The full post of the discussion about economic forecasting that I had with Professor Kashif Mansori of Colby College and Angry Bear is now up at the Wall Street Journal’s Econoblog. Here’s my bottom line:
Don’t ask for too much of your forecast or your policy, and it won’t disappoint you.
Or maybe you should read the whole thing.
Excellent discussion you had with the AB colleague who got his Ph.D. from Princeton (just had to toss that in since you linked to Colby College).
Good discussion as far as it went. I would be interested in your view of how much bad data contributes to the problem of forecast errors. I submit that even a model that was an accurate representation of the true economy and evolved to remain accurate would still have significant real time errors given the quality of the data available in advance or even after the fact.
Good discussion (I’ve worked as a forecaster in NZ). You both seem to wind up at the same point.
Since that job, I’ve considered the unpredictability of the future economy as a possible argument for free-will. If statistical techniques cannot tell us what the future will hold, perhaps it is because human beings have a space to make up our own minds? That we are not just tools of environmental determinism?
Economic forecasting: the good, the bad, and the ugly
What are the costs of an inaccurate forecast? How can forecasters do better? The latest WSJ Online Econoblog is about The Perils of Forecasting, featuring bloggers James Hamilton of Econbrowser and Kash Mansori of Angry Bear. They end up more or less o…
Very good dicussion. I have 30 years experience as a business economist and agree that no one can forecast. But that does not mean that forecasts do not have value.
I actually argue that it is impossible for the consensus to forecast a recession. my argument is that recessions occur when the business community makes a mistake. Since business plans are largely based on the consensus, for the consensus to forecast a recession it has to forecast that consensus expectations are wrong.
Or, you have to forecast that there will be unintentional excess inventories or investments.
I’d like to see more kinds of futures markets that could do forecasts directly, without a lot of interpretation and guessing. For example, stock markets and index futures forecast future recessions to some extent, and likewise for bond yields at different maturities. But maybe there could be a GDP future that would do so more directly. Or you could have options that would pay off depending on whether GDP went up or down by a certain amount.
I’ll bet such a market would do better than all those economists with their charts and formulas. Maybe they’re all afraid they’d lose their jobs if they pushed for a market like that!
Well, you can forecast. I predict that on 25 December 2006 there will be a Christmas.
I predict that retail spending will go up in December 2006 compared to other months this year.
And, more significantly, I predicted at one point that the total tax take of personal salary/wages tax would be smaller on the first month of the government’s fiscal year after the government increased that tax than it was on the same month the previous year. And it was. This was not some extreme application of Laffer’s curve. This was because most people in NZ are paid fortnightly on a Wednesday or Thursday, and most months have 31 or 30 days, so occasionally there are five Wednesdays and Thursdays in a month so the tax paid for that month is higher. And the tax increase was not large enough to offset the three payments not two effect.
“This is sometimes described as the “random walk” theory of stock prices, or, more accurately, the view that the stock price should follow a martingale.”
I had learned that the stock market is a submartingale:
“A submartingale is like a martingale, except that the current value of the random variable is always less than or equal to the expected future value.”
From the same wikipedia that you referenced.
Actually, it wasn’t me who referenced Wikipedia– WSJ threw that in.
As we discussed in a separate thread on this yesterday, the more complete statement yet is that the return on the stock minus the Tbill yield over the corresponding period minus the average value of that spread between the two should form a martingale difference sequence. But, for purposes of talking about a one-week stock return, those adjustments are quite minor, and putting in all those words kind of obscures the big picture.
On doing some research I learned that there does exist a limited market in GDP futures. It happens just one hour a day, one day a month, on the Chicago Merc exchange, and is only for the next release of quarterly GDP results.
They also do one early in the morning before the official results are announced, and as it happened, that was today. 4th qtr GDP stunned economists, coming in at only 1.1% (annualized) growth. Compare this to the 3rd quarter which was at 4.1%. So the economic community blew this call. How did the market do?
I had to register at auctions.cme.com to see the results. Turns out that the market did no better. The exchange did not even allow bids below 1.3; I guess no one considered the possibility it might come in lower than that. The imputed probability for 1.3 or below was only 2.1%. In other words the markets thought there was only one chance in 50 that 4th quarter GDP growth would be as low as 1.3. We can only guess at what the odds would be for 1.1 – probably even less.
I guess there are a couple of ways to interpret this. Maybe markets don’t really do any better than official estimates, in fact maybe they just slavishly follow the estimates. Or on the other hand, maybe markets do well and the information available just wasn’t sufficient to allow for a reasonable GDP guess. In that case, economists shouldn’t feel so bad; they were doing about as well as was possible with the information at hand.
One problem is that this market is pretty thin. It’s not widely known, it only runs one hour a month, and it’s at 7 AM Chicago time which is not too convenient. I don’t see why they can’t just keep it running all the time, similar to an index future. Maybe then the market would do a better job of attracting inside information and giving businesses a truly useful tool for planning future operations.
One thing I’ll add – I see at Angry Bear that the consensus estimate was for 3.5% growth, with “some pessimists” predicting 2.5%. The market consensus was 2.7%, better than what Kash had for the finance community overall (although not as close as the pessimists). It’s an interesting sign that the market estimate was much closer to the pessimists than the official consensus, but of course they all missed the mark so badly that none of them ended up looking good.
It should be very interesting to see Macroblog’s Monday report on market expectations of the Fed’s continual increases in interest rates. One would have to think that this will increase the odds of a suspension sooner rather than later.
Why do forecasters only report point estimates?
In a recent exchange, Jim Hamilton of Econbrowser and Kash at Angry Bear discuss the perils and pitfalls of economic forecasting. They agree on many things, most especially on the importance of maintaining a certain level of humility. The best