What the estimates say—and what they do not say

The central result in the work-from-home study is easy to quote and easy to misread. This chapter interprets the relationship between WFH and disability employment by keeping its units, outcomes, comparison measures, and occupational reach in view. The goal is to understand what a one-percentage-point change in WFH means for full-time employment, why the result is not equally clear for every form of work, and how robustness checks strengthen the conclusion without turning it into a guarantee for an individual worker.

The strongest result is specific to full-time employment

The baseline estimate says that a one-percentage-point increase in an occupation's WFH share is associated with a 1.022% increase in full-time employment for people with a physical disability. The corresponding instrumental-variable estimate is 1.268%. The wording has three parts that must stay together.

First, the change in WFH is a percentage-point change in a share. If an occupation's WFH share rose from 10% to 11%, that is a one-percentage-point increase. Second, the employment outcome is a percent change in the number of disabled workers employed full time in that occupation. The estimate does not say that the national employment-population ratio rises by 1.022 percentage points. Third, the comparison is occupation-level and relative to the study's pre- and post-pandemic windows. It is not a promise that a particular worker receiving a remote option will experience a 1.022% improvement in their personal employment probability.

That distinction is more than statistical etiquette. A percentage-point change describes movement in a share; a percent change describes movement relative to a starting amount. Suppose an occupation had 1,000 workers with a physical disability employed full time before the comparison and its WFH share increased by one percentage point. The baseline coefficient would imply roughly 10 additional full-time workers in the model's occupation-level relationship, not 1.022 additional percentage points in the employment rate. The example illustrates the unit; it does not turn an average estimate into a forecast for that occupation or worker.

The result is also outcome-specific. The paper reports that the increase in disability employment is driven by full-time work. Full-time employment accounts for 78% of employment among people with a physical disability in the study's sample, and this is where the WFH relationship is largest. The baseline estimate for total employment is 0.608, with an IV estimate of 0.734; the paper describes the IV total-employment result as not statistically significant in the reported specification. For part-time employment, the baseline estimate is −0.098 and the IV estimate is −0.842, both with large standard errors. The paper's interpretation is that WFH had essentially no net effect on part-time disability employment.

“Essentially no net effect” should not be translated into “WFH never affects part-time work.” Some workers could have entered part-time work while others moved from part-time to full-time, leaving little net change. The point is that the study does not estimate a precise positive or negative part-time effect.