Free Read of the WSJ article:
Why we can not believe the Fed
The Fed studied its own staff's forecasting performance over the period 1986 to 2006. It found that the average root mean squared error—or the deviation from the actual result—for the staff's next-year gross domestic product (GDP) forecasts was 1.34, compared with 1.29 by what the Fed describes as a "large group" of private forecasters. That is, the Fed's predicting performance was worse than that of market-watchers outside the Fed. For next-year CPI forecasts, the error term was 1.03 for Fed staff, and only 0.93 for private forecasters. The Fed's conclusion? In its own words, its "historical forecast errors are large in economic terms."
I see the problem now. The problem is that the WSJ doesn't understand math.
A large sample results in a smaller standard deviation.
A smaller sample group results in a larger standard deviation.
We expect the standard deviation for a small sample of Federal Reserve staff to be a little higher than the standard deviation of a larger group of private forecasters.
There is nothing interesting here.
RMS error = Standard deviation = Root ( mean ( square ( x - x')))
= Root ( Sum( (x-x')^2))/(n-1))
That is, you square each forecast number. You subtract the square of the actual value from it. You then add them all up. This gets divided by the total number of forecasts minus one. This then is square rooted.
Another similar measure of error is to use the absolute value of the differences.
Both methods produces a number that is internally consistent in given some scale that represents error.
One thing is that the final reported GDP itself is an estimate from a sample. So the forecasts are not being compared to the GDP, they are being compared to another estimate of the GDP which itself has an unknown error.
For a large group, n is larger so the standard deviation is smaller.
For a small group, n is smaller so the standard deviation is larger.
The problem isn't the Fed, the problem is the Wall Street Journal doesn't understand what a standard deviation is.
You really don't need to go that far. All you gotta ask is what 1.34 - 1.29 = .05 means?
What does .05/1.29=.003 mean? That is a .3% difference between the two.
If two batters have batting averages of 300.0 and 300.7, what does that mean? Is one better than the other? It's the same percentage difference.
If you were driving on the freeway with a speed limit of 65 mph and a CHP sightsyou for speeding because he clocked you at 65.195 mph, would you be okay with that?
If the target is 0 and the forecast misses by .2, the percentage error is (.2-0)/0. What is that percentage? I'll give you a clue, it is really big.
What does all this mean? What is means is that the small group of Federal Reserve staff are awesome because they can nail the forecast that it takes a huge group of private investors to get close to.
Why, you ask, would the Fed say that they sucked when, in fact, they are awesome? Because they have really high standards for themselves. They expect themselves to be awesomely awesome. For them, being just mediocre-ly awesome just isn't good enough. They are their own worse critics.
What you have to ask yourself is if you can depend on the reporting of the Wall Street Journal. You would expect that they do the research and would present the information honestly, in a manner that learns you something that you don't have the time to study. Instead, they totally misrepresent it. Is it because they don't know better or because they are simply lying? Perhaps, they will tell you what you want to hear so you will buy the stuff that is advertised on their site. Or maybe, they just don't think that you can handle the truth.
Now that you know, are you going to write a letter to the editor and complain? I would. I sure wouldn't read news from a news agency that didn't think I could handle the truth. I want to hear the truth, no matter how bad it makes me feel.
