Composite PMIs...the Best of Times; the Worst of Times-Really?

PMI data now rank observations on their range of values since December 2016. On that basis, the U.S. composite is on its high with the queue standing telling us that U.S. indicators are lower 98% of the time (this high or higher only 2% of the time). The EU and U.K. series are weaker only 8% and 6% of the time, respectively. Those extremes make this as just about the best of times and the worst of times – at least on this roughly 4-year timeline. At the same time, these extremes should also give us pause in taking PMI values or rankings too literally.

Vetting the PMIs
For example, the U.S. unemployment rate in December 2020 is at 6.7%. In September 2019, just 16-months ago, well within the four-year envelope of this data set, it was as low as 3.5%. Are ‘times’ really better in December 2020 than in September 2019? Really? EMU unemployment in November stands at 7.5%. In February 2020, just as the virus struck, it was as low as 6.5%. PMI data do compare employment situations in their details, but they are not the focus of the headlines. The overall unemployment rates tell us that taking the PMIs or their ranking literally is a fool’s errand. The U.S. is nowhere near the best it has been in four year and everyone knows that. It’s hard to understand just what these surveys are doing with the services sector in particular since it is so decimated everywhere. How could the U.S. services sector ever be assessed as better than it has been over the last four years except for 4% of the time? That makes no sense. Too many firms are shut and out of business. Too many people are jobless. I live in NYC… you cannot DINE INDOORS in any restaurant in this city. How can this be among the best of times?

Some fundamental issues
PMI data are diffusion data. They tell what proportion of a sector is expanding. I have often noted that breadth is not the same as strength. Let me repeat that here and let me also question the accuracy of these surveys in assessing breadth because their findings for the U.S. in January are ludicrous. Of course, there are also conceptual challenges those for applying diffusion metrics to data at a time like this. If a lot of services firms are going out of business- and that is happening - how does one treat that statistically? If a business is gone its output will fall the month it closes and then it will never change (since zero=zero=zero). So how or when does such an observation get removed from the data grid? And once it is removed, are the new diffusion values comparable to the old ones since they will refer to fewer businesses? In the normal course of the business cycle, these sorts of data challenges always present themselves, but at this time with so much of the services sector decimated, the challenge to the data purveyor is extreme. It is probably true that the services firms that are open are doing progressively more business as people become accustomed to the new circumstances (diffusion is rising until a new wave of the virus swirls through town as has happened in the U.K. and in Europe). But the U.S. also has a spreading virus and the strong values of the PMI in January do not seem sensible.

Europe and Japan make sense
The PMI data for Europe and Japan seem to reflect the realities we see reported in the news and about the virus. In the U.S., that does not seem true. However, two regional Fed surveys (PMI-type of just manufacturing) show very different conditions in the New York Fed district compared to the Philadelphia district in January. These are two Fed regions that are geographically quite close.

I am much more prepared to accept the survey evidence of declining conditions across Europe in the U.K. and Japan (where a state of emergency was declared in Tokyo and then extended). But the readings for the U.S. are simply hard to swallow. Maybe for manufacturing, the data are more solid, but even for manufacturing those results seem awfully strong. And late in the year since some of the government stimulus programs were running on ether or dying making a jump in output seem unlikely. But with the services sector so shuttered, it makes some sense that consumer spending would be more concentrated in the goods sectors and that would boost U.S. manufacturing as well as bloat the U.S. trade deficit. That is possible except the Markit service sector data for the U.S. show a services sector that is running hot.

What is really happening in January?
For now we have a snap shot of January that does not make complete sense. Europe seems to be in a bad way. The U.S. acts as though it has been unaffected even though data on the virus say otherwise. It’s a good time to remain a skeptic toward data. PMI data are very sensitive to shifts in economic activity and sometimes they are just too sensitive for their own good. In this instance, I wonder if they are sensitive at all at least in the U.S.

Commentaries are the opinions of the author and do not reflect the views of Haver Analytics.