Hume, hockey sticks, and The Great Forgetting
By Scott Sumner
In the 21st century, macroeconomics is entering a new Dark Ages. We seem to be forgetting much of what we learned in the last half of the 20th century.
That judgment may be harsh, but if I’m wrong then you should no longer read anything I write (including the book I have coming out this year), because in that case I wouldn’t actually know anything useful about macroeconomics.
Perhaps the most firmly established proposition in late 20th century macroeconomics is that the Great Inflation of 1966-81 was caused by central banks printing too much money. If that proposition is wrong, then I might just as well give up. Everything else I believe about macro hinges on that being true. If the Great Inflation wasn’t caused by too much money, then what can macroeconomics tell us about the world?
1. Hockey sticks, wherever you look
People often talk about “hockey stick” graphs. Deirdre McCloskey using that metaphor for the explosion in real incomes in recent centuries. Demographers use it for the Earth’s population. Environmentalists use it for carbon emissions. Tech people for gigabits of information. And there’s also a hockey stick graph for the US price level, as well as the price level of many other developed countries. The US price level was fairly flat during the first 150 years of US history, and then exploded upward 20-fold after 1933.
Back in 1752, this was all explained by David Hume. Out-of-sample forecasting is the gold standard of model success, and by that standard Hume might be the greatest macroeconomist of all time. Of course he didn’t use modern terminology like “nominal GDP”, but if you translate his core argument into modern lingo, it goes something like this (my words):
A large exogenous increase in the money supply will cause a roughly proportional increase in NGDP. There would not be a precise correlation, as velocity moves around as well. Because real growth is caused by non-monetary factors, excessive growth in money will simply lead to inflation, as NGDP growth outpaces RGDP growth.
During the first 150 years of US history, the monetary system was commodity based (bimetallic and then a gold standard.) That kept money growth close to the rate of growth in RGDP, and hence there was almost no long run inflation (although prices moved up and down erratically around a near-zero inflation trend line.)
Then we tried Hume’s experiment. After many decades of fixing the price of gold at $20.67/ounce, we let it rise sharply. Today it’s over $1800/oz. We used that freedom to print lots of money, and as the monetary base began rising much faster than RGDP, inflation also took off. The numbers were pretty much what Hume would have predicted. And most of the “anomalies” were explained during the 1950s, 1960s and 1970s by economists doing technical research into what caused shifts in money demand. The one thing Hume didn’t predict is the advent of interest on bank reserves in 2008, which caused the relationship between the money supply and inflation to weaken. But that wasn’t a factor during the Great Inflation.
2. The Great Forgetting
Doug Irwin directed me to a recent paper by Itamar Drechsler, Alexi Savov, and Philipp Schnabl on the Great Inflation. Here’s the abstract:
We propose and test a new explanation for the rise and fall of the Great Inflation, a defining event in macroeconomics. We argue that its rise was due to the imposition of binding deposit rate ceilings under the law known as Regulation Q, and that its fall was due to the removal of these ceilings once the law was repealed. Deposits were the dominant form of saving at the time, hence Regulation Q suppressed the return to saving. This drove up aggregate demand, which pushed up inflation and further lowered the real return to saving, setting off an inflation spiral. The repeal of Regulation Q broke the spiral by sending deposit rates sharply higher. We document that the rise and fall of the Great Inflation lines up closely with the imposition and repeal of Regulation Q and the enormous changes in deposit rates and quantities it produced. We further test this explanation in the cross section using detailed data on local deposit markets and inflation. By exploiting four different sources of geographic variation, we show that the degree to which Regulation Q was binding has a large impact on local inflation, consistent with the hypothesis that Regulation Q explains the observed variation in aggregate inflation. We conclude that in the presence of financial frictions the Fed may be unable to control inflation regardless of its policy rule.
I’m going to annoy almost everyone here, so let me apologize in advance. I realize that these authors are now in the mainstream and that I’m a hopeless dinosaur. It’s a perfectly fine paper by conventional standards. Nonetheless, I can’t get past the very first sentence of the abstract. Why do economists think it’s a good idea to propose a new explanation for the Great Inflation? Imagine a physics paper that began by noting that while Isaac Newton had already proposed a theory for why feathers and steel balls fall at the same rate in a vacuum tube, the authors were about to provide a “new explanation”. Why?
As I got into the paper, my frustration only increased. I expected them to push back on what I thought was the standard theory, the idea that the Great Inflation was caused by the Fed aggressively increasing the money supply. Instead, they suggest that the standard explanation is that the Fed failed to act aggressively to stop the Great Inflation, which presumably happened for some other unnamed reason:
The standard narrative of the Great Inflation places much of the blame on the Federal Reserve. By failing to act aggressively enough, the Fed had allowed inflation to get out of hand and squandered its credibility with the public (Clarida, Gali and Gertler, 1999). The loss of credibility raised inflation expectations, which made inflation accelerate further.
Now you might argue that an aggressive rise in interest rates would have reduced money growth, but that just annoys me in two other ways. First, it confuses cause and effect. Prior to 2008, the Fed raised interest rates by reducing money growth, not vice versa. Second, it ignores the income and Fisher effect, and more broadly the NeoFisherian perspective. The Great Inflation happened partly because the Fed had forgotten Humean economics, the idea that inflation is caused by printing too much money. Instead, the Fed wrongly thought that the high interest rates of the late 1960s and the 1970s were a tight money policy. It was not.
The truth is that I’ve always been a bit uneasy about the standard story of inflation in the 1970s, even though it seemed to fit the prediction of clockwise spirals. My uneasiness came from the sense that the economy never seemed to run hot enough to explain such a big rise in inflation. I actually remember the 70s! And while there were years of good job markets, they never felt as good as the 60s, the late 90s, or 2019.
Krugman’s “standard story” is not printing money, it’s not an unaggressive Fed, it’s not even Regulation Q. Rather Krugman suggests that the standard story of the Great Inflation is the Phillips curve model—high inflation is caused by low unemployment (a hot economy). But while the Phillips curve may have some limited ability to account for negative inflation/unemployment correlations at business cycle frequencies, it tells us precisely nothing about long run inflation dynamics (or, as Lucas (1975) showed, cross sectional differences in inflation.) And the Great Inflation was a long run phenomenon. There were four recessions between 1966 and 1982. The economy certainly ran “hot enough” in NGDP growth terms to fully explain the Great Inflation.
More broadly, the Great Inflation is clearly just an extreme event embedded within a much greater Great Inflation of 1933-2020, when the CPI rose 20-fold. I cannot emphasize enough that this was a monetary event. Unemployment wasn’t substantially different during the gold standard era than during modern times, indeed it was lower during 1923-29 (a period of roughly zero inflation) than during the 1970s, a period of high inflation. The price level is 20 times higher in 2020 than 1933 (after essentially no change in the previous 150 years) because we printed a lot of money, not because of anything to do with Phillips curves or Regulation Q.
3. Cognitive illusions, everywhere you look
Since the high inflation on the right side of the price level hockey stick was clearly due to monetary policy, why wouldn’t we have expected a sharp acceleration of money growth in the 1960s to lead to a sharp acceleration of inflation? Why did we forget the monetary theory of inflation and begin seeking other explanations?
I see three explanations, each of which plays a role:
1. Too much focus on Phillips curve models, which should not be used to explain persistent changes in inflation.
2. Too much focus on interest rates as an indicator of whether money is easy or tight. This makes it look like monetary policy doesn’t play much of a role, as high interest rates usually occur during periods of high inflation, and vice versa.
3. Incorrect predictions of inflation by monetarists and others when the Fed adopted quantitative easing, as pointed out by the authors of the Regulation Q study:
Scholars of the Great Inflation, such as Allan Meltzer (2009; 2013) and John Taylor (2009), worried that low interest rates and Quantitative Easing would lead to a repeat of the 1970s.
Lots of monetarists underestimated the impact of the 2008 decision to begin paying interest on bank reserves, which weakened the link between money growth and inflation. Hume’s amazing “out-of-sample” success that had lasted for 256 years came to an end in 2008.
I can’t really blame younger economists for finding the monetarist view of inflation to be hopelessly outdated. Even before 2008, the Fed was already targeting interest rates. Lots of economists don’t even know that (prior to 2008) the monetary base was 98% currency (the stuff in your wallet) and that the Fed moved interest rates by adjusting the quantity of base money, which as a practical matter meant adjusting the supply of currency.
Almost every aspect of our monetary system was designed to hide what’s actually happening. Even though the base was 98% currency in 2007, and even though almost all new base money quickly went out into circulation as currency, the new money did initially enter the system as electronic bank reserves. And even though the Fed controlled inflation by controlling the monetary base, they did so by moving the base to a position that moved interest rates to a position that was expected to lead to 2% inflation. In that regime, it’s easy to ignore money entirely, even though it was the money printing that was actually causing the inflation.
After 2008, it became even easier to ignore money. With the payment of interest on reserves the Fed no longer even had to adjust the monetary base in the background in order to move interest rates. Now they could simply adjust IOR to move rates directly.
Like journalists, economists are seduced by power. Once the powerful central bankers decided to focus on interest rates, macroeconomists decided that interest rates were the appropriate monetary variable for their models. The more “money-less” your model, the more sophisticated and up to date it seemed. What sounds more appealing to young economists, Michael Woodford’s futuristic vision of a cashless economy, or my reactionary obsession with the old days when the currency stock was more important?
Again, if you can show me that this post is wrong, and that the Great Inflation was not caused by printing money, then I’ll just give up. I’ll quit my job, I’ll stop blogging, and I’ll publicly apologize to all my former Bentley students, as it would mean that almost everything I taught them for 30 years was wrong. My entire view of macroeconomics is predicated on monetary policy driving nominal aggregates.
In my view, the “money printing caused the Great Inflation” hypothesis is not a theory in the sense that the man on the street uses the term theory (unproven hypothesis), it’s a theory in the sense that scientists use the term when discussing well established models like evolution.