Scott Sumner  

What (if anything) can we learn from 1921?

Krugman's Cursory Case Against... He Who Wills the End Wills the...

There has a been a recent flurry of blog posts on the 1921 depression--especially why it was so much less severe than the Great Depression. One unfortunate aspect of the dispute is that it has become linked to a misleading ideological debate over the role of government. Thus libertarians are often seen as pointing to the quick recovery from the 1921 depression as an example of why an activist government is not needed. Liberals then respond that the libertarians have mischaracterized what actually happened.

I happen to think the 1921 depression does have a few libertarian-friendly implications, but not necessarily the same ones that others have been pointing to. So first let me explain where I am a bit uncomfortable with the libertarian position (and equally uncomfortable with the Keynesian response.)

Many people wrongly believe that the ideological debate over small government is somehow linked to the debate over stabilization policy. Thus supporters of small government sometimes argue that a free market economy will recovery from a recession on its own, whereas big government proponents like Paul Krugman are equally dismissive of those claims. That's not really a meaningful debate, at least in the context of 1921 and the 1930s. In both cases we had a Federal Reserve conducting monetary policy. The choice was not between more or less monetary policy, but rather more or less expansionary monetary policy. Not more or less, but different.

There is no generally accepted notion of a passive central bank. A central bank that is not changing interest rates at all, may be very actively changing the monetary base, and vice versa. Or the exchange rate. There are an almost infinite number of potential indicators of the stance of monetary policy. Arguing that one central bank is more active than another is like arguing that a captain that sets his steering at ENE is more actively steering the ship than one who sets the wheel at NNE. The policies are simply different. This debate should have nothing to do with ideological disputes over the role of government in spending, regulation and nationalization. (Of course abolishing the Fed might be relevant to the debate, but it isn't relevant for 1921.)

Aggregate demand (NGDP) fell sharply during 1920-21, and then rose sharply over the next few years. Ditto for RGDP. NGDP fell sharply in 1929-30, and then continued to fall sharply all the way up to March 1933. Then it rose sharply until 1937, before falling during 1937-38, and then rising up through WWII. Ditto for real GDP, although the path of RGDP was considerably worse than I would have expected, if AD had been the only problem during the 1930s.

The AS/AD model can explain why the 1929-33 contraction was far worse than 1920-21; NGDP fell for a much longer period, and by a much larger amount. But why was the fall in NGDP so much more severe in 1929-33? That's really complicated; I have a 500-page book coming out next year (hopefully) that explains the decline using a gold standard model. But the basic idea is already there in earlier papers by myself, and people like David Glasner and Clark Johnson. Too much demand for gold.

Because we now have a fiat money system, the explanation for why the 1920-21 decline in NGDP was much shorter and milder than 1929-33 has no implications for current policy disputes.

So far I have been downplaying the idea that we can learn from the 1921 recession. But I do believe there is one message that is consistent with libertarian views---artificial attempts to control wages are very counterproductive. Unfortunately the evidence is complex, but I'll sketch out the basic conclusions here. For a variety of reasons, nominal hourly wages were much more flexible in the 1921 depression than the Great Depression:

1. Post-WWI prices were artificially high, and were expected to decline back to "normal." That made workers more willing to accept wage cuts in 1921 than 1930.

2. President Hoover pressured big companies not to cut wages during the Depression. This pressure had some effect, especially in 1930 and early 1931 (less so afterwards.)

3. An executive decree by FDR forced wages 20% higher in the summer of 1933, and there were additional wage shocks on four other occasions (in the spring of 1934, early 1937, late 1938 and late 1939.)

As a result the decline in NGDP led to a bigger fall in RGDP, and more importantly a much slower recovery, than if wages had been as flexible as in 1921. I believe that with 1921-style wage flexibility, the Great Depression would have ended about 1935, even if NGDP had followed the same path it actually followed.

Now of course Keynesians like Paul Krugman argue that with greater wage flexibility you get more deflation. They don't generally talk about NGDP, but the clear implication of this view is that NGDP would be lower with more wage flexibility. And this is where the 1921/Great Depression comparison is most interesting. The wage cuts of 1921 were some of the most aggressive ever seen in American history, and yet NGDP stopped falling almost at the same moment. In contrast, there were far fewer wage cuts in 1930, and the fall in NGDP kept getting worse and worse, partly due to feedback effects of the Depression itself.

So count me as a moderate of the 1921 debate. My best guess is the following counterfactual: With the same monetary policy and the path of NGDP that actually occurred during the period after 1929, but the Harding-era laissez-faire attitude toward labor markets, the Great Depression would have ended around late 1935. That's a six year depression if we count both the contraction and recovery to near full employment. Using the same method, the actual Great Depression was almost 12 years long (with full recovery in late 1941), but the 1921 depression was only about 2 1/2 years long (say mid-1920 to late-1922). So we went from a 2 1/2 year to a 6-year depression due to much worse monetary policy, and from a 6-year to a 12-year depression due to the bad monetary policy being accompanied by really bad labor market policies. Or at least that's my best guess.

PS. George Selgin explains why indicators like interest rates are unreliable during this period. David Glasner discusses the problem of gold hoarding, and the reasons for post-WWI wage flexibility. Nick Rowe shows why a gold standard has some "level targeting" properties (although I'd argue they are too weak to be reliable, and I'd guess he would as well.) David Henderson discusses a recent post by Bob Murphy.

Comments and Sharing

COMMENTS (15 to date)
George Selgin writes:

"There is no generally accepted notion of a passive central bank."

This is a very important observation, Scott. In principle one can ask how the Fed's stance compares to what might have occurred under this or that alternative regime, but one must have a specific alternative in mind, and the counter-factual is never easily arrived at.

AS writes:

How would a central bank that just doesn't print money or trade bonds not be "passive"?

Matt writes:

Interest post (long time listener, first time caller), but it underlines the contradiction in Progressive Policies, such as promoting long term wage contracts which result in overly sticky wages, price caps and rent control (which acts as a guard against falling rents, since these would be capped at "depression level" rates and unable to rise.) Even the entire notion of preventing deflation seems contradictory.
It is almost like a doctor prescribing constant exertion to combat the flu, since the most prevalent symptom is fatigue. I.e., having identified a disease, the policy response is to prevent the symptoms from every manifesting. It is like a doctor, having prescribed exertion to all his patients, bragging how they are never ill, because even though they drop dead from over exertion, they have never taken sick to their beds.

W. Peden writes:


Then it's a central bank with an active frozen base policy, as well as a discount policy (if it has a discount window) and regulatory requirement policies etc.


I wondered if you'd say something similar to what I said on TheMoneyIllusion. I think we seem to see the episode the same way, but I really like the comparison with the Great Depression.

Kenneth Duda writes:

AS, I think you misunderstood Scott's point. Yes, a CB could conduct more OMO's or fewer, and one could argue that a CB that's not trading is more "passive" than one that's not. Fine.

Scott's point is different. His point is that pundits are making a mistake when they cast the comparison of 1920-21 to other recessions in terms of "more government" or "less government", because the real issue is monetary policy, and with monetary policy, you can't have "more" or "less" monetary policy. The CB has a policy and follows it. If that policy is to not conduct any OMO's and leave the monetary base fixed, that is a policy as much as any other. Even if you abolish the CB, you still have a monetary policy, for example to expand the money supply by digging metal ore out of the ground and shrink the money supply by consuming that metal ore in industrial production or jewelry. If that sounds like a silly monetary policy, then you're not a goldbug.


Andrew_FL writes:

It's worth noting that a return of the private economy to it's trend line at about 1941, was then almost immediately followed by nationalization of about half the economy for the sake of the war effort. After that point, none of the usual macroeconomic indicators are meaningful according to their conventional interpretation-not the unemployment rate, in light of an unprecedented movement of young men from the labor force to the armed forces, not to mention war industries employment with draft exemption; not GDP, a very large portion of which consisted of defense goods at government, not market prices, and very little of which consisted of private investment; not inflation indices, which were artificially kept level by such extensive price controls as have not been seen before or since.

In my view, this was another Depression, directly caused by the war, and hidden by it at the same time. So the recovery from the Great Depression was, in fact, brief. And if we count this all as one nearly continuous economic crisis it would be 18 years long (to 1947) rather than 12.

Also, George's post is brilliant, as usual.

Larry writes:

I've always thought of "activist" monetary policy as meaning that the bank changes policy in response to events. Passive means "steady state", in which case things like fiscal policy can affect nominal variables.

CrisisMaven writes:

"... with 1921-style wage flexibility, the Great Depression would have ended about 1935 ..." Well, this goes for ALL prices - and FDR had a penchant to regulate everything he could "put a figure on". Though regulation is now en vogue again in the banking etc. sector, it generally presupposes some bureaucrat to know something "we don't". Never do we get an explanation how that happens. Were markets free to decide about prices, they would also make short shrift of unreliable or over-leveraged banks. But the government "insures" them and then in turn "regulates" them. That said you are being imprecise about "deflation" - PRICEs sinking or monetary contraction? When prices sink it can be for any (good!) reason: less demand is the main reason, as is rising supply. Manipulating prices for fear of "deflation" blunts that indicator and leads to ... recessions or booms. There is no truth in the argument that "deflation" must be avoided because people will not buy, but wait for the future. Those obese citizens speak louder than words that they do buys, even if prices get lower ... Equally "inflation", if there were any truth to the first argument, would then lead to hold off selling - because they wait for higher prices. And wait some more. And wait. And wait. All nonsense.

Scott Sumner writes:

AS, Consider the monetary base from August 2007 to May 2008, which was roughly unchanged. Was policy passive? In one sense yes, but in lots of other senses no.

1. The Fed repeatedly cut the fed funds target during that period.

2. The growth rate of the MB slowed sharply from the rate during the previous decade.

In those senses policy was active.

Thanks W. Peden.

Ken, That's right.

Andrew, I agree that there was a sort of "consumption depression" in the early 1940s, but it's better to use language in a very precise way. Thus I would not call the early 1940s a depression, I'd call it a bad period for the economy and living standards.

Larry, It depends what you mean by "policy" And "changes policy".

Travis Allison writes:

Is PK's argument that wage cuts are contractionary only when we are at the ZLB?

The 1930s episodes might be more of a better test of PK's proposition about wages. If a wage cut is contractionary at the ZLB, shouldn't a wage raise be expansionary?

Ray Lopez writes:

I have James Grant's book on 1921 and will read it with these comments in mind, but frankly this blog post was more on 1929 than 1921, except to say '1929 was bigger than 1921'.

This caught my eye, and I hope Dr. Sumner reads this: "I have a 500-page book coming out next year (hopefully) that explains the decline using a gold standard model." (I hope this is not just a rehash of arguments in Eichengreen's Gold Fetters).

RICHARD N. COOPER, Harvard University, "The Gold Standard: Historical Facts and Future Prospects" (written in a neutral tone, awesome summary, shows that even back in the Gold Standard the central banks practiced tricks as done today, such as sterilizing gold inflows, so essentially gold standard becomes modern fiat money management)

Blurb from Cooper article, 56 pp long: "How did this system [classic Gold Standard] fare in terms of economic performance? The idealized gold standard as it appears in textbooks conveys a sense of automaticity and stability-a self-correcting mechanism with minimum human intervention, which assured rough stability of prices and balance in international payments. The actual gold standard could hardly have been further from this representation"

Andrew_FL writes:

@Scott-Actually, primarily the trough is in private investment, rather than consumption.

Nevertheless I understand your reluctance to use language in a non standard way. I just don't think the standard way makes sense, personally, in this case. But, that's just me I guess.

Michael Byrnes writes:


I think Scott's point is that in marked contrast to the Great Depression, the negative impact of WWII was not in any conceivable way a business cycle phenomenon. I think WWII would be better thought of as a real shock and for most people a decline in real income (imposed via rationing/shortages).

Andrew_FL writes:

@Michael Byrnes-Ah, that makes sense. And I agree, it's certainly not a "business cycle" so much as it is an event. And it was a shock to real output, not nominal spending. All of these set it apart from pretty much every business cycle since then.

If that's what Scott meant, and it seems probable to me that he did, then I pretty much agree completely.

J.V. Dubois writes:

Scott: I think this is an interesting debate, especially the difference between Krugman and you in how cutting wages impacts AD. Krugman explains his position in this paper:

I have a beef with this part:

"Our model suggests, however, that when the economy is faced by a large deleveraging shock, increased price flexibility – which we can represent as a steeper aggregate supply curve – actually makes things worse, not better"

Should we not represent perfectly flexible AS by a horizontal curve? This is how LRAS curve is usually drawn - any changes in AD affect only price level. Or to present a counterfactual - if wages would be fixed, would it mean that it would make SRAS curve horizontal in Krugman eyes? This seems fishy to me.

How does Krugman save this? Fisherian debt deflation. But I don't think this is quite right. Sure, AD shock can cause worse debt situation but that is self-explanatory. AD shock (lower incomes) causes worse income-to-debt ratio. Yes, but again thinking about the counterexample how would fixing wages help in this regard? We would end up with lot of unemployed people which would leave us with very bad income-to-debt ratio anyways.

PS: I think that welfare implications of monetary policy is a serious issue (how the risk of nominal debt is divided between debtor and creditor). That is why I prefer NGDP level targeting as nicely explained by David Eagle over at Lars Christensen's blog:

However this should not get in the way into business cycle stabilization role.

Comments for this entry have been closed
Return to top