Gene Callahan links to my discussion of the shifting unemployment-investment curve that John Taylor posted, and points out that this is true of the social sciences in general. We are capable of identifying meaningful relationships - even meaningful causal relationships. But the parameters of those relationships, because we're dealing with such a complex system, tend to shift. That means we have a real phenomenon but our observation of it is a little different over time. OK, big deal. That is the nature of the social sciences and any science of a complex system. Gene talks specifically about Zipf's law, a neat little regularity that is near and dear to my own heart.
While we're still on Taylor's post, I want to point out a good response from Noahpinion that essentially makes all the points I have made over the last couple days (this isn't surprising at all, everyone should know this, the curves shift over time, etc.).