Michael Housman, chief analytics officer for Evolv, discusses workforce science with Stephen Dubner (Think Like a Freak):
We looked specifically at pay in a research study that we just finished. We found, there is no question that pay enables people to stay longer, and they perform better. But the magnitude of the effects were actually not as big as we had expected. So for every 10 percent increase in pay, there’s a 5 percent reduction in quitting behavior. So it’s a less than one-for-one offset. And what’s more, is that when someone receives a raise, there are kind of these warm fuzzies that are associated with receiving the raise. There’s this halo effect. We found that that effect lasts longer than a week, but not as long as a month.
[...]
Your supervisor alone accounts for about as much variance in terms of longevity in these roles as everything else combined. The effects are staggering. Anecdotally, this seems to resonate with people because everyone has had a bad boss that made them leave the job. And we’ve really made understanding that supervisor/employee relationship a priority of ours because I came into this thinking that it was all about raw talent. You get the right person in the job and everything will work itself out, and that’s really the key decision. Our research has actually show that that’s actually a relatively small piece of the pie, something in the range of 10 to 15 percent.
[...]
What we found was that people who said they were honest actually were 33 percent more likely to be terminated for policy violations. So, learned our lesson, which is you don’t ask people if they’re honest because you tend not to get an honest answer.
[...]
We came up with a very creative way of measuring what we think is honesty and integrity, which is that we asked them upfront, early in the assessment, how are your computer skills, what’s your typing speed, do you feel comfortable with the keyboard and mouse, toggling between the screen and so on and so forth. And then guess what? About five or six screens later we tested them. We asked them what’s the shortcut for cutting and pasting text using a word processor. We actually measured their typing speed and accuracy. And what we found when we compared their self-assessed responses to their actual technical proficiency is that there were two groups of people that came out. One group was relatively honest. They were what they said they were in terms of the technical skills. And the other group we will call a little bit creative in that they claimed to be exceptional with the keyboard and mouse, but they couldn’t type more than 10 words a minute.
Evolv found that the honest employees tested better on just about every performance metric — except sales.