Services PMIs Show Broad Weakness in February As Do Composites

The Composite PMI and the Service Sector The PMI readings globally are not as comprehensive a set of data as for manufacturing. Still, there is a broad rather representative group of data we can observe to track the overall PMI and the global service sector. In January, among the twelve reporters of service sector data, eight weakened showing that weakening members outnumbered strengthening members two to one. That is decisive. In December, nine members weakened month-to-month. That compared to eight weakening in November.

The service sector globally These monthly changes demonstrate (data not shown separately) that the service sector has been under siege over the last three months with declining sectors outnumbering advancing sectors by a factor of at least two to one for three months running. That is 'impressive' in a negative way.

The chart shows that among the countries and the EMU region whose data are plotted there, the U.S. has been a very different animal with the service sector building to a crescendo while the other service sectors ran either a more restrictive cycle (like the EMU) or simply waffled while moving mostly sideways (Japan shows a bit more uptrend than the EMU or China).

The service sector ranks below its median (on data from January 2018 to date) in eight of twelve sectors with those below their median outnumbering those above their median by two to one again. The relative strongest service sectors are in Brazil (83.7%) and Canada (72.2%). Among the world's four largest economies (the U.S., China, Japan and Germany, the strongest standing for a service sector is Germany at its 36th percentile). Among the twelve global service sectors, eight of twelve have weaker PMI values than their level before the Covid virus stuck in January 2020 (one country, Brazil, is unchanged). The only countries with higher service sectors on that timeline are Canada, France, and the U.K.

The Composite PMIs The service sector usually dominates the composite reading but the composites are more comprehensive, and more countries report a composite PMI than report both individual sectors. Twenty countries report an up-to-date composite PMI in the table.

In January, the composite PMI slows month-to-month in 16 of 20 jurisdictions but dips below 50 (the diffusion boom-bust line) in only six (30% of reporters). The median reading is 51.0, a skinny gap between the median and the boom-bust line.

There has clearly been a worsening in the last two months when the proportion of reporters showing deterioration has risen sharply and stayed high. This is probably a result of the highly transmissible Omicron virus, although some health experts are now concerned that Omicron may not have spread as widely as initially suggested and there may still be a good deal of Delta in the mix. This just points out how much health authorities are stabbing in the dark at a moving target. The U.K. does a great deal of detailed testing. The U.S… not so much, and the tests that the U.S. deploys often only test for 'Covid-19' not for the particular variant. And lot of what we 'know' about the virus is still derived from models and if there is anyone who knows how dodgy depending on a model can be, it's an economist. The initial 'model results' given around Christmas by a U.K. group for the spread of Omicron in the U.S. appears to have been 'overstated.' So, we will have to listen to the health authorities to see what they tell us. Whatever is going around, it is spreading fast and it may be a mix of Omicron and Delta.

A world of 'hurt' Whatever is going on in the world of virus, it is affecting the world of economics and has had a large impact over the past two months. Infection curves are now dwindling (GOOD NEWS!) and although deaths are low relative to infections the infections have been so broad-based that in raw numbers the deaths have been high.

Virus impact on economy Obviously, what happens next is going to depend on what the real virus facts are and where we go from here. The virus has an outsized impact on the service sector since that sector puts a premium on face-to face contact and people who are engaged in heavy mitigation strategies simply avoid as much contact as possible. They stay home. They let other people shop for them. They use the internet, and so on… I live in NYC on the Upper West side of Manhattan, a densely populated area. I see a less grocery store shopping, less traffic on the streets, fewer people on the street, a less crowded subway system. People are mitigating or maybe migrating or even hermitting. Even though they still shop, that behavior hurts growth.

Diffusion data, queue rankings and high-low percentile readings The global composite PMI data show several interesting trends. I just wrote on the deterioration in the last few months. But note the queue standing column in the table…what is going on there? An average standing of 43% means that on average reporters are significantly below their median (medians occur at a queue ranking of 50). Now this is different from the median of the diffusion data which is at 51 and shows a very small tendency to expand (PMI values above 50 signal expansion; values below 50 signal contraction; on the queue ranking data 50% identifies the MEDIAN value of the underlying diffusion value). But these two readings are not incompatible -in fact together they enhance our understanding of events. As a final matter, the column labeled percentile provides the percentile standing of the month's observation in its range- between the sample high and low. A 50% reading on that is simply the middle of the high-low range.

Making the metrics work together One of these metrics, the median, points to a barebones expansion; the other (queue standings) says that countries are posting results well below their historic medians. These two findings are quite compatible; in fact, a barebones 'skinny' PMI level just above 50 is also below most nations' medians (in almost all cases). We can also see that the percentile column shows an average across reporters of 78% and a median of 82%. Again, that is compatible with the other two results. What the table shows is that there is only one reading in the top 10 percentile of its historic high-low range of values (Sweden). However, there are 12 of 20 readings that are in their top 20th percentile on this gauge. While there may be broad queue percentile standing weakness, there is not deep high-low weakness.

A slowdown in the making... If you stop to think about this, it makes a lot of sense. And what it tells us is that despite all this below median queue standing readings, we are not necessarily devolving into recession – maybe a period of sluggishness- but nothing more. The queue standings use ALL the observations to create a rank of this month's 'number' among all numbers since January 2018. The percentile of the high-low only places this month's number as a percentile of the range, the high-low range: it uses only three numbers to do this, the high, the low, and this month's number. The question is when you fill in the blanks on the high-low and rank this month among all observations, how does the distribution of values from 2018 to date look? Well, PMI data tend to cluster on both sides of 50 with the median above 50 - since growth is the order of the day. The distribution of the PMI data appears to be roughly 'normal.' A reading may be low relative to this clustering but may not have to fall very far to have a low percentile standing; that drop depends on the density of the distribution in the neighborhood of the current value. On the other hand, when recession or deep or a real slowdown comes, diffusion values generally fall by much more. So, what is it that we are seeing here? It's just a significant deviation from normal not necessarily anything more than a significant slowdown. It could become a more corrosive episode, but it is not signaling that right now.

An illustration Think of it this way. Your normal heartbeat is in a certain range. But when you exercise the heart rate goes up and it may go up quite a lot. But that elevated heart rate is not part of a problem even though an elevated heart rate might also precede a heart attack. Some variations in heart rate are normal and even desired and they accompany exertion. It is the same way with the economic data. No country runs as a set PMI or GDP growth rate every day, every quarter, every year and not all deviations from 'normal' are trouble.

Evaluating our data However, stepping away from this analog, the question is this- what kind of deviation is this? Our three metrics show us the overall median value is low – but there is still growth and little contraction in the system. The number/proportion of jurisdictions slowing over the last two months is broad. Their collective PMI ranking is low (and we can see that it is no statistical fluke- the individual country measures/rankings are low). But that PMI measures have not fallen as low in their range as they do pre-recession. So, we do have a slowing and we can connect the dots to an event of Covid-19. Since the infection curves broadly are getting lower, we can expect that this will be a self-correcting issue next month or in two months as the virus has let up and business has gone more back to/toward normal. The metrics and our knowledge of the cause of slowing lead us to treat this as a less than a worrisome event.

Ahead! Onward! Upward? But what we don't know is how durable the recovery will be. Is this the last cycle under Covid-19? Are we able to head back to normalcy? Is Omicron mild enough and has infection spread broadly enough that immunity will protect us from a repeat of the past several months at least as far as the severity of some the infections is concerned? Frankly, we just don't know. But, for now, we see the light at the end of the tunnel, and it looks a lot like daylight. Let it shine.

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