Jack Welch insists that the September jobs report, released last week, is “strange” and “implausible.” Specifically, the economy’s sluggish growth doesn’t support the reported drop in unemployment last month to 7.8% from 8.1% in August. It just “doesn’t make sense,” the former head of General Electric writes in The Wall Street Journal. The problem, he explains, is that the methodology behind the household employment survey, which is the source of the unemployment number, is less than perfect. Agreed. But why stop there? If we’re fired up about finding reasons to question economic indicators–one at a time, in a vacuum–the opportunity is unlimited for casting aspersions across the statistical landscape. This is a dead end, however, if we’re looking for deeper insight about the economy and the business cycle. Pointing out flaws for a given data series has merit, but not much. The bigger issue is deciding how to interpret the numbers in search of reasonably reliable perspective. Fortunately, the outlook isn’t as bleak as Welch’s essay implies.
It’s well established that economic data generally is noisy. In other words, it’s best to remain cautious about the number du jour. Welch and others correctly warn that we should be suspicious of the household survey’s estimate that the number of employed persons rose by a dramatic 873,000 last month. Actually, we should remain suspicious of monthly numbers generally and focus instead on year-over-year percentage changes and compare the trend in employment with a broad set of economic and financial numbers for a stronger read on the current state of the business cycle. On that score, the employment data don’t look so “strange” after all.
For example, consider how the rolling 1-year percentage change in the household employment numbers compare with the establishment series, an alternative methodology that surveys businesses as opposed to households. Clearly, the household data (red line) is considerably more volatile than the establishment numbers, even on a year-over-year basis. Over time, however, the two series generally track one another, although at times there can be relatively wide divergences.
For a clearer look at how the two data sets compare let’s zoom in on recent history: the past 10 years. Household employment last month rose roughly 2% vs. the year-earlier level. Meantime, the establishment number increased 1.7% for the year through September. We can argue about which number is a better read on what’s happening in the labor market, but both series are essentially telling us the same thing: the economy is creating jobs, somewhere in the 1.7% to 2.0% range. That’s roughly in line with the August reading. In other words, not a heck of a lot changed last month in terms of the broad trend in employment data, even if some pundits are making a fuss about the last data point.
Sure, you can focus on the month-to-month changes, or even the 3-month or 6-month fluctuations. But that’s problematic because of the short-term noise that can and does mislead us. Looking at year-over-year changes reduces a fair amount of that noise. Even so, we should still be skeptical. Fortunately, we can enhance clarity a bit further by looking at additional labor market series, such as initial jobless claims and the hours worked. We can and should track other economic and financial indicators as well. In fact, I regularly do just that on these pages via The Capital Spectator Economic Trend Index (here’s the latest update).
Not surprisingly, looking at a diversified set of indicators, primarily on a year-over-year basis, provides a statistically robust measure of economic activity. By contrast, playing up the limitations of any one indicator is the equivalent of a high school prank. If you cherry pick numbers you can see anything you want. But if you’re interested in a relatively high confidence read on the business cycle, it’s best to ignore the noise—in the data and in the media.
This post was originally published at The Capital Spectator and is reproduced here with permission.