Another misleading argument by Cole and Ohanian

I agree with Cole and Ohanian that the NIRA aborted a promising recovery after July 1933.  I disagree with Paul Krugman on this issue.  And unlike most Keynesians, I don’t think the recovery from the Great Depression under FDR was very impressive.  Much of the recovery was due to productivity growth (until 1941.)

And yet I find myself once again to be very irritated by an argument against the demand-side view put forward by Cole and Ohanian:

The main point of our op-ed, as well as our earlier work, is that most of the increase in per-capita output that occurred after 1933 was due to higher productivity – not higher labor input. The figure [at the link] shows total hours worked per adult for the 1930s. There is little recovery in labor, as hours are about 27 percent down in 1933 relative to 1929, and remain about 21 percent down in 1939. But increasing aggregate demand is supposed to increase output by increasing labor, not by increasing productivity, which is typically considered to be outside the scope of short-run spending/monetary policies.

I originally read this quotation over at MR, and immediately thought; “When has a Keynesian ever argued that there was a robust demand-side recovery from 1933 to 1939?”  I’ve read just about everything ever written on the subject, and I’ve never heard that argument made.  Instead, Keynesians argue that demand stimulus led to a fast recovery during 1933-37, and then tight monetary and fiscal policies caused a severe relapse in 1938.  So why would Cole and Ohanian pick those dates?

As soon as I clicked over to the Stephen Wiliamson post where Tyler found the argument, I immediately knew the answer.  Cole and Ohanian present a graph that strongly supports the AD view of the recovery from the Great Depression.  Hours worked went from being 27% below normal in 1933, to only 17% below normal in 1937, the cyclical peak.  That means an extra 2.5% per year.  Using Okun’s Law, I’d guess that gets you about 5% RGDP growth per year.  Now the actual rates were substantially higher during 1933-37, as productivity also grew briskly.  But the hours worked finding basically follows the predictions of AD models.  Even Keynesians believe the economy was still far from full employment in 1937.

Then hours worked plunged between 1937 and 1939, in response to the sharp fall in AD (as measured by NGDP) during 1938.  Again, this is perfectly consistent with demand-side explanations of the 1930s.  Indeed it’s the standard view.   BTW, I happen to think a massive adverse supply-shock also reduced hours worked and output during 1938, so my position is actually intermediate between C&O and the Keynesians.  Looking at the entire period from 1929 to 1939, the blue line (hours worked) is highly correlated with changes in AD (i.e. NGDP.)

I think aggregate supply mattered a lot in the Great Depression.  But none of the data presented by C&O refutes the argument that demand played a major role in the Depression, indeed it strongly supports that view.

PS.  I’d be interested in whether the C&O data include hours worked on government jobs programs.  Official government unemployment data from that period is highly inaccurate, as they treat millions of WPA/CCC workers as “unemployed.”

PPS.  In case anyone wonders why I view the 1933-37 recovery as disappointing, despite high RGDP growth rates, consider that industrial production grew 57% between March and July 1933, due to dollar devaluation.  Then FDR raised nominal wages by 20% in late July, as part of the NIRA.  Monthly industrial production data fell immediately, and didn’t regain July 1933 levels until after the NIRA was declared unconstitutional in May 1935.  This led to rapid growth in late 1935.  Because of the way annual GDP data averages over entire years, the RGDP growth from 1933-35 looks deceptively steady and impressive.  It wasn’t.

Update:  I just noticed that Matt Yglesias is just as puzzled as I am by their chart.



32 Responses to “Another misleading argument by Cole and Ohanian”

  1. Gravatar of David Pearson David Pearson
    29. September 2011 at 07:55


    I have often thought that, had you been magically transported back to 1935, you would have been a major critic of monetary policy at the time. The reason is FDR and the Fed allowed a huge NGDP level short fall from trend to occur during the 1933-1936 recovery. Even on a trajectory basis, they would not achieve a level target for multiple years. For that reason, you might have argued that efforts to raise NGDP, while welcome, were woefully short of what was necessary.

    Which begs the question, what is the relationship between an NGDP shortfall and a real recovery? Does a persistent shortfall imply a newly-recessionary or stagnant real economy? Or just one that grows but not at trend? Obviously, 1930’s NGDP growth was above trend, just not at a trajectory that made a level target achievable within a reasonable time frame. Nevertheless, what does that above-trend growth tell us about the meaningfulness of the NGDP level shortfall construct?

  2. Gravatar of David Pearson David Pearson
    29. September 2011 at 08:10

    BTW, by the same token, Japan’s real per capita growth, while not stellar, is arguably at trend given the country’s aging population and well-known structural issues. Certainly, the country’s employment experience is far superior to ours. How could the economy be growing at anything close trend with such a large, persistent NGDP “gap”? How would economists even determine what “trend” is after such an important demographic shift and persistent deviation from the historical trajectory? Is the concept of an “AD shortfall” supposed to explain such a long-run phenomenon?

  3. Gravatar of Scott Sumner Scott Sumner
    29. September 2011 at 08:51

    David, Wages fell sharply between 1929-33, and hence I don’t think we needed to get back to 1929 NGDP. Fisher suggested going half way back in price terms, as I recall. In my view there was plenty of NGDP for a fast recovery, and the reason the Depression didn’t end quickly (like 1921) was the NIRA.

    In my view the natural rate of unemployment in Japan is about 2%, Because of near-zero NGDP growth since 1994 the actual unemployment rate is more like 4% or 5%. That produces a one time reduction in RGDP, but doesn’t affect the long term growth rate–so in that sense I agree with you. But I still think moving up to 2% or 3% trend NGDP growth would give them a one-time boost in jobs, and that this would be quite desirable.

  4. Gravatar of David Pearson David Pearson
    29. September 2011 at 09:40


    I thought NIRA raised nominal wages. That is, by 1935 nominal wages had recovered much of their previous fall. Thus, the NGDP-adjusted nominal wage was still way above its 1929 level in 1935. Maybe this is your point: NIRA was at fault for high unemployment. However, why not equally blame the Fed for “allowing” the NGDP-adjusted wage to be so high? Wasn’t it under their control? When it comes to real wages, what government policy does, the Fed can undo.

  5. Gravatar of David Pearson David Pearson
    29. September 2011 at 09:44

    The above reminds me of today’s debate. The “structuralists” argue part of the reason behind high UE is government-induced uncertainty about total future employment costs (i.e. because of the new health care law). You reply to them, “you have a point, but its still the Fed’s fault for keeping NGDP too low.”

  6. Gravatar of Morgan Warstler Morgan Warstler
    29. September 2011 at 10:20

    Minor Annoying Point:

    WHY don’t all of you: David, Stephen, DeKrugam, Matty all make a constant and immediate link to:

    It drives me crazy.


    “Per capita real GDP was about 27% below trend in 1939, with more than three-quarters of this shortfall due to the continuing depression in labor. Our research indicates that New Deal industrial and labor policies, such as the National Industrial Recovery Act and the Wagner Act (the National Labor Relations Act), were the main reasons. The NIRA, for example, fostered monopoly and raised wages well above underlying worker productivity by a quid pro quo arrangement of relaxing antitrust enforcement in exchange for industry paying substantially higher wages.

    The Wagner Act substantially increased unionization and union power. This, in conjunction with government’s toleration of sit-down strikes, in which union workers forcibly seized factories to stop production, increased wages further.

    In the absence of these policies, we estimate that labor input would have been about 20% higher than it was at the end of the 1930s and would have returned the economy to trend by that time.”

    The SAME thing would happen with my Guaranteed Income plan – which Scott supports.

    EVERYONE would get a check from the gvt. but they would auctioned off at WHATEVER rate made their labor PROFITABLE to the private sector.

    This would:

    1) reduce prices on all kinds of things like daycare, or yard work / remodeling in poor neighborhoods. Increasing consumption.

    2) increase productivity, the human that Scott wants to talk to would speak English as a first language – because all the telemarketers would be mommies working from home, instead of Indians up all night in New Delhi.

    small fun fact: they teach the girls there to speak with a southern drawl because it best covers up their accent. one of the most surreal experiences of my life (and ther ehave been many) was watching an Englishmen teach a roomful of Indian ladies to to say ya’ll.

    3) Increase hours worked.

  7. Gravatar of Charlie Charlie
    29. September 2011 at 11:50

    “I’d be interested in whether the C&O data include hours worked on government jobs programs. Official government unemployment data from that period is highly inaccurate, as they treat millions of WPA/CCC workers as “unemployed.””

    It’s a minor point, but I think “officially” they fixed this in the latest Historical Statistics of the U.S.

  8. Gravatar of Charlie Charlie
    29. September 2011 at 11:51

    Obviously, I don’t no what C&O used.

  9. Gravatar of Bob Murphy Bob Murphy
    29. September 2011 at 13:37

    Scott wrote:

    Cole and Ohanian present a graph that strongly supports the AD view of the recovery from the Great Depression. Hours worked went from being 27% below normal in 1933, to only 17% below normal in 1937, the cyclical peak.

    Scott, I think you are here engaging in the same “very misleading” trick that O&C did in their original op ed, for which Krugman, you, and Glasner sent them to the woodshed. In particular, you are giving us endpoints, when if we look at the graph for the entire period, we see the opposite story.

    (BTW sorry if I’m missing the whole debate here, but I think this is right…)

    O&C in their latest graph are trying to show that wages and total labor hours move in opposite directions, right? So that fits their story that unemployment was because of New Deal policies that drove wage rates above market-clearing levels. (As Krugman just recently reminded us, Keynesians do NOT blame unemployment during the 1930s on wages being too high.)

    That view is supported not only by the endpoints of their chart, but in the middle too. Yes, total hours per worker went up, but not when the gold devaluation occurred. Instead it started in 1934, and that was a period when wages were *falling* not rising. From 1933-1934, wage rates spiked (which I guess shows that a devaluation can turn around prices and wages, just like you tell us), but that period also showed total hours per person continuing its downward trend.

    And when you get the collapse in hours per worker starting in 1937, that also goes hand in hand with a huge jump in wages.

    I agree O&C had some misleading stats in their original piece, but I don’t see how your critique here to this later chart works.

  10. Gravatar of happyjuggler0 happyjuggler0
    29. September 2011 at 14:33


    Forgive me for asking a question that I suspect you answered a long time ago, but what in your opinion finally ended the Great Depression? Was it WWII spending or monetary policy changes or what?

    P.S. If you did address this elsewhere in a post, perhaps you might add that to your “important links” page. I can’t be the only one who wants to know what you think on this issue.

  11. Gravatar of Tommy Dorsett Tommy Dorsett
    29. September 2011 at 15:29

    That NGDP and RGDP were nearly perfectly correlated between 1929-1939 pretty much blows away the O&C argument that AD wasn’t an important driver of that cycle.

  12. Gravatar of Scott Sumner Scott Sumner
    29. September 2011 at 16:44

    David, No, it is not my argument that The Fed should try to undo the damage from things like the health care law. I think the Fed should focus on keeping NGDP growing at a constant rate, regardless of the supply-side of the economy. If the government is determined to screw up the economy, the Fed cannot and should not try to stop them–it would merely lead to an inflationary spiral.

    Morgan, Why link to a misleading article.

    Charlie, Thanks for the info.

    Bob, I have no problem with their argument about wages, or their claim that Krugman is wrong. I also think Krugman is wrong. I do have a problem when they present misleading data to try to suggest that the evidence is inconsistent with the AD-view of the Depression. In fact employment and AD are highly correlated throughout the Depression, contrary to what they claim. They are attacking an argument that no Keynesian is making. But they imply Keynesians are making that argument.

    Their chart has nothing to say about the effect of gold devaluation on hours worked. Nothing. I very much doubt that industrial output could rise 57% in 4 months with no increase in hours–that would be the fastest productivity growth in history.

    happyjuggler, That’s a complicated question. It depends how you define monetary policy. If you define monetary stimulus as rising NGDP, then one might argue that monetary stimulus did the job. On the other hand perhaps NGDP rose because of wartime spending in 1940-41. More pro-business policies might have also helped a bit.

    Tommy, That’s right, which is why it bugs me that C&O try to suggest otherwise. Bad AS policies explain the slow recovery, but the Depression itself was caused by a big fall in AD. (And then a smaller fall in 1937.)

  13. Gravatar of Bob Murphy Bob Murphy
    29. September 2011 at 19:20

    Scott Sumner wrote:

    Their chart has nothing to say about the effect of gold devaluation on hours worked. Nothing.

    What do you mean by that? The chart (I think) shows that hours per adult continued to fall, even after the gold devaluation occurred. Are you saying I’m reading the chart wrong, or that no matter what the chart showed, we wouldn’t be able to conclude anything about the effect of an event happening at a certain point in the chart?

  14. Gravatar of Eric Morey Eric Morey
    29. September 2011 at 19:58

    I thought that you would like to know that a Google search for “Cole and Ohanian” returned the following top 3 results

    How Government Prolonged the Depression – – Similar – Add to iGoogle
    You +1’d this publicly. Undo
    Feb 2, 2009 – Policies that decreased competition in product and labor markets were especially destructive, write Harold L. Cole and Lee E. Ohanian.
    The Great Depression in the United States From A Neoclassical ……rep…
    You +1’d this publicly. Undo
    File Format: PDF/Adobe Acrobat – Quick View
    by HL Cole – 1999 – Cited by 216 – Related articles
    23, no. 1, pp. 2-24. The Great Depression in the United States. From A Neoclassical Perspective. Harold L. Cole. Lee E. Ohanian. Senior Economist. Economist …
    Another misleading argument by Cole and Ohanian
    You +1’d this publicly. Undo
    12 hours ago – I agree with Cole and Ohanian that the NIRA aborted a promising recovery … against the demand-side view put forward by Cole and Ohanian: …”

  15. Gravatar of Benjamin Cole Benjamin Cole
    29. September 2011 at 21:03

    Well, I suppose re-hashing the Great Depression through modern-day ideological lenses is inevitable. I am glad Cole/Ohanian were exposed as near-frauds. (No relation btw).

    But the example market monetarists should cite is Japan. That way we avoid the ancient battlegrounds and intellectual tugs-of-war over the Great Depression. Which neither side will ever “win” as partisan politics befogs everything.

    Market monetarists should talk about today, and Japan. Crickey-Almighty, there is plenty to talk about.

  16. Gravatar of Morgan Warstler Morgan Warstler
    29. September 2011 at 22:48

    “Morgan, Why link to a misleading article.”

    Um, so your readers don’t suspect you are afraid of linking to it.

    No one cares about your “expertise” or your “curation” – you take your idea out of your pants and you flop it on the table and say match this…

    If you don’t want it laying right next to somebody else, you sound like a guy who needs low lights to be comfy.

    They weren’t selling Pol Pot memorabilia. And you could use future links from the WSJ. Google LIKES direct reference to blue chip properties.


    Anyway, pls note you have already AGREED with their findings… it uses the same logic as my Guaranteed Income.

  17. Gravatar of Gabe Gabe
    30. September 2011 at 06:47

    I have a great idea!…lets pay people not to work at all and if they get a part-time job or entry level job to gain experience in a new field and make the 1st step towards independence then we stop all payments to them…this way we can provide extra incentives not to try some new career.

  18. Gravatar of JimP JimP
    30. September 2011 at 07:13

    Martin Wolf – support demand:

    And will Bernanke do this?

    No – because he might just get himself impeached by the wonderful Republican deflationists.

    Of course – real actual tax reform – fewer deductions and lower rates, and eliminate the corporate tax – along with credible long term spending reform – now that might well do it – with a little cooperation from the Fed.

    But Obama and the Dems are not interested

  19. Gravatar of JimP JimP
    30. September 2011 at 07:14

    Its a pincer operation

    Moron Republicans vs moron Democrats. And we are in the middle.

  20. Gravatar of John John
    30. September 2011 at 12:10


    I’d be interested to hear your response to graph Ohanian presents around 6:00 in this video

    The graph shows M1, M2, and manufacturing hours from 1929-Fall of 1930. During this time period, M1 and M2 were fairly consistent with 1929 values while manufacturing output collapsed. All this happened before the first banking panic. Money supply had nothing to do with the collapse in output! Today we see that the CPI is higher than it was in 2007 or 2008 while manufacturing output hasn’t recovered. The data points to a supply shock, not a fall in AD.

    Btw, I studied at UCLA and took several classes from Ohanian, so I’d naturally side with him on this one.

  21. Gravatar of More evidence on how Cole & Ohanian were deceptive | Historinhas More evidence on how Cole & Ohanian were deceptive | Historinhas
    30. September 2011 at 13:54

    […] a recent post Scott Sumner writes: “And yet I find myself once again to be very irritated by an argument against the […]

  22. Gravatar of Steve Steve
    30. September 2011 at 19:13


    I don’t have time to listen to the entire Ohanian talk, but in the part you reference he seems to be blaming “Labor Market Failure” starting in late 1929 as the cause of the Great Depression. It wasn’t AD shortfall or money supply decline. He goes on to explain the “Labor Market Failure” was rising REAL wages starting in late 1929 (as other nominal prices started to fall). He blames price collusion and Hoover’s support for trade association for the rising real wages.

    It sounds like he’s arguing that sticky wages during a deflation caused the depression but he refuses to acknowledge sticky wages as a legitimate economic concept. Instead he calls sticky wages “Labor Market Failure” and blames regulation for exacerbating the stickiness.

    On your supply shock question: I think there was a supply shock in 2009, but it was caused by the tight money and the subsequent financial crisis. As an example I present into evidence the rig count, which fell almost 60 percent during the financial crisis. This lowered AS, i.e., led to oil shortages early in the recovery at levels of AD lower than the previous peak.

    I also think that sticky prices are creating resource misallocation now (but Ohanian would call it “Labor Market Failure” because labor or regulation is a priori always at fault).

    Paul Krugman linked to an interesting paper on PLOG (Prolonged Large Output Gaps) — in multiple case studies, inflation tends to fizzle at small positive numbers even as the economy shrinks.

    I look forward to hearing Scott’s reaction to your question and my observations. TIA, Scott.

  23. Gravatar of Scott Sumner Scott Sumner
    1. October 2011 at 16:24

    Bob, No, the chart doesn’t show what he claims, because it uses annual data, which is very misleading. Hours soared after the dollar devaluation–as industrial production rose 57% between March and July 1933. Then hours fell sharply after wages were raised 20% in late summer–this is why hours fell slightly in 1934. The effect of devaluation on hours was strongly positive, and the effect of the NIRA was strongly negative.

    Thanks Eric, I’m number three!

    Ben, Yes, Japan is what we should be worried about.

    Morgan, There are billions of articles I don’t link to, do you think I am afraid of all of them?

    Gabe, Yes, it’s called unemployment comp.

    JimP, Yes, that’s the problem.

    John, Money was extremely tight in late 1929 and early 1930. Between October 1929 and October 1930 the monetary base fell at one of the sharpest rates of the whole 20th century. If you are going to talk about the money supply that’s the one to pay attention to. But as you may know I think all monetary aggregates, even the base, are unreliable indicators of policy. We know money was tight because NGDP plunged in 1929-30.

    Steve, I mostly agree. However I’d say the biggest negative supply shocks in 2009 were not in oil regulation, but higher minimum wages and extended UI benefits.

  24. Gravatar of Mark A. Sadowski Mark A. Sadowski
    1. October 2011 at 20:43

    I’ll be as clear in my staked position as I can be Scott:

    Cole and Ohanian are absolutely wrong. Welcome back to the fight.

  25. Gravatar of sdfc sdfc
    2. October 2011 at 02:10

    Unemployment benefits supported household income. Lower household income would mean lower spending and lower investment. Let’s not forget household debt to GDP was not far south of 100% in 2009. Not extending benefits would have been a disaster.

    Do you really believe that cutting benefits would promote growth and that job openings would magically appear?

    By the way according to the BEA NGDP rose 17% in 1934, 11% in 35 and 14% in 36. Don’t expect much do you.

  26. Gravatar of Morgan Warstler Morgan Warstler
    2. October 2011 at 12:22


    if it is the main source of a discussion, no matter how much a a circle jerk is going on, directly linking the article is the only “not afraid” thing to do.

  27. Gravatar of John John
    2. October 2011 at 13:59

    “We know money was tight because NGDP plunged in 1929-30.”

    That sounds so incredibly crazy. What if the government had set the minimum wage to $100 an hour? What if they had made it illegal to produce, distribute, or use oil? NGDP would collapse and it would have nothing to do with money being tight or not. This is basically what Cole & Ohanian are arguing and you’re saying it only comes down to money. It’s absurb. Money is a medium of exchange not the fountainhead of economic growth.

  28. Gravatar of Scott Sumner Scott Sumner
    3. October 2011 at 15:43

    Thanks Mark.

    sdfc, Don’t put words in my mouth. I never said it was a mistake to extend UI in 2009.

    I favor higher NGDP as a solution to unemployment.

    You obviously don’t understand my argument on the Depression, as your NGDP data actually supports my argument.

    Morgan, I link to the articles I discuss.

    John. Supply shocks don’t directly affect NGDP. It’s possible that they might indirectly cause monetary policy to tighten, but that certainly wasn’t a problem in the 1970s, when NGDP growth soared.

  29. Gravatar of Scott Sumner Scott Sumner
    3. October 2011 at 15:46

    sdfc, Do you really think France’s unemployment rate “magically” rose from 3% to 10% and stayed there for 30 years, or do you think there might have been a reason?

  30. Gravatar of Dave Thomas Dave Thomas
    25. November 2012 at 21:17

    I came across this thread a year late while doing some economic history research, and something struck me reading this thread.

    How does this NGDP macroeconomic mindset deal with the economic growth spurred by marginal tax cuts. The first historical example came from Andrew Mellon’s marginal tax cuts following the First World War. The second example was LBJ’s passage of the Kennedy tax cuts inspired by Robert Mundell. Finally the Reagan marginal tax cuts coupled with Volcker’s sound money interest rate policy.

  31. Gravatar of Saturos Saturos
    25. November 2012 at 21:31

    Dave, we allow disinflation to accomodate the RGDP boom. We don’t deviate from the NGDP level target. Supply side issues are important too, more important in the long run. But right now NGDP matters more – a lot of the supply-side problems in the US right now are the consequence of demand-side problems, of tight money.

  32. Gravatar of Let’s Talk About: Economic Growth | Guy Debord's Cat Let’s Talk About: Economic Growth | Guy Debord's Cat
    13. March 2015 at 05:34

    […] and Ohanian are comprehensively defenestrated in this blog. Hannan isn’t interested in reality and like all right-wingers of his ilk, he exists in the […]

Leave a Reply