Cross-posted at the Angry Bear blog.
I was searching for some information and I stumbled on a post Scutt Sumner wrote last year about Robert Skidelsky’s biography of John Mayndard Keynes. I haven’t read Skidelsky’s book, nor do I know Skidelsky, and its been awful long time since I read Keynes, but this seems an odd complaint:
I’m afraid that his analysis is both misleading and inaccurate. The US gradually depreciated the dollar between April 1933 and February 1934. During that period unemployment was nearly 25% and T-bill yields were close to zero. Keynes argued that monetary stimulus would not be effective under those circumstances, and Skidelsky seems to accept his interpretation (which was published in the NYT during December 1933.)
[Note that Keynes certainly did believe in the "pushing on a string" theory--I frequently get commenters insisting that Keynes didn't believe in liquidity traps.]
Unfortunately, Keynes and Skidelsky are wrong. The US Wholesale Price Index rose by more than 20% between March 1933 and March 1934. In the Keynesian model that’s not supposed to happen. The broader “Cost of Living” rose about 10%. Industrial production rose more than 45%.
Sumner goes on to impugn Skidelsky:
The “disappointing” results that Skidelsky mentions come from cherry-picking a few misleading data points.
All that seems very odd to me. If I were making an argument that conventional monetary policy doesn’t work in a liquidity trap, but that the traditional Keynesian prescription does, I’d start that argument with something very much like the sentences Sumner wrote right after stating “Unfortunately, Keynes and Skidelsky are wrong.”
(Note – I can imagine a “monetary” prescription that I think would help tremendously in a liquidity trap, but it doesn’t look at all like what was done in the 1930s, or what was done since 2007, or from what I can tell, what Sumner suggests. That can be a post for another time.)
Using the graphing tool from FRED, the Federal Reserve Economic Database maintained by the St. Louis Fed, we can show the one year percentage change in both PPI (producer price index) and CPI (consumer price index) from January 1932 to December 1935.
Here’s what we see: after some massive deflation during the Great Depression, prices start to rise more or less when FDR took office. The annual percentage change in PPI peaked around 23% and change in February 1934, and the CPI peaked a few months later at about 5.6%.
Elsewhere, Sumner attributes that to:
We all know what happened next (well not exactly, but I’ll explain that in another post), so let’s jump ahead to 1933. FDR takes office in March, promising to boost wholesale prices back up to pre-Depression levels. He uses several tools, but the most effective was loosely based on Irving Fisher’s “compensated dollar plan.” Fisher’s plan was to raise the price of gold one percent each time the price level fell one percent. An obscure agricultural economist named George Warren was a big fan of Fisher’s idea, and sold it to FDR with all sorts of fancy charts.
And it worked.
Initially it worked better than any other macroeconomic policy in American history. But at first the policy’s success was mostly accidental, just a matter of talking the dollar down, not enacting Fisher’s specific plan. Nevertheless, prices immediately began rising sharply. Industrial production rose 57% between March and July, regaining over half the ground lost in the previous 3 1/2 years. Then in late July FDR decided to cartelize the economy and sharply raised wages (the NIRA) and industrial output immediately began falling. By late October FDR was desperate for another dose of inflation, and asked Warren to come up with a plan. They decided to have the US government buy gold at a price that would be continually increased in order to reflate the price level.
Sumner even helpfully tells us:
It was a very confusing plan, as they never bought enough gold to equate the government buying price with the free market price in London.
I agree that what Sumner describes is confusing. And yes, the times were desperate, and FDR was flailing around throwing all sorts of things against a wall to see what would work, but when I look at the graph above, and take into account the extremely rapid economic growth that took place during the New Deal era, I see a much simpler story.
1. Aggregate demand was very slack when FDR took office..
2. FDR showed up in Washington with a plan to start spending a lot of money and thus boost aggregate demand.
3. The immediate effect was to convince factories they’d be running down their inventories. That boosted producer prices. It had a much smaller effect on consumer prices because everyone knew the gubmint was going to buy a heck of a lot more producer goods than consumer goods. (The government did buy some consumer goods for the various programs, plus there was a spillover effect, but as the graph clearly shows, the action was on the producer side.)
4. After a bit of time, the public realized FDR wasn’t planning just a one-off, but rather a sustained program of purchases of industrial items. That led them to start using some of their idle capacity, which meant not just selling the fixed amount that was in inventory. The rate of price increases thus dropped.
5. GDP increased at the fastest rate in the United States peacetime history since data has been kept. There was a big hiccup, of course, in 1937 when the government cut back on spending for a while.
By contrast, here’s Sumner explaining his theory:
There is a great deal of evidence that I won’t get into here that suggests the suspension of the gold standard in March 1933, and gradual devaluation between April and February 1934, almost certainly explain most of the increase in goods prices, stock prices, and industrial production during that period. But why? Not because it boosted our trade balance, which actually worsened as the rapid recovery pulled in imports.
Both Gauti and I believe that only the rational expectations hypothesis can explain these events. He focuses on how the regime change led to higher inflation expectations, and thus reduced real interest rates. I prefer to think in terms of specific policy signals sent as rising gold prices changed the future expected gold price, and hence the future expected money supply. I don’t see any non-Ratex explanation that can account for the extraordinary rise in prices and output during March-July 1933. Nominal interest rates didn’t change much, and open market purchases in 1932 (under the constraint of the gold standard) had accomplished little or nothing.
So…. his story requires the devaluation of the currency to worsen the trade balance, and rational expectations to cause a one time explosion in industrial prices and a rather smaller recovery in consumer prices. Rational expectations, however, that came an abrupt halt, at roughly the same amount of time one would predict companies might decide that demand will be sustained enough to start producing more rather than just selling off inventory sitting in warehouses. And his story doesn’t explain why growth was so much faster during the New Deal era than any other period of peacetime since the US began keeping data, nor why there was the big hiccup in 1937.
Sumner is essentially trying to tell a story about an unusual set of events, but his story seems to assume that most extraordinary events of the era (and what sets that era apart) kind of just happened to occur for no particular reason so he misses the big picture and ends up focusing on details. With all due respect to Sumner, I prefer to think the US economy is not Forrest Gump.