Scott Sumner is actually discussing discrimination, and how discrimination is often "proven" in social studies
The economy operates in very subtle ways, and often when I read academic studies of issues like discrimination, the techniques seem incredibly naive to me. They might put in all the attributes of male and female labor productivity they can think of, and then simply assume than any unexplained residual must be due to "discrimination." And they do this in cases where there is no obvious reason to assume discrimination. It would be like a scientist assuming that magicians created a white rabbit out of thin air, at the snap of their fingers, because they can't think of any other explanation of how it got into the black hat!
Most alarming climate forecasts are based on the period from 1978 to 1998. During this 20 year period world temperatures rose about a half degree C. People may say they are talking about temperature increases since 1950, but most if not all of those increases occurred from 1978-1998. Temperatures were mostly flat or down before and since.
A key, if not the key, argument for CO2-driven catastrophic warming that is based on actual historic data (rather than on theory or models) is that temperatures rose in this 20 year period farther and faster than would be possible by any natural causes, and thus must have been driven by man-made CO2. Essentially what scientists said was, "we have considered every possible natural cause of warming that we can think of, and these are not enough to cause this warming, so the warming must be unnatural." I was struck just how similar this process was to what Mr. Sumner describes. Most skeptics, by the way, agree that some of this warming may have been driven by manmade CO2 but at the same time argue that there were many potential natural effects (e.g. ocean cycles) that were not considered in this original analysis.