A may survey of 1000 people reported that 39% would consider quitting if their work didn’t provide flexibility.
I’m no statistician but the red flags are n=1000 and “would consider”. This is a hugely hit button issue right now, great for clicks.
It's been awhile since I've taken stats, but people semi-frequently overestimate the necessary sample size needed for even a 99% confidence level.
Depends on the standard deviation for sure, but some quick napkin math suggests that a sample size of ~700 people is more than enough to draw inferences about the entire western hemisphere at a 99% confidence level.
Reasons for enlarging samples have less to do with accuracy than with resolution. If you plan on looking at specific cross-tabulations or sub-populations, you need a sufficient sample size (typically n>30 for large-sample statistics) to draw inferences.
Error decreases with the square root of sample size. To halve error you must square the sample. A 1,000-element sample has half the error of a 31-element sample. To halve error again would require a 1,000,000 element sample.
"Would consider" is a fair catch, though here the question is how that compares against previous measures / trend.