Bob Murphy is nothing if not ambitious. He argues that there’s very little value in empirical economics today, mostly because you can make the numbers add up to whatever pre-conceived notion you want. If you believe that X causes Y, you can find cases in which your proxy for X correlates somehow with your proxy for Y close enough to declare victory. But at the same time, your opponent could just as easily come up with another proxy for X which conclusively proves (because of a lack of correlation with his proxy for Y) that you’re an idiot. He then pushes the envelope a little further, noting that empirical counter-factuals are not very meaningful when you can’t predict the future very well at all.
A classic example is the “stimulus.” After supporters put out a graph showing that unemployment could reach 10% if the government doesn’t spend a huge amount of money, we saw Congress pass and the President sign into law a bill authorizing $800 billion of government spending. Once this passed, we saw unemployment rates consistently above the “no stimulus” estimate. Supporters of the bill said afterward, “Gee, good thing we passed this—otherwise unemployment would have been even worse!”