InTrade’s Strangely Misnamed Recession Bet

US Recession 2008

Chart via Intrade



I am only a partial believer in The Wisdom of Crowds.

Why? Most of the time, the crowd is correct. However, at turning points, or in manias, the crowd can be totally and expensively wrong. That’s why I always watch the trend, and what the crowd is doing, and whether we are in one of those rare moments when the profitable thing to do is fade popular opinion.

Hence, it caught my attention in the Spring when the Intrade Recession forecast for 2008 suddenly went into freefall. The Economist community dropped the odds of recession dramatically over the same March to June period. Was the population of betters suddenly like minded?

This was one of those things I have been meaning to address. I nearly forgot about it, too, until suddenly a spate of articles and television pundits began relying on the above Intrade chart as evidence there will not be an economic contraction.

Off I go to the Intrade site to see what there was to see. It didn’t take very long to discover the flaw in their betting process. According to Rules and Legal Info at the site, this bet actually is defined as:

“For expiry purposes, a recession is defined as two successive quarters of negative real GDP growth.

Expiry will be based soley on the data reported by the U.S. Department of Commerce (Bureau of Economic Analysis, Table 1.1.1, “Percent Change From Preceding Period in Real Gross Domestic Product”) as reported by the BEA.

If the table as reported at that time indicates that any two
consecutive quarters between (and including) Q4 of the previous year
and Q4 of the year specified in the contract are negative, then the
contract will expire at 100. Otherwise, the contract will expire at 0.
For example, if Q4 of 2007 and Q1 of 2008 both experience negative
growth then the contract for 2008 will expire at 100.”

Aha! This bet is misnamed — it is not a trade based upon whether the US is in a Recession; rather, it is a bet as to whether the US has 2 consecutive quarters of negative GDP. As we have so painstakingly discussed, is not the definition of a recession.

Intrade recognized that GDP numbers tend to get revised long after the fact. So thety added yet another twist to this strange bet:

“The final figures will be used for expiry – not the advance or preliminary numbers…Any changes to the result after the contract has expired will not be taken into account – Exchange Rule 1.4.”

So it is the Final GDP numbers, reported on the 3rd month after the quarter has ended, that will be the basis of determining the winner, and not the more accurate revised data, such as we just had for Q4 2007.

Astonishing. What a strange wager: It does not use the actual definition of recession, and it does not use the most accurate data available.

Regardless of the title, this is not a recession bet. Perhaps the trade “USA two Consecutive Quarters of Negative Final (but Unrevised) GDP” has less of a ring to it than USA Recession 2008.


Continual Spot Price, US RecessionPrice for US Economy in Recession (*see contract rules for definition*) at


Update: August 5, 2008 10:30am

Portfolio’s Felix Salmon points out that Don Fishback made a similar observation — Don observes that according to Intrade’s bet, there was no recession in 2001 !


Recessions Often Begin With Positive GDP Data (May 2008)

Why Prediction Markets Fail (January 11, 2008)

Pervasive Pollyannas of Prosperity (July 2008)

Whining US CEOs: Economy is “Dismal”  (July 2008)


Intrade Recession Bet

Rules and Legal Info

Business Cycle Expansions and Contractions
National Bureau of Economic Research

Print Friendly, PDF & Email

What's been said:

Discussions found on the web:
  1. jf commented on Aug 5

    Yeah. And where can I place my bet that the homeland security threat level gets elevated prior to Nov. 3? What happens if they change the definition of GDP growth?

  2. Rex commented on Aug 5

    Nice one. I’d noticed this odd bet too. Glad you took the time to correct them.

    Why not just have a contract that pays off if the NBER declares a recession? I guess the problem with that is it would have an indeterminate end to the contract.

  3. Smokefoot commented on Aug 5

    I was looking at the Recession 2007 bet back in late 2007, but at the time it didn’t look like it made sense at all – the bet ended on 12/31/2007, long before the Q4 GDP numbers were released! I didn’t bet on it because I didn’t know how the dealt with a bet which couldn’t be decided on the expiration date!

    Did they fix this with the Recession 2008 bet?

  4. James commented on Aug 5

    I was trying to read the article on Fishback’s site, but when I click on the link, it give me an error?

  5. observer commented on Aug 6

    I have been screaming about this for ages, but you have a louder megaphone than I do. Thanks for mentioning it.

    This basically amounts to a bet that both q3 and q4 will have negative gdp growth – a bet I wouldn’t make.

    A better definition (but one that might not get decided until early or mid 2010) would be something like “the Business Cycle Dating Committee of the NBER establishes that a recession occurred during a period that contains at least one month of 2008.”

    “Don observes that according to Intrade’s bet, there was no recession in 2001 !”

    Technically, yes, but it depends on when you look at the series! Most people don’t have any idea how much the gdp figures get revised.

    For example, the original “final” estimates as reported:

    2000q1 5.5
    2000q2 5.6
    2000q3 2.2
    2000q4 1.0
    2001q1 1.2
    2001q2 0.3
    2001q3 -1.3
    2001q4 1.7
    2002q1 6.1

    In July of 2001, prior quarters got revised as follows:

    2000q1 2.3
    2000q2 5.7
    2000q3 1.3
    2000q4 1.9
    2001q1 1.3

    In July of 2002, the revisions changed the series to:

    2000q1 2.6
    2000q2 4.8
    2000q3 0.6
    2000q4 1.1
    2001q1 -0.6
    2001q2 -1.6
    2001q3 -0.3
    2001q4 2.7
    2002q1 5.0

    Note we had three consecutive quarters of negative gdp growth at that time.

    December of 2003 brought more revisions:

    2000q1 1.0
    2000q2 6.4
    2000q3 -0.5
    2000q4 2.1
    2001q1 -0.2
    2001q2 -0.6
    2001q3 -1.3
    2001q4 2.0
    2002q1 4.7

    Still three consecutive quarters of negative gdp growth. Note that 2000q1 got revised all the way down (starting at the “final” estimate, remember) from 5.5 to 1.0! And 2000q3 went from 2.2 to -0.5.

    Somewhere along the way, more revisions have occurred (I can’t date these easily), but today the figures come out like this (with no change for any quarter in 2000 from the December 2003 revisions):

    2000q1 1.0
    2000q2 6.4
    2000q3 -0.5
    2000q4 2.1
    2001q1 -0.5
    2001q2 1.2
    2001q3 -1.4
    2001q4 1.6
    2002q1 2.7

    With 2001q2 back to positve, we (again) no longer have even two consecutive quarters of negative gdp growth. Note that 2002q1 went from 6.1 down to 2.7 after three rounds of revisions.

    The actual recession dates (200104 to 200111) doesn’t contain any months from the first two quarters of contracting gdp, but encompasses all of the third.

    No number in this series is ever safe. Sometimes the BEA changes the methodology resulting in revisions of every number back to 1929!

Read this next.

Posted Under