting, but I am not convinced yet. This study found that output (which is measured well) didn’t decrease from WFH. Discussions should not ignore that.

Either way though, selection effects mean WFH doesn’t have to work for every company, and it certainly won’t. That is completely consistent with huge positive productivity effects for the economy overall.

cc @Richard_Florida @arpitrage. With credit to @mattsclancy for a useful discussion of this paper.