14 comments

1. 

   Gregory Sokoloff · October 4, 2010 - 2:02 pm · Reply→

   A good article, albeit too short. It would have been more fun with a few examples, especially with so many hilarious ones to choose from. For example, just today, Krugman was poking fun at Peter Schiff’s brilliant predictions of imminent hyperinflation.
   Good luck with the Economy Lab–such a section is long, long, overdue.

2. 

   Robillard · October 4, 2010 - 3:47 pm · Reply→

   There has got to be a good short description for forecaster behaviour. I would nominate “The Price is Right Effect”, so called for the “One Bid” segment in which the four contestants try to guess the price of the prize lot and get as close as possible without guessing above the actual value. In most cases, the last contestant to guess will bid one dollar more than the previously highest bid.
   In the same way, forecasters may be incentivised to just be produce forecasts that are only slightly higher or slightly lower than those of their competitors so that they can claim that they came the closest, or at least were not much worse than the other forecasters.

3. 

   anon · October 4, 2010 - 3:51 pm · Reply→

   Some prediction markets (betting exchanges) trade contracts on macroeconomic forecasts such as U.S. GDP growth, whether there will be a recession in the near future, etc. If these contracts are sufficently liquid, I’d trust the market odds more than an expert’s forecast. This goes even more so for variables which have large, well-established markets, such as interest rates, inflation, credit events etc.
   Of course, private sector forecasts could still be valued for a comprehensive economic analysis, or for providing additional scenarios which would be excluded by the consensus, but may provide guidance against “tail risk”.

4. 

   Daniil · October 4, 2010 - 5:05 pm · Reply→

   Yes, blame forecasters. Except that most forecasts are ‘conditional’ and talking about forecasters and not mentioning that – usually means that objectively portraying forecasters is not on the agenda.

5. 

   Stephen Gordon · October 4, 2010 - 6:37 pm · Reply→

   Oh, I understand the difficulties of forecasting and what the limitations are. I just think it would be better if forecasters were less nervous about dissenting from the consensus and if they got into the habit of announcing error bands to go with their projections.

6. 

   Matt Nolan · October 4, 2010 - 6:37 pm · Reply→

   Good article – captured a lot of the issues in a short space. I’ll be sending that around work.

7. 

   Just visiting from Macleans · October 4, 2010 - 6:51 pm · Reply→

   BMO is one of Canada’s protected banking oligopolists. Could that explain why “Dr. Sherry Cooper, executive vice president and chief economist for the Bank of Montreal will receive the [Lawrence R. Klein] award, which is awarded annually to the most accurate economist”?
   Read more: http://www.montrealgazette.com/business/fp/Cooper+wins+award+forecasting/3596145/story.html#ixzz11QuuBRMU

8. 

   Determinant · October 4, 2010 - 8:51 pm · Reply→

   Only if you tell us why your post isn’t a Straw Man argument.

9. 

   Just visiting from Macleans · October 4, 2010 - 9:00 pm · Reply→

   If that was directed at me, then the hypothesis being tested is this:
   As ever, the problem is one of incentives. A plausible hypothesis is that private-sector forecasters face the same incentives as do fund managers: it is relative, not absolute, performance that matters
   If a protected industry (Canadian banks) face less competition/private incentives, then their forecasts should theoretically be better. No?

10. 

    Geoff Castle · October 4, 2010 - 10:25 pm · Reply→

    Congratulations. A worthwhile effort on the part of the Globe.

11. 

    Jim Rootham · October 4, 2010 - 11:04 pm · Reply→

    With respect to the length. You made comments about word count earlier. I would expect you to have continual trouble with this. Unfortunately this is part of the Globe redesign into shorter pieces. AKA dumbed down.

12. 

    Dan · October 5, 2010 - 12:02 am · Reply→

    Then suggest a meaningful way to put confidence bands around conditional forecasts based on judgement calls. It’s makes perfect sense to put confidence bands around unconditional forecasts which are purely mechanical function of the data, but for some reason there aren’t many of those. Maybe it’s because those statistical ones rarely outperform random walk out-of-sample.
    On the deviations from consensus – there are certainly some forecasters who feel that way (probably mostly in the public sector), but there are also wall-street forecasters, where there’s demand for both ‘contrarian’ and pretty strong directional forecasts.

13. 

    Simon van Norden · October 5, 2010 - 3:10 am · Reply→

    Actually, most forecast consumers don’t want conditional forecasts; they want plain old (i.e. unconditional) forecasts! Most investors, for example, don’t want to know which stocks to sell if China revalues its currency; they want to know which to sell unconditionally!
    Conditional forecasts protect forecasters from serious evaluation; that is not good for forecast consumers.

14. 

    Dan · October 5, 2010 - 1:48 pm · Reply→

    Change ‘most’ to ‘some’ above and it would be an accurate statement.