"We also obtain information about employees’ gender, seniority, and the number of skills listed on their individual LinkedIn profiles, which serves as a proxy for employees’ skill level."
'Best talent' can be defined only in relation to a particular role, i.e. the job title, expectations and credentials.
If the study were to show that, of the software engineers at level X at a particular company, RTO mandates caused the loss of a greater percentage of the better performers, than of the lower performers, that would:
- be interesting
- match my intuition.
Unfortunately, the study does not seek to estimate this, nor does its design make that possible.
Using seniority as proxy for 'best talent' seems flawed because seniority doesn't mean someone isn't underperforming, as this is always relative to expectations.