Back in February, I threw cold water all over the idea that the Institute for Supply Management’s closely watched monthly surveys of U.S. manufacturing are worth closely watching. Today the ISM’s latest report (for October) came out, so it’s a good occasion to see if the organization’s performance has gotten any better.
The short answer is “No.” But let’s refine the question in two ways. First, since it’s not reasonable to assume prima facie that government manufacturing data is always or usually the reality check, let’s change the question to “How well do the ISM results match up with Washington’s official data?” Second, rather than simply measure the ISM’s headline figure with the Federal Reserve’s statistics on manufacturing production, let’s compare the ISM’s specific output figure with those Fed output figures, and the ISM’s employment numbers with their counterparts from the Bureau of Labor Statistics.
Unfortunately, however, the ISM’s more specific gauges of manufacturing’s health stack up just as badly with federal government output and payrolls numbers as the headline – which incorporates many measures – stacks up with Washington’s production numbers.
My Marketwatch.com article looked at the data all the way back to 1993, but to get this post up sooner rather than later, I’m restricting my analysis to the current calendar year. Still, the results should worry anyone concerned with domestic manufacturing’s future – which should include everyone concerned with the U.S. economy’s future.
The gap between ISM data and U.S. government data is especially wide for production, and the problem began at the very beginning of the year.
Most everyone now agrees that the last harsh winter played some role in depressing overall economic activity in America, including in the manufacturing sector. The Federal Reserve production figures conform with this assessment, showing a 1.03 percent drop in inflation-adjusted manufacturing output between December and January. But the ISM January output number showed solid expansion that month, with a 54.8 reading. (Any ISM number above 50 indicates expansion, a below-50 reading indicates contraction.)
Most everyone also agrees that, once the worst of winter was over, overall economic activity, including in manufacturing, began recovering. Again, the Fed figures compare well with this assessment, showing a sharp 1.34 percent monthly growth rate for February. But the ISM output figure reported a significant manufacturing production drop – with a 48.2 result.
The divergence continued in March, though not so dramatically. The Fed figures showed manufacturing’s monthly real growth remained strong, but the expansion slowed to 0.89 percent. But the ISM showed a huge bounce-back from contraction to a 55.9 reading.
In April, May, June, and July both gauges reported roughly comparable results, both in the direction of the basic trend and in the magnitude of change. But problems reemerged in August. According the Fed, after-inflation manufacturing production shrank that month for the first time since January – by 0.46 percent. But according to the ISM, manufacturing output clocked in at 64.5 – its best monthly reading of the year.
In September, the Fed said manufacturing production recovered, growing by 0.47 percent in real terms. (This figure is still preliminary and will be revised.) The ISM reading improved as well – but only to 64.6. We won’t get the October Fed figure (along with that September revision) until November 17. For the record, the ISM manufacturing output figure released this morning was 64.8 – yet another 2014 record.
The ISM employment readings correlate better with the Bureau of Labor Statistics data on manufacturing job gain and loss, but questions still arise. And firm conclusions are a little harder to draw because, at least according to the BLS, the actual job changes month to month have been so small lately. But you be the judge.
In January and February, the ISM payrolls numbers were identical – 52.3 results that indicate so-so growth. The BLS showed that manufacturing employment in those months increased by 9,000 and 20,000, respectively – a strong acceleration in percentage terms but not in absolute terms.
The March and April comparisons between the two were good, but weakened some in May and June. For both months, the ISM reading was 52.8 – again, in the so-so-growth neighborhood. But the BLS figures showed quickening growth – from 15,000 to 21,000. More encouragingly, the BLS’ absolute job gain levels correlated decently with the absolute levels of the ISM labor indicator.
July’s match-up wasn’t nearly as good. The ISM payrolls indicator jumped all the way from 52.8 to 58.2. The BLS July job gain rate increased, too – but only from 21,000 to 24,000.
The August and September BLS figures are still preliminary, but so far, they differ significantly from the ISM results. For August, the ISM jobs figure fell, but only from 58.2 to a still strongly expansionary 58.1. But BLS claims that 4,000 manufacturing jobs were actually lost that month. For September, ISM reported somewhat weaker manufacturing job growth, as its reading dropped to 54.6. Yet the BLS reported that August’s 4,000 job loss had turned into a 4,000 job gain in September.
I still prefer the government data for several reasons. Mainly, they try to measure actual performance rather than reports of perceptions. And largely as a result, they’re not so distorted by the survivorship bias I’ve discussed in previous posts. But there’s no doubt that the above numbers make at least one conclusion clearer than ever: Over any serious length of time, you can reasonably trust either the federal government’s manufacturing data or the ISM’s. But you can’t trust both.