Thursday, July 11, 2013

Economic Predictions in News Media



Have you noticed the headlines and articles about economic data, things like the unemployment figures and growth rates, that often include references to predictions by a group of economists? The headline will say, “New jobs exceed expectations!” “Growth rate falls below predictions.” What predictions? So what?

I have seen no complete explanation of where these predictions come from. In a couple articles I have read, the author has offered a one or two sentence note to give some credence to the prediction, but nowhere I have I seen any statement about why the expectation has any meaning or what that meaning could be.

From what I can tell, the set up is something like this: A news company has contracted or at least asked several economists, presumably people with the appropriate background, who will periodically provide their expectation as to what the figures in soon to be published reports will be. This practice is used in at least a few different countries.

Which economists are used isn’t mentioned or if they are academic, government, or private economists. Nor has any mention been made about what methods of prediction are being used by any individual predictors.

One reporter mentioned that the news company did drop extreme predictions, i.e., ones that were considerable different from the majority of responses from the predictors. Otherwise, the process seems to be that when the responses are in, the news company averages the numbers, and that becomes the standard for evaluating the real number when it is announced.

One wonders if the predictors are evaluated in any way. For example, if one economists consistently offers numbers that are way off or always in the wrong direction, would he be stricken from the list? Should there be any weight given to the predictor is usually closest to the real number? One wonders if there is any consistency between the methods used by the various economists offering predictions. If one, for example, uses the Mystery 8 Ball, another uses a computer model based upon the proportional orbits of the planets, and a third uses a model based upon Keynesian precepts, what could it mean to average the predictions? That is an extreme example (who in their right mind would use Keynes?), but if everyone’s methods were inconsistent, what would an average mean? What would any comparison of answers mean?

But the real thing is that the meaning of the prediction is the opposite from what the news organization suggests. What should be happening is an evaluation of the method and underlying reasoning for the prediction in comparison to the real numbers, i.e., reality (assuming that the “real numbers” are themselves generated by a rational method). If the predictions by a particular method and theoretical framework consistently provide a figure reasonably close to the reality then the validity and truthfulness of the theory is supported. Possibly, if the theory is consistently correct in its predictions, then that particular approach could be used in the future for predictions for some reason. But that still wouldn’t support the present news organization approach. You certainly wouldn’t take a poll.

But the relation of the real number to the prediction means nothing. The relation of the two provides no knowledge about the consequences or importance of the real number.

No one should care. (Well, except the economist who truly wants to understand the economy.)

1 comment:

  1. Spot on. They just collect an average, and that means you can almost get the same thing by naive extrapolation of the short-term trend.

    I wonder if anyone has done that test. Have a prediction that takes absolutely no knowledge into account other than the recent past (say last 4 readings) of the particular metric being predicted. Then, take the "consensus" estimate, and see which one is the better predictor.

    ReplyDelete