Scott Sumner  

Recessions often begin before the thing that caused them occurs

Bio of Eugene Fama... A nice summary on secession...

I hope the title of this post has caught your attention---if not please reread it.

In the previous few recessions, the consensus of economists did not even forecast a recession until it was already underway. That is, not only can economists not forecast the economy, they can't even "nowcast" the economy.

One reason for this is that the early stages of a recession are often quite mild. In July 2001 the unemployment rate was only 4.6%, up from 4.3% in March, when the recession began. In April 2008 the unemployment rate was 5.0%, the same as in December 2007, when the recession began. Yes, unemployment can be a lagging indicator, but other monthly indicators were also relatively stable during the early months of each recession.

The title of the post relates to a strange quirk in the definition and dating of recessions. Not all fluctuations in the economy are regarded as recessions, just large ones. But the recession is assumed to begin when the data first starts dropping, even if the initial drop was too small to constitute a recession. It's contingent on what happens later. Thus if the Fed had done massive monetary stimulus in July 2001, or April 2008, and the economy had grown strongly, then there would have been no recession. And yet that stimulus would not have even occurred until 4 months after the recession began! If (like me) you believe the recession was caused by tight money (or lack of monetary stimulus if you prefer that language), then the recessions of 2001 and 2008 began before the event that caused them.

Here's an analogy. A bus normally drives down the center of the lane. Every so often it veers 3 feet to the right or left, due to inattention by the driver. Then the driver corrects. But one time after the bus veers 3 feet to the right, the driver is distracted by his cell phone ringing. The driver does not take the normal corrective action. Now the bus continues to veer off the road, into an accident. The key factor that caused the accident was the driver being distracted by the cellphone, but the beginning of the accident would be defined as the time the bus started to veer off course, if the NBER was doing the accident investigation.

Commenter BC asked me this very interesting question:

If recessions are caused by declines in NGDP and if we believe that NGDP futures markets are good at predicting changes in NGDP, then doesn't that mean that recessions are predictable, at least by NGDP futures markets participants? I can understand why EMH implies that it should be difficult *to profit* from knowledge of an impending recession, but I don't see why recessions should necessarily be unpredictable.

Conversely, if recessions are unpredictable, then does that imply that central banks always or almost always do a good job: they rarely have reason to believe that their policies will do anything other than to maintain stable NGDP? For example, when the ECB raised rates in 2011, was it predictable or unpredictable that such policy would cause a double-dip recession. If unpredictable, then was the rate increase the right (a priori) policy, the policy that would make future NGDP just as likely to be above target as below target, given information available at the time?

I hope you can see how the previous analysis relates to this question. Here was my response:

BC, That's actually a very good question, and the answer is very complicated. My view is that a NGDP futures market would not have been able to predict the 2008 recession until is was underway, but none the less would have been able to prevent it, or at least make it milder. Here's a possible sequence:

Late 2007, NGDP futures market would have predicted slowdown in 2008, but no recession. Had we been following NGDPLT, then monetary policy would have been made more expansionary at that time.

Then the slowdown occurred. During the second half of 2008, an NGDP futures market would have been predicting that the slowdown would turn into an outright recession. Notice that this is a successful prediction of a worsening of the recession, but it doesn't count as predicting a recession because the recession actually began in December 2007.

Here's a driving analogy. I can predict when the steering of the bus will lead it to drift off the center of the lane, but not when the bus will veer off the road and crash. If I nudge the bus back to the center each time it begins to veer off, then the bus never crashes.

An NGDP futures market cannot predict when the central bank will take the necessary nudges to keep NGDP track, but it can actually help the central bank do those nudges if it wants to.

As Jim Hamilton once pointed out, it would be a bad sign if the central bank could predict recessions:
You could argue that if the Fed is doing its job properly, any recession should have been impossible to predict ahead of time.

After all, if you could predict it, why wouldn't you prevent it? (At least for demand side recessions. Supply shocks are also hard to predict.)

As far as the 2011 double dip, my view is that futures markets might have missed the recession prediction at the time. But the policy was still a mistake because the eurozone economy was quite weak at the time, it was already skirting on the right edge of the highway. An NGDP futures market would have told the ECB to do a more expansionary policy.

Comments and Sharing

COMMENTS (16 to date)
Greg G writes:

I thought this was an original, interesting and beautifully written post Scott. All that is hard to find in writing about economics. I'm often unsure how much I am persuaded by your stuff but it always makes me think and it's always well argued.

Colombo writes:

History shows that definitions are flexible, by definition.

Why only one cause?

Scott Sumner writes:

Thanks Greg.

Colombo, Yes, multiple causes are possible, indeed likely. But I thought the basic point I was trying to make was easiest to explain with a simple demand-side policy failure.

Philo writes:

Rather intriguing. Let us grant--what I take to be your assumption--that the *cause* of an event E is the last event C that made E inevitable. Now, we might define a 'cession' as a period from six months after the start of a recession to the end of the recession, in which case the cause of a cession (which means: the cause of the period's *being a cession*) will--unsurprisingly--occur just before the start of the cession. But this very cause will also cause *the recession* which, however, will have started almost six months earlier--to the casual thinker, a seeming anomaly. Of course, the cause of the recession must be distinguished from the cause of *the start of* the recession; this latter cause will probably have been monetary policy just before the start--i.e., six months before the cause of the recession itself.

This sort of logic-chopping won't take us very far, but if we can't do it we're likely to get bogged down in a mire of confusion.

E. Harding writes:

I've noticed this, too. If the 2008 recession ended in August, it would have been viewed as a fairly mild (2001-level) inflationary recession linked to high gas and food prices and a slowing housing market.

Also, remember the non-recessions of 1951-2, 1966-7, 1986, and 1995.

James writes:


What effect do you think would NGDPLT have on growth rates? I'm asking for a number.

If NGDPLT only reduces the severity and frequency of recessions without affecting expansions, then that would have the effect of raising long term growth rates. You had a post a while back suggesting that policies can only change levels but not growth rates.

If NGDPLT reduces the severity and frequency of recessions but also reduces the magnitude and frequency of expansions, then that is consistent with the idea that policies cannot affect growth rates. But then it's also a free lunch: we get the same long term rate of growth with less volatility and no other costs.

P.S. I have looked but failed to find any specific predictions from you concerning what actual results you expect from NGDPLT, e.g. a X% reduction in the duration of median unemployment, a Y% reduction in the standard deviation of percent changes in per capita consumption, etc.

Colombo writes:

Well, perhaps the "demand-side policy failure" approach was too simple for my thick skull. I'm sorry.

Who said "make things as simple as possible but not simpler"?

Rajat writes:

Interesting post Scott, and I agree. It might be a slightly closer-run thing if the two consecutive quarters of negative GDP growth definition were used. As in, if the Great Recession didn't commence until July 2008, it's certainly possible an NGDP futures market would have signalled something awry by June 2008.

BC writes:

I think I get it. We can't predict recessions, only prevent them after they've already started. Not preventing a recession in the future causes it to have started happening in the past!

Scott Sumner writes:

Philo, Those views sound reasonable. Since you are a philosopher, you undoubtedly know that some philosophical puzzles are actually about the ambiguity of words.

James, As a first approximation I'd say no effect on the average growth rate. The business cycle would get substantially milder, but I can't say how much. I'd guess at least half as volatile.

Colombo, My assumption is that most recessions are caused by demand shocks, and the few that are not are also unpredictable, as they represent things like oil shocks.

Rajat, My hunch is that expected NGDP growth would have dropped really sharply after June 2008.

BC, Exactly.

michael pettengill writes:

On "prediction markets"...

"Wall Street" is a prediction market.

If prediction markets worked, the Wall Street would have predicted that the stock market would be where is ended this week, basically exactly where it was two weeks ago, and thus predicting no change over two weeks, the prediction market would not have predicted doom by going down by 10% in the five days starting under a week into the two weeks just ended.

Also, on price equalling value, if the stock market prices indicate the value of the corporations listed, then a week ago it seems like 10% of factories were blown up, 10% of key employees killed, by on Wednesday an 11% increase in factories happened by aggressive construction, and a flash job training increased the key employees by 11%.

Since 1980, with huge increases in individual stock investors influenced by emotion instead of logic and theory - ie, tens of millions of anti-Warren Buffetts, the stock market prices are increasingly irrational - driven by mood, fear, hope, the idea that the Wall Street market predicts anything is absurd.

And the stock market no longer functions as it did for centuries: as the means to fund industry investment. 99.99999% of stock buying is 100% speculation that tends to force capital destruction by CEOs. Unless the CEO announce firings or plant closings, the stock buyer is a loser because profits won't increase.

Of course, recession happen when all the CEOs are driven by stock speculators demanding higher share prices to respond by firing workers.

Fred Anderson writes:

Michael Pettengill;

I am not trained as an economist, so I risk looking the fool here, but much in your comment seems to me in error.

(1) It seems obvious to me that the people "predicting" on Monday weren't the same ones "predicting" on Wednesday. I should think that prediction markets, in order to work, would require a reasonably complete survey of all those predicting.

(2) Price isn't equal to value. Price is what one must pay to get a thing. In normal purchases, the buyer believes the item is worth more than he/she is paying (else, they wouldn't buy). (And the seller believes the item is worth less than he/she is asking; else they wouldn't sell.) They can both be right; what it's worth to me may not be what it's worth to you. Wednesday's buyers not only were different people than Monday's sellers, but they necessarily had two more days information on which to base a valuation.

(3) Re: the huge increase in individual investors. I thought this had spurred a huge increase in financial intermediaries. That these institutional investors were there precisely to buffer the "mood, fear & hope" you're worried about.

(4) The assumption that the only way profits (& hence stock price) can increase is by cutting labor costs seems especially antediluvian. People aren't buying iPhones because they're cheaper than the old drop-a-nickel wall boxes (which, come to think of it, imposed no purchase cost on their users). Sam Walton made his fortune by lowering the cost of goods to consumers, and he did that primarily by re-engineering the supply chain to drastically slash distribution costs. You won't buy that miracle drug that could save your little girl's life because it's cheaper; you'll buy it because it offers you high value.

Yes, any well run business will be looking for ways to do whatever they do more efficiently. But if you want lots of customers who'll gladly pay you high prices, you're focussed first & foremost on how can we deliver greater benefit to those customers. That costs money. I think investors understand that companies can't scrimp their way to prosperity. (Consider, for example, stock prices for pharmaceutical companies that are adjudged to have little in the pipeline.)

Nathan W writes:

It seems to me that playing with definitions and word games is mostly what is at stake here.

If we did away with the word "recession" and instead said "indicators of GDP were in decline for four months before actions which may have made it worse" then we wouldn't be playing semantic games such as actions which "cause" the recession by causing declines in GDP indicators which extend the period of decline from four months to six months (the definition of a "technical recession").

Also, I'm not sure that it's quite accurate to say that it is the job of the central bank to prevent recessions. It is the job of macroeconomists to search for policies which reduce the negative impacts of recessions, but from early chapters of any intro to economics text, my understanding has always been that recessions are an important part of long term growth, in that they facilitate shedding of less productive/profitable enterprise, thereby freeing up resources for growth sectors.

Jose Romeu Robazzi writes:

Predicition Markets are not perfect. It is perfectly possible that collectively market participants were focusing on wrong price drivers. But, once new information comes in, prediction markets will adjust accordingly, and will do it very fast. That compares much better with a "board of experts" that can have their views biased by groupthink and other cognitive biases and therefore be slow to ajust. And more importantly, this small group failures affect the lives of virtually all people in the planet, if we think of the the FRB. I prefer an imperfect market.

James Alexander writes:

It's not entirely surprising central banks cause recessions when their primary goal is low inflation. The recession is an accidental by-product.

Tough on the unemployed and the employed who suffer low or no wage growth. But at least everyone can take comfort that the central banks' credibility is intact and that we don't have hyperinflation.

iVALIS writes:

Interesting post Scott, I must agree with Greg G re quality!
A question, if you or others would be so kind:
I have recently come to the conclusion that inflation has been demonised in the public consciousness, and aggressively targeted by Central Banks for ONE reason: it tends to redistribute wealth downwards.
I do not have the figures here to prove this, however it is a "gut feeling" from numerous articles I've read recently comparing wage/ wealth/ inequality in US/ UK post Second World War.
Particularly the contrast between the Mixed Economy model, in place until around the end of the 1970's, and the changed "free market" approach thereafter.
Anyone have any thoughts on the validity (or otherwise) of this opinion?

Comments for this entry have been closed
Return to top