Tyler Cowen has an excellent Bloomberg article discussing the fact that economists did a poor job of forecasting the economy of 2023.  But the article has one paragraph that contains a troubling ambiguity:

Krugman has lately further explained his position — complete with unironic headline — suggesting that the untangling of broken supply chains had helped lower the rate of inflation. That point, too, is correct. He didn’t mention that there also has been a massive negative shock to aggregate demand: High rates of M2 growth became slightly negative rates of M2 growth. Fiscal policy peaked and then retreated. The Fed raised interest rates from near-zero levels to the range of 5%, and fairly rapidly. It also sent every possible signal that it was going to be tight with monetary conditions.

The meaning of the phrase “massive negative shock to aggregate demand” is more ambiguous than you might assume, perhaps even more ambiguous than Tyler assumed.  Here are two possible meanings:

1. There were massive policy initiatives that should have reduced AD, but failed to do so.

2. There was a massive reduction in AD that should have caused a recession, but failed to do so.

In the first case, we’d see what looks to many people like contractionary fiscal and monetary policy, combined with continued rapid growth in nominal spending.

In the second case, we’d see a sharp drop in nominal spending growth, but no significant drop in output and employment.

Either case would represent the failure of an economic model, but they would be very different models.  If contractionary policy failed to reduce AD, then our understanding of policy tools is flawed.  If a sharp drop in AD failed to cause a recession, then almost all versions of the Phillips Curve model would be flawed.

So what actually happened in 2023?  While we don’t yet have the 4th quarter data, it looks like it was mostly a failure of the first type.  Some supposedly contractionary policy tools failed to significantly reduce AD.  I say “significantly”, because while nominal GDP growth will likely be around 6% to 7% in 2023, that is a bit less than in 2022.  Even so, if you’d told economists in late 2022 that we’d have 6% to 7% NGDP growth in 2023, I doubt that very many would have predicted a recession.  So the events of 2023 in no way refute the assumption that a sharp slowdown in AD will generally trigger a recession—we failed to have a sharp slowdown in AD.

The model that did get refuted is the assumption that the fiscal policy and interest rates have a reliable impact on aggregate demand.  It isn’t just 2023, these tools have NEVER had a reliable effect on AD.  Economists who assume that rising interest rates are “tight money” are engaging in the fallacy of reasoning from a price change.  (Previous prediction failures of this sort were generally waved away with vague comments about mythical “long and variable lags”.)

This is not to say that all of the Phillips Curve models performed well in 2023.  When economists use inflation as a proxy for demand shocks, they are engaged in reasoning from a price change.  The Philips curve model that uses changes in the inflation rate did poorly in 2023.  In contrast, NGDP did pretty well as a policy indicator. The poor performance of the inflation/unemployment Philips Curve is probably due to two factors, both mentioned in Tyler’s article:

1. In 2022, actual inflation was above underlying inflation due to supply problems.  Then in 2023, actual inflation was below underlying inflation as supply chains healed.  Thus underlying inflation fell by much less than headline inflation.

2. The Fed had been targeting inflation at roughly 2% for 30 years and therefore inflation credibility was not completely lost during 2022.  Thus, it was possible to reduce inflation with less of hit to real output than we saw back in 1981-82.

Many economists might be inclined to re-evaluate their model after the failed predictions of 2023.  In my case, 2023 has caused me to double down on my key beliefs:

1.  Aggregate demand models should focus on NGDP growth, not inflation.

2.  Interest rates are a poor indicator of the stance of monetary policy.

3.  It’s almost impossible to forecast turning points in the business cycle.

Tyler mentioned that the Fed raised rates “from near-zero levels to the range of 5%, and fairly rapidly”.  I’m not sure if he thinks adding “fairly rapidly” is a way to avoid reasoning from a price change, but it isn’t.  It’s true that raising rates rapidly often makes policy more contractionary than otherwise, but it doesn’t provide any indication of whether policy was contractionary in an absolute sense.  In retrospect, it’s clear that the Fed was way behind the curve in early 2022, and the sharp rise in rates merely moved policy from highly expansionary to something close to neutral.  There was no massive negative demand shock.