PMI as a Leading Indicator

TL emailed me and asked:

What are you thoughts on using PMI indexes to measure economic conditions vs GDP? (See below video) https://m.youtube.com/watch?v=ydBzV9KTF0Y

**********************************************

So, what’s PMI (in the context of macroeconomics, there are other abbreviations too).

PMI is short for Purchasing Managers Index. It is produced by the Institute for Supply Management (ISM). It’s based on a survey of purchasing managers in manufacturing about how they feel about their supply chain: is it indicative of a stronger economy, or not? It’s an index that’s scaled so that when it is above 50 the economy is expected to expand, and to contract when it’s below fifty.

In principal, this is probably a good group to survey. They should have a feel for how the economy is performing at the wholesale level, a month or two before that hits the rest of the economy at the retail level. So it ought to be a decent leading indicator.

The video is informative about all of this. But I’d say it’s careful to note that it isn’t meant to think critically about PMI.

PMI would be useful for forecasting if it was a leading indicator. Leading indicators are variables that match the patter of business fluctuations in the overall economy, but which lead it in time. For example, when the leading indicator peaks, the economy is expected to peak shortly after.

**********************************************

I replied with:

I think PMI is worth close to zero. But not zero.

In order to be useful for forecasting it would need to lead the business cycle. In order to be useful to use alongside GDP it would need to be coincident with the business cycle.

If I go to Google and enter "is pmi a leading indicator" I get this https://www.google.com/search?client=firefox-b-1-d&q=is+pmi+a+leading+indicator But a quick look confirms these are all stock market tip sites that are probably selling something. That's not encouraging.

The Conference Board is the think tank that keeps track of the Index of Leading Indicators. They have a list and the PMI is not on it. Another index created by the same group (ISM) is on their list. So that would work better. Even so, leading indicators have a poor record of forecasting turning points. Better than nothing, but ... not much better.

The problem with using anything like this is false positives and false negatives. All indicators give a ton of them. Good ones give less.

This is not to say that early information on how the economy is doing wouldn't be useful, since GDP is produced with a lag. But experience suggests there are no magic ones.

And, there's a huge problem with people looking at the numbers after the fact and assessing whether they're useful. This is sometimes called a post hoc ergo propter hoc fallacy. They do this towards the end of of the video you linked to. The real test is always can you state what the indicator says about the economy before it happens, and then confirm that it does?

I added this later:

One other thing, if you go to Google Scholar, and put in purchasing managers index, you will find stuff, but none of it is very recent, and none of it is from the bigger journals.

P.S. When you do macro, you get pretty used to thinking a new thing you've discovered is going to work, only to have your hopes dashed. The world's just a big, complex, weird thing to predict.

Get used to that. It doesn’t mean that it isn’t fun or interesting. Nor does it mean that it isn’t useful, or that you can’t make a killing by being better at it than others.

But, what’s left unsaid is that many novices impression of what should be predictable is based on success they’ve had with predicting a few things in the past. But, what one learns with experience is that there are some things that are inherently less predictable. Short-run and medium-run business fluctuations are one of those.