Curves shift - it's not a big deal

This morning I was complaining about people who got excitable about John Taylor's nice graph of unemployment vs. investment percent of GDP. Not really all that surprising, I said.

Now, it seems people are going from "wow it's remarkable" to "it's meaningless".

Specifically, Justin Wolfers shares this graphic going back to 1948:

And concludes it is "decidedly unimpressive". This is the wrong way to look at this sort of data. Relations in economies are not stable. They shift over time. Technology changes. Preferences change. What you should see in this graphic is about four curves like the one John Taylor first showed laid on top of each other, each shifted a little because of some fundamental change in the economy. We see it with the Beveridge Curve. We see it with the Phillips Curve. Every time these curves shift people want to call the time-of-death on a very solid underlying phenomenon. That's bad economics. Taylor's original graphic was fine - what was a little questionable was Taylor's interpretation of it. We know employment is highly correlated with GDP. We know investment moves a lot and consumption doesn't over the business cycle. We should expect unemployment to be highly correlated with investment.

If it moves over time, who cares? Curves shift. All this means is that any business cycle theory worth its salt has to answer the questions (1.) why does investment change, and (2.) why don't we stay on the production possibilities frontier - why isn't the reduction in investment matched by an increase in consumption?