Macroeconomics makes progress, one dead prediction at a time
By Scott Sumner
This will probably be my most important post of the year, but I predict that almost no one will pay any attention.
Macroeconomics is a deeply flawed field, because it is extremely hard to do controlled experiments. Ideally, you’d like governments to do wild and crazy things, like a sudden 30% cut in the money supply—just to see what happens. But after the Great Depression, governments have little stomach for these sorts of experiments. Nonetheless, they do happen on rare occasions.
June 2016 was one of those occasions. The British public voted for Brexit, creating enormous uncertainty about the future course of the UK economy. This was widely treated as the biggest shock to hit the UK since the Suez crisis, if not WWII. There was great uncertainty about how it would play out. Would City banks still have (passport) access to the Continent? Would it be a soft Brexit (with continued free trade) or a hard Brexit? The consensus among economists was that this uncertainty shock would lead to a dramatic slowdown in growth.
The Economist magazine reports the consensus growth forecast for 2017. In early July, they reported a sudden drop from 2.0% in June to 0.8% in July:
But even that understates the shock, as not all forecasters had revised their predictions. By the August issue the forecast for 2017 had dropped to 0.5%, where it remained in September. That’s actually consistent with a couple negative quarters, and hence a mild recession. For instance, the US experienced a mild recession in 2001, despite 1.0% RGDP growth (calendar 2001 over calendar 2000.) Since September, however, the Economist’s consensus forecast has been steadily creeping higher.
Then a few days ago the UK came out with shocking 4th quarter RGDP data, showing a 0.6% increase (an annual rate of 2.4%.) The second such increase in a row. Now the Bank of England has revised its 2017 growth forecast back up to 2.0%:
Alongside news that the BoE had kept its key interest rate at a record-low, the British central bank lifted its 2017 economic growth prediction to 2.0 percent from a 1.4-percent forecast just three months ago.
It now appears that Brexit (i.e. the vote, not the policy) had absolutely no impact on RGDP growth. Sound familiar? Recall when the 2013 austerity was widely expected to produce a sharply showdown, and growth actually sped up. (Unlike Brexit vote, the austerity began January 1, 2013, so I use Q4 over Q4 figures.)
In the field of macroeconomics it is very unusual to see an almost perfect test of the “uncertainty theory of business cycles”. I certainly never expected to see such a nice natural experiment. Now that it is over, we can remove one suspect from the list of possible causes of business cycles, which include monetary shocks (falling NGDP) and real shocks (things that cause RGDP to fall even if NGDP growth is kept stable by monetary policy.)
My own view was more optimistic than the consensus, but even my view turned out to be wrong. I thought unemployment might rise by about 0.5%, whereas now it looks like unemployment will not rise at all.
As of today I am even more a market monetarist than I was last year. I’m even more convinced that NGDP shocks are what cause employment fluctuations. That’s not to say that real shocks don’t matter. I still think a complete shutdown of oil from Opec or a $20 minimum wage would cause a recession, even with steady NGDP growth. But I don’t expect to see either of those shocks.
Back in March 2011, we learned that even a devastating earthquake/tsunami, which led to the complete shutdown of the extremely important Japanese nuclear power industry, would not impact the Japanese unemployment rate. So we can also rule out natural disasters as a cause of business cycles for big diversified economies. (On the other hand a small economy can be tipped into recession by a natural disaster, as we saw a few years ago in Haiti.)
And of course we now know that a sudden $500 billion cut in the budget deficit will get offset by Fed stimulus, even at the zero bound.
I started off with the pessimistic prediction that this natural experiment would not matter. I believe that economists who favor uncertainty theories of the business cycle will simply wave away this near-perfect experiment. They might argue there actually wasn’t all that much uncertainty. But that would merely prove uncertainty theories to be useless for a different reason—because we are unable to ascertain in real time what uncertainty looks like. Or they will argue that some other mysterious factor prevented their theory from working, as Keynesians did after the failed 2013 austerity predictions.
We don’t live in a Bayesian world where macroeconomists update their priors after failed experiments; we live in an ideological world. But the minority of us who do have open minds learned a lot in 2016–and we should celebrate that success even if we were wrong (as I was somewhat wrong.)
PS. The effect of Brexit itself (expected in 2019) is a completely different issue. I do expect Brexit to reduce UK RGDP growth, but only modestly. I remain opposed to Brexit. But I’ve changed my priors on uncertainty theories from “a negative real shock but less important than the consensus believes” to “not a factor at all.” Maybe I’ll later change my mind on Brexit itself.
PPS. If you are puzzled about how an economy can grow in a recession year, imagine the following thought experiment. Suppose an economy grows at a two percent rate all throughout the year before a recession. It ends the year 2% above the previous year. Now let it decline a couple tenths of percent for two quarters in a row, and then level off. For the year as a whole, RGDP is actually up about 0.8% over the average of the previous year, even though the year saw slightly negative growth on a Q4 to Q4 basis.
PPPS. Those reluctant to change their minds should reflect on the fact that the “losers” of intellectual debates are actually the winners. The “losers” learn something, becoming smarter. The winners learn nothing, and simply become even more arrogant jerks.