Sunday, May 06, 2012

What is a "financial crisis"? (reply to John Cochrane)

The conventional wisdom says that recessions that follow financial crises last longer than other recessions. In a recent blog post, John Cochrane challenges the conventional wisdom:

Financial crises certainly don't always and inevitably lead to long recessions, as the factoid suggests... 
In a nice article for the Atlanta Fed, Gerald Dwyer and James Lothian went back to the 1800s, and find no difference between recessions with financial crises and those without. Some, like the Great depression and now, last a long time. The others don't.   
Michael Bordo and Joseph Haubrich wrote a somewhat more detailed study of US history, (which I found through John Taylor's blog) concluding 
recessions associated with financial crises are generally followed by rapid recoveries. We find three exceptions to this pattern: the recovery from the Great Contraction in the 1930s; the recovery after the recession of the early 1990s and the present recovery. ... 
In contrast to much conventional wisdom, the stylized fact that deep contractions breed strong recoveries is particularly true when there is a financial crisis. In fact, on average, it is cycles without a financial crisis that show the weakest relation between contraction depth and recovery strength 
This had pretty much been the "stylized facts" when I went to grad school: US output has (so far) returned to trend after recessions. The further it falls, the quicker it rises (growth). Financial crises give sharper and deeper recessions, followed by sharper recoveries, but not, on average, longer ones. This "recovery" is in fact quite unusual, looking more like the Great Depression but unlike the usual pattern.  
As I did minor searches for the facts however, it's clear there is an explosion of work on this subject, so it's hardly the last word.  
When I read this, the first question that popped into my mind was "OK, but how are they defining financial crises?" It turns out that the two sources Cochrane cites disagree on this point. The article by Dwyer and Lothian says:
No U.S. recession since World War II [other than the current recession] has been associated with a financial crisis. 
While the paper by Bordo and Haubrich says:
Consequently, the recessions we associate with a financial crisis are those that start in 1882, 1892, 1907, 1912, 1929, 1973, 1981, and 1990.
That's a big difference! Three post-WW2 financial crises versus zero?

The problem, of course, is that it's difficult to define a financial crisis. Many people seem to use the term to mean "a large drop in asset prices" (this is the definition that leads to the claims that economic models can't forecast financial crises). But there are other possible meanings of the term. For example, "financial crisis" could also mean:

  • A large number of bank runs, leading to a liquidity crisis
  • A solvency crisis, in which most large financial institutions are found to be insolvent

It seems to me that this third definition - the simultaneous insolvency of most large financial institutions - is the kind of "financial crisis" that most people would casually associate with long recessions and slow recoveries. Note that a steep fall in asset prices is neither necessary nor sufficient to generate a solvency crisis, since firms may have different degrees of leverage.

If we define financial crises as asset price drops, it seems pretty obvious that "financial crises" will be associated with most recessions. This is because asset prices are forward-looking and quick to react to events; any shock that will cause a recession over the next year will cause an asset price almost as soon as the shock is realized. I strongly suspect that Bordo and Haubrich are using this definition, since they claim that a financial crisis happened just before the 1981 recession. Stock prices would obviously fall in reaction to an interest rate hike by the Fed (which most people believe caused the recession in '81).

Dwyer and Lothian seem to be using a different definition of "financial crisis". I'm not sure what their definition is, but my guess is that it is a liquidity crisis or solvency crisis type of definition.

Now, both of Cochrane's sources reach the same conclusion, which is mainly based on pre-WW2 data. I don't know much about pre-WW2 economic history, so I'm not really qualified to evaluate their claims. But I think that the confusion over definitions merits a long hard look into exactly what happened in and before all of those pre-WW2 recessions.

Two more points:

1. Actually, my intuition says that financial crises are not the cause of slow recoveries. My intuition is basically the "balance sheet recession" story, which is about a sudden regime change in people's behavior toward debt and consumption after a long build-up in debt. My intuition says that people's sudden shift to a "balance sheet rebuilding" regime will tend to cause both a long recession and a financial crisis. But then again, this is just my intuition...As Cochrane says, there is lots of research being done on the subject.

2. Cochrane definitely does do a good job of rebutting a factoid tossed off by Bill Clinton, which is that recessions after financial crises last "5 to 10 years". Cochrane cites evidence from Reinhart and Rogoff that shows that the international average has been just under 5 years for recent crises. Although I'm not sure I believe the Reinhart/Rogoff numbers (the Great Depression lasted only 4 years??), it is definitely the case that countries such as Sweden, Finland, and Korea have managed quick recoveries after financial crises. I think we should look at what they did, and see if maybe we can replicate their best practices.


  1. malcolm10:33 PM

    Figure 3 looks at increases in unemployment rates across the historical comparison
    group. (As the unemployment rate is classified as a lagging indicator, we do not include
    the current crisis.) On average, unemployment rises for almost five years, with an
    increase in the unemployment rate of about 7 percentage points. While none of the
    postwar episodes rivals the rise in unemployment of over 20 percentage points
    experienced by the United States during the Great Depression, the employment
    consequences of financial crises are nevertheless strikingly large in many cases.
    Figure 3

    Duration in years
    4.8 years

    Rogoff and Reinhart clearly believed in their AEA 2009 article that
    unemployment tends to rise for 5 years during a crisis plus recession in their figure 3.

  2. financial crisis are slow because the debt workout process is slow.

    nominal debt contracts like mortgages are 30-year fixed price contracts. When the house price drops, your textbook grad macro 101 sticky price model suggests that disequilibrium will occur until contracts negotiated under the previous price level are renegotiated.

    renegotiation can occur through principal paydown or default.

    Most industry and academic research suggests that the cost of default is 40-60% of the value of the house (FICO score stain for 7 years, lose the house, etc etc). Plus it takes a long time to foreclose (right now, 2 years).

    So a mortgage is a 30-year fixed price contract with a 40-60% cancellation fee.

    Thats why 70% mortgages that are 115-150 LTV (the mortgage is 15-50% more than the house) remain current. (meander on over to calculatedrisk mortgage delinquency tables).

    Mortgages have a dual trigger due to the cost of default: people stay current until they lose their ability to pay (or walk away "strategically default" when sufficiently underwater.

    Not surprisingly, delinquencies are highly correlated with the unemployment rate. we are only 1/2 way through this mess because it takes 700+ days right now to foreclose.

    And: as long as unemployment remains above ~6%, delinquencoes will remain elevated. 25% of homeowners are underwater making them suceptible to default if they lose their job.

    Fishers debt-deflation!

    now implications for macro policy:
    1. clamp down on stupid lending. stupid lending lowered the cost of owner-occupied housing (financing) pumping up prices.

    the Fed clamed down on this in 2006; by 2008 we were 18 months into the housing recession and:

    2. when there are a lot of underwater mortgages dont let unemployment skyrocket as it will create a delinquency-default-homeprice decline spiral that will make even more people suceptible and further pressure home prices.

    in other words, yep here it is: dont let ngdp drop because of that dual trigger.

    the paydown or renegotiation of principal needed to "reset" the debt contracts to the new market level is slow either through foreclosure (which takes 2+ years) or through simple paydown (which might also take years). Either way, disequilibroum will continue until most debt is reset at the new lower levels.

    we are about 1/2 way through the mess. delinquencies are declining with UE, but will not get back to normal levels until UE reaches about 6%. meantime, home prices will still decline, but not nearly as fast. As the debt gets worked out, construction should pick back up (at new lower prices, just as you would expect from a sticky price model).

    financial crisis are slow to resolve because debt is the stickiest price around.

    1. and by the way, by implication, the speed with with one merges from a debt-deflation recession is directly a function of inflation.

  3. The duration number quoted in Cochrane's piece may be peak to trough - not peak to recovery.

    A financial crisis in a small open economy is not the same as a financial crisis in the USA. Failed domestic institutions can be readily replaced by foreign ones and falling exchange rates can cause rapid expansion of exports.

    (In the summer and fall of 1981 interest rates went through the roof as a matter of deliberate Fed policy and the economy did a face plant.)

  4. Anonymous12:45 AM

    Seriously, this whole debate shows what's wrong with the economics academy. Why is the actual mechanism of a given crisis and recovery such a complete fucking mystery to academics that the best that can be done is this, pathetic stamp-collecting of past crises. It really is the laziest, sloppiest waste of time that ever passed for an academic field.

    How about this? Ban toy models, and any conclusions drawn from them. They are not publishable - full stop, and any economists that offers them is rejected and embarrassed. Economists have to explain the real mechanisms of real world events - and only to the extent they can with a proper standard of proof- or admit that they can't, which itself is information that society desperately needs to know to help plan and respond to events.

    1. Anonymous3:03 PM

      Economics is a social science. Accept it as such.

    2. or admit that they can't, which itself is information that society desperately needs to know to help plan and respond to events

      I really agree with your post

      you are seeing "incentive caused bias," aka lying for money, in action.

      If economists admitted how little value they have, their stock (and earnings and salaries) would fall through the floor.

    3. Anonymous11:04 PM

      Economics is secretly useless and everyone knows it, but there's a huge cover-up among economists to protect their wages and EVERYONE is playing along. Economics continues to be taught, funded and listened to only because the Illuminati makes the whistle-blowers disappear.

    4. Anonymous11:58 PM

      Here, here! and I agree, but this leads to the question of what therefore promotes the field as relevant and valuable to "society"? Self serving interests among the academic economist's community cannot sustain the field or it's academic studies, conclusions, speculations, or "theories". Until this question can be objectively answered then the issue remains unexplained, which therefore also negates the subjective feeling that the profession is worthless.

  5. I don't think there is much of a best practice to emulate - Sweden and the others had the good luck to have financial crises in relative isolation from the rest of the world economy.

  6. Anonymous4:41 AM

    I decided to stop going on Cochrane's blog after reading that article, and I'm really surprised that you're not coming down harder on him on this one.

    As commenters above said:

    > financial crisis are slow because the debt workout process is slow.

    Well, isn't it obvious? If instead of borrowing ever greater sums to fuel consumption you spend the next 20 years repaying your mortgage, of course it's going to slow down the economy. Especially as creditors as well (such as banks) want better balance sheets and accumulate cash instead of spending / lending again, so debt that gets repaid enters a black hole instead of getting pumped back into the economy.

    > I don't think there is much of a best practice to emulate - Sweden
    > and the others had the good luck to have financial crises in
    > relative isolation from the rest of the world economy.


    > Seriously, this whole debate shows what's wrong with the economics
    > academy.


    1. most borrowing is used to finance investment - houses, factories, etc. - or durable goods like cars. oh, and student loans.

  7. Anonymous12:57 PM

    About the great depression lasting more than 4 years ("Although I'm not sure I believe the Reinhart/Rogoff numbers (the Great Depression lasted only 4 years??"), an explanation:

    When you take averages, some data points will be BELOW the average, and some ABOVE. This is the nature of an AVERAGE.

    Let me know if you need more help with this concept.