The Possibility of Social Science Fakery

There are many scandalous cases of physical scientists or medical researchers faking their data. E.g., Schon's plastics research, Ninov's "discovery" of element 118, Scott Reuben's faked medical trials, and many more.

For anyone contemplating inventing their own data, it's a risk-reward calculation: The risk of being discovered and drummed out of a profession versus the reward of whatever fame accrues to new scientific discoveries or publications. But it's amazing that physical scientists would ever fake data, given the high risk of being discovered when other scientists are unable to replicate their findings.

But some do it anyway. Which makes me wonder about social science, where the risk would be far lower. There's no general expectation that social science "discoveries" are going to be replicable 100% of the time in any new dataset or research that comes along. For example, if one psychologist does a psychological experiment of some sort, and someone else does a similar experiment with different results, that can always be explained by the fact that the experiment happened at a different time and place involving different people as subjects, or maybe there was some seemingly minor difference in the experimental conditions, etc.

So if people are willing to fake results even when they surely know of the high risk of being caught, how much more willing would they be to fake results when the risk is far lower? How much social science fakery is there?