Consider the following two paradoxes:

1. Falling wages are associated with falling RGDP. Falling wages cause higher RGDP.

2. Falling interest rates are associated with falling NGDP growth. Falling interest rates cause higher NGDP growth.

You often hear people say correlation doesn’t prove causation, but rarely do correlation and causation go in opposite directions as often as in these two cases. I’ve talked a lot about interest rates in this blog, is there anything to be learned by comparing interest rates to wages? I believe the answer is yes.

The following is going to be an ad hoc model, which just happens to be true during most of recent US history. The “never reason from a price change” maxim warns us to not assume that it will always hold true.

Let’s suppose that most employment fluctuations are caused by the combination of NGDP shocks and sticky nominal wages. For simplicity, assume that when NGDP falls by X%, nominal wages fall by only one half times X% in the short run, that is, only half as much as would be required to keep the labor market in equilibrium. We might observe three countries that see NGDP plunge by 2%, 10% and 20%. In those three countries nominal wages fall by only 1%, 5%, and 10%, that is by one half as much as NGDP fell. That’s what we mean by sticky wages.

Now think about what that implies. The country where wages fell by the most (minus 10%) is the country where wages are the furthest above equilibrium. And that’s likely to be the country with the highest unemployment rate. What conclusion would the average person draw from these stylized facts? They’d conclude that wage cuts “don’t work”. Cutting wages just causes NGDP to fall even further, and is thus self defeating. I’d argue that this is one of the most important themes in Keynes’s General Theory.

And yet I believe this view is completely wrong. For the country where NGDP fell by 20%, a larger wage cut, say 15% or 18%, would have moved wages closer to equilibrium, and this would have led to lower unemployment. Keynes had causation exactly backwards, on an issue that is central to his critique of classical economics.

I hope that by now you see the connection to interest rates. Falling interest rates are a sign of a weak economy. As with wage cuts, a Fed decision to cut interest rates makes the economy stronger than otherwise. So then why are falling interest rates usually associated with a weak economy? Because a weak economy puts downward pressure on market interest rates, and the vast majority of Fed rates changes are merely reacting to changes in the economy, not causing them.

Here’s a good recent example. Since the December rate increase, short-term interest rates in the fed funds futures market have been trending downwards. Instead of the 4 rate increases in 2016 predicted by the Fed last month, markets are now forecasting only one or two at most. So what are we to make of this change in the expected path of rates? It’s theoretically possible that this reflects an expected easing of monetary policy. That is, the Fed is now less likely to raise rates, even assuming no change in the macroeconomic environment. An easier money policy, which would be expected to boost growth.

In fact, it’s far more likely that these lower interest rate forecasts are a prediction of a weaker than expected economy, and also a prediction that the Fed will react to the weaker than expected economy by raising rates less that previously expected. Do we have any evidence for this claim? Yes, a mountain of evidence. All sorts of other asset markets are becoming much more bearish about the economy. If the Fed’s likely decision to move away from an aggressive path of rate increases really were an expansionary policy, then asset prices would be rising as bond yields fell. But asset prices are falling with bond yields.

Interest rates usually fall when there is a NGDP growth slowdown. And yet it’s also true that interest rates are usually too high when there is a NGDP downturn. This implies the Fed’s target interest rate is sticky; it falls more slow than would be required to maintain stable NGDP growth. Sound familiar? So periods when interest rates are falling are also periods where interest rates are becoming increasingly too high.

I believe that Keynes noticed that the deeper the depression, the lower the level of interest rates. Because he viewed low rates as being an expansionary monetary policy, he wrongly concluded that interest rate cuts were relatively ineffective in a deep depression. Unlike with wages, he did not wrongly reverse causation; he was too good a monetary economist to do that. (We’d have to wait for the Neo-Fisherians for that error.) But he did become excessively pessimistic about the potency of monetary stimulus in a depression, and for much the same reason that he wrongly thought that wage cuts would be ineffective in a depression.

His erroneous views on wage flexibility led him to reject classical solutions for depressions. In fairness, wage flexibility is not the best solution, even if Keynes was wrong about causation. The second error led Keynes to reject the views of progressives like Fisher, Hawtrey, Cassel, and even the Keynes of the Tract on Monetary Reform, who favored using monetary policy to stabilize the price level (or NGDP.) Keynes wasn’t hostile to their suggestion, he just didn’t think that monetary policy alone could get the job done.

By the 1990s, New Keynesians had moved away from Keynes on these issues. They thought monetary policy was enough for stabilization of aggregate spending. And they thought wage cuts were expansionary. Due to the zero bound, economists have recently drifted back to old Keynesian ideas. My view is that they were that view was wrong in the 1930s, wrong in the 1990s, and they are wrong today.

PS. When would wage changes be associated with output moving in the opposite direction? Perhaps if they were caused by exogenous policy shocks, such as the higher wages after July 1933 implementation of the NIRA, which slowed the recovery, or the lower wages resulting from Germany’s labor market reforms of 2004, which helped boost growth and reduce unemployment.