Why Seasonal Adjustment Defenders Are Wrong- It’s Like Fake Abstract Impressionist Art

I have been inveighing against the use of seasonally adjusted economic data for a long time. The following comments are adapted from posts I wrote back in 2011-12 regarding seasonally adjusted first time unemployment claims and nonfarm payrolls. These comments apply to virtually any kind of seasonally adjusted data, aka manipulated, massaged, or fudged, data.  Meanwhile, here’s a current example of how actual, unfudged, unmanipulated, unmassaged data should be analyzed.

Below are some of the reasons why SA data should be at least de-emphasized in favor of actual data.

November 10, 2011 – Every week the US Department of Labor dutifully reports two sets of numbers on first time and continuing unemployment claims. One set is the actual number of people filing claims as tabulated by the 50 states. The other is a fictionalized number which purports to smooth out the seasonal variances so that the trend can be represented more clearly. One number is real. The other is fake, but is like a work of abstract impressionism. It represents the statistician’s impression of reality, and for some people that works. Hey, beauty is in the eye of the beholder. But I prefer photo realism.

Most of the time the fake seasonal numbers paint a pretty picture which more or less represents the trend. But they make it impossible to see how the rate of job loss is really doing in real time. For that we need to look at the actual snapshot photo and compare it with past such photos. That way it’s possible to quickly determine whether the picture is getting better, worse, or about the same as past periods.

Using the seasonally massaged data, that’s not possible. That becomes critical at real turning points when the seasonal data, like any smoothed data, is likely to obscure or lag the turn for months. It’s why we don’t look at just moving averages of stock prices when we turn on our computer screens in the morning. We look at the actual prices, and the actual price patterns, and then maybe the moving averages. It should be the same with any economic data. The actual patterns are absolutely critical to our comprehension. Seasonally adjusted data ignores them.

While the Feds dutifully reports both sets of numbers, the mainstream financial media dutifully reports just the fake numbers. Their view is that reality is just too complicated. So they give us a cartoon instead.

January 15,2012 – Seasonally adjusted numbers frequently veer away from reality by the very nature of the arbitrary seasonal adjustment process. Conomists and the media focus almost entirely on this nonsense, which attempts to compare one fictitious number with another fictitious number to derive a fictitious month to month change.Some even go to the erroneous extreme of comparing two annualized fictitious numbers a year apart to calculate a rate of change. Meanwhile they ignore the actual data. I mean, if you’re going to emphasize the seasonally adjusted crap, you could at least check it against the trend of the real data.

The not seasonally adjusted (NSA) data is the actual data. The seasonally adjusted data is fictional, and it frequently over or understates the real rate of change of the trend, and sometimes even goes in the opposite direction of the actual momentum of the trend. The conomic establishment tries to downplay the actual data by calling it “not” data as in “not seasonally adjusted,”  which carries the connotation that somehow “seasonally adjusted” (SA) without the “not” in front of it is the real thing, when the opposite is true.

Wall Street and academic conomists cloud the facts by sticking them into a numbers grinder hoping the product will come out as a nice tasting sausage that everyone can eat. Sausage, while tasty, isn’t good for you when it’s your entire diet. And if you actually knew what you were eating, you probably wouldn’t eat it.

Are conomists too lazy (or too crooked) to do the simple analysis of comparing year to year changes and the monthly rate of change in the real data to past years to get a clear picture of where the trend is headed? Or do they just think you are too dumb to comprehend it?

It’s so easy to do when the data is placed on a chart (like below- reported here). But unlike technical analysts who actually know how to read patterns and trends in noisy data, the vast majority of conomists who have sold their souls to the Wall Street devil, have no clue how to to do that. The process is so simple. Maybe that’s the problem. They like complication because they think it makes them look more like the experts they pretend to be. In reality, the public gets it. It’s the conomists that don’t.

Sometimes SA data represents reality to some degree, and sometimes it doesn’t. If you are following that data, at any given time you have no way of knowing which it is. One thing is certain–photo-realism it ain’t. There are ways to measure trends using actual data. One is to show the year to year line as of the current and corresponding date. Another is to view the annual rate of change.

One glance at the chart shows you exactly where things stand at the present moment and in relation to the past. And it takes only a paragraph or two to describe that. It’s a simple task that mainstream would be journalists absolutely refuse to do. They do it one way and one way only, because that’s the way everybody does it, and that’s the way everybody has always done it. Independant thought does not apply.

So while I’m sure Renoir could have shown the statisticians a thing or two, as Impressionism goes, statistical impressionism is misleading often enough that it is inexcusable to rely on it to the exclusion of reporting and analyzing actual data. Mainstream pundits refuse to do it because they must walk in lockstep with their crowd.

But they’re not impressionist artists. They’re copycat clowns.