I have written before about the difference between precision and accuracy, and the challenge of distinguishing signal from noise. With the constant barrage of economic and stock market data, much of which is taken to ridiculous levels of precision—one part in a thousand, for example, for the unemployment rates—the problem is particularly acute. What makes it even more ridiculous, of course, is that the figures will often be revised substantially. Precision is an illusion in this area.
This isn’t anyone’s fault; it’s just the reality of collecting messy data across a massive economy in a massive country. It is remarkable that the statisticians do as well as they do.
I mention this today because, over the past couple of days, some stats have come out that have been interpreted as showing a renewed slowdown. Specifically, the new unemployment claim figure came in higher than expected, and housing starts came in lower. Other bad news included a decline in industrial production and a decrease in capacity utilization.
Taken together, these stats suggest a slowdown or worse. In context, however, which is how they must be considered, things don’t look so bad. Housing starts, for example, did dip, but quite possibly due to labor and material shortages. That’s actually a good thing, suggesting that other industries will have to step up production and hiring. Not only that, permit applications went up substantially, suggesting that future construction is likely to continue to increase.
For unemployment claims, the figure did tick up—but from very low levels per recent history. Moreover, based on multiple employer surveys, hiring is expected to continue to increase, and anecdotal evidence suggests that employers such as Wal-Mart will have to start hiring again. The unemployment figure isn’t great, but, in context, it’s far too early to get worried.
The decline in industrial production is something to be more concerned about, as much of it represents a decrease in foreign demand due to economic weakness in Europe and elsewhere. Looking at it more closely, though, much of the decline can be attributed to warmer weather, which decreased utility output. If I don’t have to pay to heat my home, that actually helps other areas more than it hurts industrial production. Nonetheless, it’s something to keep an eye on.
Another example of a statistic that needs context is the April retail sales report, with a headline increase of 0.1 percent. Not good, right? Well, if you take out the decrease in gasoline sales, which was due to prices dropping substantially (a good thing), the increase was 0.5 percent, which is excellent. Not only that, the prior two months were revised upward as well. Overall, this number was a heck of a lot better than it looked.
And that’s the point: we need to look at economic statistics in context and with an understanding of the weaknesses. Trends are important, as are the underlying stories.
I want to close with a number that’s a good illustration of how I look at a data point. Today’s consumer confidence survey by the University of Michigan came in at an almost six-year high. The actual number doesn’t matter, per the above arguments, but the trend and the underlying reasons do. A six-year high means back to 2007, which is pre-financial crisis. Think about that.
In the spirit of looking at the full picture, though, a big part of that increase was due to lower gasoline prices and the strong performance of the stock market recently. If either of those was to change, and they easily could, confidence would be at risk. We continue to face uncertainty, but the trend remains positive—and that is the relevant signal amidst the noise.
When the time comes, I’ll be among the first to start pointing out things to worry about. (They don’t call me Eeyore for nothing.) In fact, I’m already doing so for the stock market—see about 50 percent of my recent posts. As far as the real economy goes, though, I believe the signal still indicates continued recovery, but I will continue to look for reasons to change my mind.