Mises Daily Articles
Charting Fun with Krugman
In a few recent blog posts, Paul Krugman used bar graphs and tables to (allegedly) prove the superiority of his views over those of the Austrians. Yet, as I'll show in this article, I can use Krugman's own data to demonstrate the exact opposite.
Krugman on the Fed and Banking Panics
Perhaps spurred by his Bloomberg debate with Ron Paul, Krugman posted the following regarding financial panics and the US central bank:
There's a very widespread belief on the right that banking crises only happen because either the Fed or Barney Frank cause them; go back to a gold standard, and there would be no need for financial regulation or anything like that.
This is, of course, nonsense; Walter Bagehot knew all about financial crises, which have been a constant feature of modern economies since at least the early 19th century. Just to drive the point home, I thought it might be worth posting Gary Gorton's chart (see pdf) of "panics" before the Fed went into operation:
Panics will happen; the question is how they are contained. (emphasis added)
Now although Krugman doesn't explicitly say "Ron Paul" or "Austrian economists," I think he has to have them in mind here. After all, before the Austrians rose in popularity, hardly anybody talked about the gold standard, let alone abolishing the central bank. It was the Austrians, and most notably Ron Paul, who put those ideas back into the limelight so that Paul Krugman feels the need to address the issue.
In that light, Krugman is simply making stuff up when he says such people think banking panics never happened before the Fed. Murray Rothbard's doctoral dissertation was The Panic of 1819, and Rothbard also wrote on the history of the Fed, so I'm pretty sure he wouldn't be shocked by Krugman's table.
But besides the cheap debating ploy — setting up his opponents as believing something obviously ridiculous — Krugman leaves open the door to his own demise in his final sentence, after the chart, when he writes, "Panics will happen; the question is how they are contained."
Fortunately, Krugman himself provides the answer in a follow-up post on the very same day. In this new post he writes,
A followup on this post. We had frequent banking panics before there was a Fed; how bad were they?
Well, Christy Romer has an old paper (see PDF) on long-run volatility, in which she created a metric: percentage-point months of industrial production lost until previous peak was regained:
So some of those pre-Fed panics were worse than the big slumps of 1974 and 1981, although far short of Great Depression stuff.
By my estimate, the current number using industrial production data is 455; it will get a bit bigger but not much if the recovery continues.
So we're doing worse than after, say, the Panic of 1907 — but not that much worse.
How do you like that? By Krugman's own admission, the two worst panics occurred after the Fed was formed. And if we take Romer's numbers from the table above, and plug in a decline of 455 for the most recent recession (which Krugman himself says will be an understatement), we get that the average "output loss" (measured in the units Romer defines in the chart) during recessions from the pre-Fed era was 158.1, while in the post-Fed era it was 356.4.
Does everyone see the significance of this? Krugman himself said that panics will always happen, and the question is how they are contained. Using Krugman's own source, we find that the establishment of the Fed generated (a) the two worst panics in US history and (b) a string of panics that were on average more than twice as bad as the average panic from the pre-Fed era.
Steve Horwitz does a good job explaining why Krugman's understanding of US banking history is flawed, because we didn't have laissez-faire banking in the late 1800s. But we don't even have to rely on such explanations for the matter at hand. Remember, these data weren't pulled from Krugman after a session of waterboarding. He volunteered them as if they were somehow supposed to embarrass the critics of the Fed. What would the numbers have to look like, for Krugman to have admitted, "Hmm, it seems like for once, empirical reality has turned against my Keynesian nostrums"? Would the post-Fed panics have to be three times as bad?
One last thing before I leave this topic, because I just want to make sure everyone sees what a rhetorical illusionist Krugman is. After he ballparks the current number (using Romer's metric for output loss) at 455, he writes, "So we're doing worse than after, say, the Panic of 1907 — but not that much worse."
That word "say" in his quotation makes it sound as if Krugman threw a dart at Romer's left column, and picked the panic that happened to come up. But no, the Panic of 1907 is the worst one of the pre-Fed era shown in Romer's list. With as much justification, Krugman could have written the following summation instead: "So the second-worst recession of the Fed era — which we're experiencing right now even though the guy I said there's nobody I'd rather be Fed chief has been at the helm the whole time — is already worse than the absolute worst episode from the pre-Fed era. But hey, as long as this recovery continues — and I often warn that we are on the verge of disaster if Republicans take over — we're not doing that much worse. It's not like right now we're anywhere close to being as awful as under the Great Depression, which happened 20 years after the Fed was formed to prevent things like the Panic of 1907."
The Structural Signature
A bit earlier in the month, Krugman returned to his familiar theme of saying that the employment data best match aggregate-demand explanations of the recession:
How do you assess stories about what's going on in the economy? You can go with your prejudices, of course.… But the way I usually try to do it is to ask whether the available facts fit the "signature" the story seems to imply — that is, do we see the general pattern that the argument would suggest we'd see?
Now consider the argument that our problems are mainly structural. The way this story is usually told is that we had too many workers in the wrong industries, that we have to expect a depressed level of overall employment as workers are moved out of these "bloated" sectors.
OK, so what should be the signature of that story? Surely it is that job losses should be concentrated in the bloated sectors, that employment should if anything be rising elsewhere — and wages should be rising in the unbloated sectors more rapidly than in the bloated ones.
Kind of looks like job losses everywhere, doesn't it?
And on wages,
Who's bidding for workers?
You can try to refine this stuff by disaggregating, but on first pass the signature of a structural problem just isn't there.
Before diving into the substantive issue, let me note that — once again — Krugman is engaging in sleight of hand. Ask yourself: Why is he using the absolute change in job numbers in the first chart, when he (quite correctly) used percentage changes in the second chart? The answer, I suspect, is that percentage changes in the first chart look like this:
Now let's make one more tweak. Although it's true that Arnold Kling (and perhaps others) have been pushing his "recalculation" story as if it continues to be the dominant theme throughout the entire recession, the standard Austrian position was that the initial downturn was necessary because of the structural imbalances of the boom years. Yet no Austrian to my knowledge said that the various rounds of quantitative easing, TARP, or the Obama stimulus package would have a disproportionate effect on construction.
If anything, those programs would have dampened the blow to housing and construction more generally, and the Austrian point was that we weren't doing the economy any favors by preventing the needed recovery. To put it in other words, the standard Austrian story was that the various interventions (all of which Krugman favored, though with caveats as to their design and implementation) would make Americans collectively poorer, and would hurt "the economy" in general. There is no reason to think that, say, the Obama stimulus package would yield a larger drop in construction versus services in Q4 2009.
To get a much better test, then, of whether the recession exhibited the "structural signature" before Krugman's favored (and what he considered inadequate) measures kicked in, we can look at employment in the three sectors Krugman chose for the test. The only difference (besides the obvious one of using percentages rather than absolute job losses) is that I'll change the time period from 2006 to 2008. The BLS link Krugman gave only has annual data, so these are the best start and end dates to isolate the popping of the housing bubble and the official onset of recession (in December 2007) but before most of the Keynesian "medicine" was administered (starting in late 2008 but kicking in especially in early 2009). Here's what the revised chart looks like:
That sure looks like a sector rebalance underway, to me.1
This is not the first time that Krugman has erroneously used data to cast aspersions on the Austrian position. In my reply to the first (and last?) blog post where Krugman specifically mentioned me, I wrote,
I can point to at least two episodes where the "sectoral-readjustment" story of the Austrians clearly has more explanatory power than Krugman's "insufficient demand" story. Specifically, in late 2008 Krugman argued that the housing bust had little to do with the recession, because the latest BLS figures showed that unemployment at the state level bore little relationship to the declines in home prices across the states.
However, I pointed out that looking at year-over-year changes in unemployment at the end of 2008 was hardly the right test. If we looked at changes from the moment the housing bubble burst, then five of the six states with the biggest housing declines were also in the list of the six states with the biggest increases in unemployment.
On another occasion … Krugman once again thought he had dealt the readjustment story a crushing blow when he pointed out that manufacturing had lost more jobs than construction. I pointed out that this too wasn't a valid test, because manufacturing had more workers to begin with. When we looked at percentage declines, then construction did indeed crash more heavily than manufacturing. Furthermore — and just as Austrian theory predicts — the employment decline in durable-goods manufacturing was worse than in nondurable-goods manufacturing, while the decline in the retail sector was lighter than in the other three.These are very important episodes. When Krugman thought the numbers were on his side, he was happy to cast aspersions on the sectoral-readjustment story; he thought his own model was perfectly able to explain the situation if the crash in housing really didn't have much to do with the upheaval in the labor markets. And, as Krugman himself argued, had he been using valid tests, then the outcomes would indeed have been challenging to the Austrian story.…
Because Krugman was the one who set up these two challenges, it is significant that the Austrian theory passed with flying colors.
In light of his recent blog posts on banking panics and structural signatures, I think we need to amend my statement to read that Krugman has now set up four separate empirical tests, and using his own data we see that the Austrian approach does far better than the Keynesian.
- 1. For the interested reader, I should note that the wage changes from 2006 to 2008 don't match up with my chart on employment changes the way one might expect: construction wages rose 3.8 percent annually; manufacturing rose 2.4 percent; and services rose 3.8 percent. But this isn't a devastating blow to the Austrian-versus-Keynesian explanations, because the wage changes don't "work" in Krugman's story either. He is explaining the drop in employment across the various sectors as due to insufficient aggregate demand. Yet even in that story, one would have expected the microfoundations of those job losses to match up with wage changes; since construction (both in percentage and absolute terms) fell far more than services from 2007 to 2011, even on Krugman's terms (invoking a shortfall in demand) we would presumably expect to see much lower wage growth in construction. Yet we don't; the wage growth in construction and services (as Krugman's second chart shows) was about the same between 2007 and 2011. I have some possible theories for why wage changes aren't lining up with employment changes, but it's not my responsibility to offer them in order to "rescue" the Austrian story from Krugman's critique, because his own explanation has the same problem.