What is the Purchasing Managers’ Index™ (PMI™)?
The Purchasing Managers’ Index ™ or PMI™, is an economic indicator. It moves around a neutral level of 50, which indicates no change on the prior month. Readings above 50 signals that the economy is expanding and below 50 it is contracting – this allows you to see the economic cycle over time.
We can see here the dip down in the global financial crisis early on in this series, and more recently, another dip as the Corona virus (COVID-19) epidemic hit business in early 2020.
Why is PMI™ a critical and more accurate data set?
PMI™ was born out of a need for economic data. Other data republished by governments such as GDP, but these data suffer weaknesses. They're often published infrequently with a delay in publication. Quarterly GDP numbers, for example, only published several months after the quarter in question. Poor coverage of the whole economy as well any monthly data are largely restricted to manufacturing or retail, excluding the vast service sector. The data can also be volatile making signals hard to read. And even when they are published, they can be prone to revision or error.
Here's an example. This is UK GDP. The gray line here is the initial estimates that we see. These are the numbers that are published first by the government that you see in the newspapers, telling everyone what the economy has just done. Below this green line, these are the revised data. This is what the government now thinks happened during these occasions. We can see here some of these revisions are quite substantial. Prior to the global financial crisis, we can see that some periods of slowing growth have been revised away to show very strong expansions. More recently, since the global financial crisis, we can see fears of a triple dip recession been revised away. So, there was in fact, one recession.
Why does this discrepancy in data accuracy matter?
Revisions like this can cause huge problems. Policymakers at central banks, for example, will struggle to set interest rates effectively if they don't know what's happened in the economy. Investors and business decision makers, likewise, will struggle to make efficient decisions. Our mission here S&P Global in publishing the Purchasing Managers’ Index™ is quite simple. We want to provide the earliest the most accurate, most comprehensive, internationally comparable economic indicators available.
How do we build our Purchasing Managers’ Index™ Survey Panels?
Well, the PMI™ are actually surveys of companies polled every month. Not just any companies. We must build many replicas of the economy. And we do this by obtaining official data on the true structure of the part of the economy that we want to survey.
Here's an example: The is US manufacturing. We can see the contribution to total output in the US manufacturing sector, by categories such as food and drink, textile and clothing. And within each category, we can see the correct mix of small, medium, and large firms that we need to accurately reflect the true structure of manufacturing. Here is the equivalent data for the US service sector: hotels, restaurants, and within them, the mix of small, medium, and large service sector includes transport and storage, telecommunications, financial intermediation, the banking sector, business services, and consumer services. It's a huge disparate part of the economy. Importantly, these matrices that we use to build these survey panels we update every year to make sure they're accurately reflective of the economy at any given time. And of course, they will change from country to country. No two countries’ manufacturing sectors look the same. We take that into account when we're building our panels.
Finding the right contributors
Once we know the right mix of companies to get we approach companies to see if they would like to join the panels. This is a protracted process that can take huge numbers of man hours looking through marketing and mailing lists and trade association memberships to find willing participants. Now Purchasing Managers actually form an ideal survey base, especially in manufacturing. They have access to a wide variety of information. This didn't apply to many other managers. They will need to know when firms are taking on more staff because they'll have to buy more computers or more equipment for the staff to use. They will need to know if the company is planning a new sales push because they will have to be boosting their inventory levels. They will also know if the company is looking to cut back on capacity in times of a downturn probably earlier than anyone else will because they will have to adjust their purchasing inventory, so they form this ideal kind of base. However, we've run into problems with smaller companies and service providers. Small firms tend not to have purchasing departments and service providers where they do have a purchasing department is often buying office supplies. So these turned out to be the wrong people to be talking to in trying to get a handle on what their companies are doing on the at the coalface. So, we widen that net to executives such as finance directors and other senior manager roles — often C-suite professional.
Now within company, we try not to just have one contact. We broaden the net to encompass a number of people within each firm ideally, so that we can ensure high response rates from each company. And importantly, to retain those companies over time in case one of the executives we are contact with decides to leave the firm. This means that we can achieve response rates in the region of 80%, sometimes more than that, and retention rates, whereby we lose very few companies each year to attrition. Now, we don't pay these companies for their participation in the surveys, because we fear that paying people may induce them just to tick boxes and not really think about the answers. Instead, the survey results are provided as the incentive which forms a valuable give-get model. We know that companies use the survey results in their own business decision making and planning, and as such, it seems unlikely that they will not put reliable information into those survey results themselves. Now, for all countries importantly we use mother tongue languages. And we use locals with expert knowledge of the customs and culture. So again, boosting response rates and also forming a dialogue with companies so we can extract anecdotal information from them, as well as just the straightforward survey responses to add color to our surveys.
Collecting the data
In terms of collecting the data, is a very simple question that we ask. The one example here this is: Please compare the level of output at your company with the situation one month ago, is it higher, the same or lower? They tick just take one of these responses. Now we also asked them if there's a reason for this change-- this can give us some of that important color I was referring too. Companies might say that they've just embarked on a new sales push new marketing campaign, they may say that an exchange rate effect is helping boost their exports, for example. This gives us some anecdotal evidence we can use when we compile the analysis of these data. But the key thing, of course, is that response whether output was higher, the same, or lower than one month ago.
Creating the diffusion index
Now, each of these responses are given a weight. So, the percentage of the survey panel that reported higher is given a weight of one and the percent that was reporting the same gets given a weight of naught point five, and no weight goes to those reporting lower. We add up these responses to get what's called a diffusion index. And this varies around 50 as we discussed, because we can see from here if 100% of a survey panel said that business conditions stayed the same, that'll be 100% times naught point 5 -- 50. The upper bound is 100. If 100% of the survey panel said there was an improvement, it would be 100% times one.
Similarly, the lower bound is zero because if 100% of the panel said, things have deteriorated a lower response, that would be 100% times zero. We can trace the cycle. Any reading above 50 represents growth or improvement and the higher above 50, the faster the rate of improvement. As the index comes down, the rate of improvement slows fall below 50. And we have this decline and the further away from 50. The faster the rate of deterioration signaled when it moves closer to 50 as a slower rate of deterioration.
Reading the index
Now this is important to get right. There's lots of mistakes made by analysts seeing a PMI index move from 40 to 45 and believe that represents an improvement. It does not, it simply means that the rate of decline has eased. Similarly, an index falling from 55 to 52 does not mean a deterioration – it’s just growth at a slower rate.
Now, that was the output index we talked about, but there's actually a whole number of questions we asked companies every month. We asked them about their new orders, and of which, what were their new export orders looking like. And their backlogs of work-- this is an idea of how many orders they've not yet completed, that are waiting to be processed. We asked them about their employment levels. And we asked them about both their input prices and their output prices-- their selling prices. In manufacturing, we ask some more questions as well because we need to know more about manufacturing supply chains. These questions aren't so relevant to service sector providers. But in manufacturing, these are key indicators. We look at the amount of goods purchased by companies, and the length of time it takes for those goods to be delivered -- supply delivery times. We also asked two questions on company's stock levels. So, inventories: stocks of purchases -- that's inputs for use in production -- and their stocks of finished goods-- goods that they produce sitting in the warehouse waiting to go out to customers. All these questions ask companies about the situation current month compared to one month ago. There's one exception. That's a future expectations question where we ask companies to report on what they think their own output will be in one year's time. That's the only subjective question in the PMI surveys. The rest of these questions are based on hard metrics. They are not objective. This is not business confidence; this company is reporting what actually happened at their firms compared to one month ago for all of these variables.
Manufacturing, we compile what we call the Headline PMI. This is a weighted index of five key survey variables. These are weighted according to their ability to lead the economic cycle. New orders get the highest weight of naught point three, output gets a weight of naught point two five, employment naught point two, supplies delivery times which gets inverted in the PMI. It's a weight of naught point one five and stock purchases a weight of naught point one now those weights sum to 1.0. So, these five components get aggregated into a PMI which gives you an overall indicator of the health of the manufacturing economy. We don't produce this composite PMI for the service sector. Simply because we don't collect supplies delivered times and stocks of purchases data in the service sector. Instead for services, we focus on output as the key variable.
Collecting the data
In terms of collecting the data, we normally start collecting information from companies around the 12th of each month. Try to end as late in the month as possible, usually a few days before the end of each month.
However, rather than waiting for all the survey responses to come in, we can produce an early estimate of what the numbers are likely to look like. This is what we call the Flash PMI is released around one week prior to the end of each month, and typically includes information from around 80% of the total number of questionnaires that will normally get back for a month. This allows an early insight into those final PMI numbers. Those final PMI is published on the first working day of each month for manufacturing and the third working day for the service sector. So, as you can see, these are rapidly produced data. They are the quickest available in most countries in terms of economic indicators published.
Timeliness is the key
Now, this really helps to be able to produce early estimates of official data such as GDP having data that's produced very rapidly means we get an insight into how economic trends are changing faster than any other data. Now, here's an illustration. This is a timeline for the first quarter of 2020. This is the United States GDP timeline, if you like. So, for the first quarter, covering January, February, March, the official data don't get published until nearly May. This is what they call the advanced estimate of GDP. And then that gets revised, the quality of it the information included gets improved with a second estimate than the third estimate. And then finally, around early July, so several months after the quarter, do we get the final PMI data with a breakdown by the industry sectors such as manufacturing and services. So, there's a huge delay here with which policymakers have to wait and avoid until they can decide what's appropriate to do for the economy. Now, this is where the PMI comes into its own. Because as you can see, throughout the quarter in almost real time, we're getting information on the performance of the economy starting in January with a flush estimate, then final PMIs. Then February data come through and by the start of the next quarter, we have all three months data for that quarter for manufacturing and services. This means we can produce estimates of GDP, much quicker than official data can be produced. In fact, using the flash data around mid-month, you get a fairly good idea of what the chord is going to look like usually. So by the middle of each quarter, we're starting to get a good GDP estimate firming it up at the start of the second quarter.
This means that PMI base GDP estimates and in fact all that data, such as employment and manufacturing output are available at least a month ahead of official first estimates. If you look at this chart, by the time GDP data for industry are published, we've already got PMI data for the following quarter.
How accurate is the PMI data?
Okay, timeliness is great. But is it accurate? There's no point in being early if you're going to send the wrong signals. The PMIs have developed a really strong track record of accurately anticipating data. European Central Bank paper for example, in 2015, referred to the PMIs as the most closely monitored business surveys in the world. Bank of England paper rated the global PMI is their best single measure of global economic activity in real time.
Let's take a look at these in action. So, the left chart here shows our global PMI. So the aggregate of all of our national surveys, which is charted against global GDP, the official data shown by the bars. As you can see here, we were early in signaling the global financial crisis and indeed early in indicating the upturn from the global financial crisis. And we track things like the eurozone sovereign debt crisis, cooling global growth in 2012, and some ups and downs since then. Or recently 2018 we saw the start of global trade wars started dumping growth and more recently, of course, the corona virus pandemic, causing the most severe economic downturn that we've so far seen. GDP data records not yet available at the time. And we can also see breakdown data for example, here's our global export data on the right-hand side charted against official global export data from statistical bodies. PMI here you can see there's an 86% correlation. But importantly, not only is that correlation high, but the PMI is available three months in advance of those data. If we look at this in how it played out in the global financial crisis, this is just one illustration of how that timeliness really helps. So, the chart here, this was our Eurozone PMI that we published on the 21st of November 2008. This was the flash estimate, and it fell to a level that we hadn't seen before down-- below 40 signaling contraction at 39.7. Signaling that GDP was falling in excess of 1% quarterly rate. This is the first sign that anyone had heard that the global financial crisis was taking a big toll on the real economy. If you were waiting for the GDP data, though, it wasn't until the 13th of February 2009. So that was around three months after we publish that PMI number that you got GDP data, indicating that indeed, the economy was collapsing. So, if policymakers had waited for that official data, they would have been tweaking policy far too late for it to be effective.
In fact, there's some papers on this. If you contact your S&P Global sales team, we can send you some papers on just how to use PMIs to track GDP and other variables as well to really get the most value out of the data.
Caution also: sometimes GDP data gets published and they maybe don't say the same thing as the PMI. But as this chart nicely illustrates, sometimes that because the GDP data get revised. Which goes back to the beginning exactly why we started this venture. The chart you see here has the green line is our Eurozone PMI. And the gray bars again, are the official estimates of GDP going back 20 years now. However, the dark blue line, that's the first estimates of GDP that were published, those are the so there's the numbers that you saw in the newspapers when GDP was first reported. As you can see, those numbers get revised, it seems that they often get revised higher, but the revised data the gray bars are more in line with the PMI. So, this means that those official first estimates are often wrong. So, you can use the PMI as a better guide to what actual GDP data will be when the numbers finally get revised. So huge advantage over official data. So not only is the PMI, faster than those first estimates of GDP, but also it tends to be more accurate.
This is also something that we've seen in the US as well where GDP data get published, but they contain errors. So once again, not only are the official data late, but they can be misleading. Illustration here, I've picked out three quarters where there seems to be what we call residual seasonality depressing first quarter GDP growth. Time and time again, we saw in the US that first quarter GDP data came out very low and they bounced back in the second quarter. They weren't able to properly remove seasonality from the data, it seems. But the PMI tended to cut through the noise of that seasonality and provide a better indication of what the economy was actually doing.
Now let's look at where these data are available for now. So we're now covering every major developed and emerging market around the world. So we started back in 1991 in the UK, and that's coverage is steadily growing. We're now up to around 80% of global GDP coverage, Kazakhstan, our latest addition highlighted on the map there. Now in these countries-- 44 countries around the world now-- we're surveying amounting to over 28,000 companies each month around the world. So, this is a vast database that we are now compiling. Updating each month information from these firms around the world, allowing us to produce this unrivaled suite of national, regional and sector time series data. Extremely powerful. Here's an example. Here's some of the major countries all ranked by their manufacturing output growth. This is in March 2020. Only India and China were expanding at the time. And again, what's great about the PMI is that we don't just look at things by country, but we can look at data by sector as well. So, what we have here is a similar ranking, but this is global sectors. So, during that initial first impact of the Coronavirus pandemic, only food producers were reporting any growth -- everyone else in decline. Now we have these data obviously not just for output, but all the other survey measures such as employment prices, order books, inventories where appropriate and so forth. So, a massive database that's available to look at economic data not just by country, but by sector as well. Here is a current list of the countries which we have data available, manufacturing data, service sector data available for all major developed and emerging markets. Some economies we have what we call whole economy PMIs, notably in Africa and the Middle East. And of course, this list continues to grow.
To learn more about PMI data or the PMI index, contact us.