Was all this a conspiracy on the part of the Bush administration to make things look worse than they actually were? Hardly seems like rational political behavior.

The "problem" comes from the methodology. There’s no exact data for any of those statistics; they have to get as much data as they can and then make estimates. Part of the process of estimation uses previous trends. It’s as if we were using past performance of a mutual fund or stock to project future returns. Even though we look at the past performance, we should know that past performance isn’t indicative of future results. Just look at some of the top-performing value-oriented mutual funds in the recent bear market, like superstar Bill Miller's Legg Mason Value Trust fund (LMVTX) - the after-fee returns of which had beaten the S&P 500 index for 15 consecutive years, from 1991 through 2005. It did rather poorly last year, even in comparison with the S&P, which was horrid. Past performance is interesting, but it can disappoint. And sometimes rather viciously.

Now, just as saying that a fund on average will produce a 10% return doesn’t mean it will yield 10% every year, neither do government statistics work that way. While the methodology for each series of data is different, they all are more or less trend-following: They take past relationships in the data they can gather and use them to estimate current numbers. And -- this is important -- on average and over longer periods of time, they’re pretty accurate.

They’ll revise the data many times over the coming years, getting closer and closer to the actual numbers. For instance, I can't remember exactly when, but it was several years later that we learned we were already in a recession in the third quarter of 2000, at the very time most economists were calling for a robust economic future! (Except for your humble analyst, who was predicting a recession, and had been for some time because of the inverted yield curve, but that's another story.)

But in the short run, at economic transitions, they’re going to get it wrong, because the backward-looking data is mean-reverting. But how else would you do it? One of the keys to economic transitions is to look at the direction of the revisions. Recently, the revisions have all been negative. Things are actually getting worse than the initial data suggested. And during the last recovery, the data kept getting revised upward, especially 6 months and one year later.

The Fault, Dear Brutus, is Not in Our Stars

Look again at the very useful chart above (great work, wish I had thought of it!). Non-farm payrolls, which for some odd reason everyone pays attention to, are especially wrong at the turns. Anyone trading on non-farm payroll data deserves the losses they’ll get.

One of the reasons that non-farm payrolls are so often revised is that the Bureau of Labor Statistics (BLS) is forced to estimate the number of new businesses being created each month that are simply under the radar screen of government statisticians. This number is called the birth/death ratio. You couldn’t create a useful payroll number without this estimate, yet it’s simply a wild-eyed guess based on past trends, which by definition, we know will change at economic turning points.