I was struck to discover that there exists a remarkably simple statistical explanation for why people so often believe punishment to be more effective than reward. I found out about it in the following excerpt from Enlightenment 2.0 written by substacker Joseph Heath .

“When a dog is trying to perform some new trick, what you typically see will be a series of fairly average performances, along with occasional deviation—every so often you will see a particularly good performance or else a particularly bad one. It is, however, a statistical fact that uncommon events are uncommon, and so an uncommon event is more likely to be followed by a common event than by another uncommon event. This is called regression to the mean. As a result of it, an exceptionally bad performance, because it is uncommon, is likely to be followed by an average one, just as exceptionally good performance, because it is uncommon, is likely to be followed by an average one—regardless of whether the bad performance is punished or the good performance rewarded. But if you do start punishing and rewarding, a casual observer is likely to be fooled into thinking that the punishment worked, because performance improved after it was applied, while the reward not only failed, but actually had the perverse effect of encouraging bad performance. The result is an extraordinarily common bias that leads people to overestimate the effectiveness of punishment.“