1.97 per cent of scientists admit to faking their data at least once.

From Science Fictions by Stuart Ritchie.  Page 68.

So, if you combine the Leaderboard’s heavy hitters with all the other scientists who retract articles for dishonest reasons, how many scientists actually commit fraud? The overall proportion of papers retracted – around 4 in 10,000 published studies, or 0.04 per cent – is reassuringly low. It also isn’t very helpful, as we know on the one hand that some retractions aren’t due to fraud, but on the other that some journals either don’t catch false findings or don’t bother retracting them. What happens if instead you simply ask scientists, anonymously, whether they’ve ever committed fraud?
 
The biggest study on this question to date pooled research from seven surveys, finding that 1.97 per cent of scientists admit to faking their data at least once.  As if one in fifty scientists admitting to being fraudsters wasn’t alarming enough, consider that people are naturally loath to confess to fraud, even in an anonymous survey, so the real number is surely much higher. Indeed, when surveys asked scientists how many other researchers they know who have falsified data, the figure jumped up to 14.1 per cent (although of course, some of those polled might have been mistaken, paranoid, or exaggerating problems in their rivals’ research).