You cannot control for ”all” other factors in an observational study. You can control for ”how people choose to answer your survey questions,” which is two steps removed from the truth.
Virtually every correlation study can be criticized through this method. Modern correlation studies 'account for' or 'control for' all observed variables in an attempt to prevent criticism along this line. This type of correlation study exists essentially to say "this correlation is interesting", and spark investigation as to why they are correlated.
Sociology is primarily correlation studies because of the logistical and ethical challenges around experimentation. For example, if we could force a randomly selected group of males to be unemployed we could avoid the third variable problem and establish causation. However no ethics review board would consider "I'm curious if your marriage will survive" a valid reason to fire someone.
It's not just ethics, people may figure out what you are doing and change their behaviour.