Many years ago, whilst still at a previous employer, I wrote one of those short, forced pieces sprung from a complete lack of ideas – you have seen a fair share of those in this forum as well. Nonetheless, and to my surprise, it spread wide and large and just last week a colleague and friend reminded me of that particular piece.
What I did was to look at the relationship between the constituents of the PMI and their “real” equivalents and then weigh together a “real PMI”. That gave me an opportunity to double-check the validity of PMI-outcomes as there is always a risk that this type of survey not only reflects actual developments but to some degree also reflects the level of confidence among respondents. In extension, if for instance the official PMI figure was considerably higher than the real PMI, it could be interpreted as respondents being more upbeat than what is warranted by actual data due to, e.g., strong stock market developments (which themselves might be right or might be wrong). And vice versa, of course.
In the first step, we look at how well each constituent is explained by its “real” equivalent. As Markit PMI is an add-on database, I have chosen to use US ISM data instead, included in the base Macrobond package, as an example. Do note that the ISM-index gives equal weighting to each constituent: Orders; Output, Employment; Delivery times, and; Stocks of purchases are each given a 20% weight, whereas in the Market PMI a different weighting scheme is utilized. For those of you who do not have access to local PMI data, similar calculations could probably be performed for other, similar, business surveys like Ifo etc.
Graph 1: The five ISM/PMI constituents and their real-world equivalents
Questions in the PMI-surveys are posed as changes in relation to previous month(s) and taking typical seasonality into consideration. That is why I am using three-month log changes in seasonally adjusted real data, standardized. As can be seen in the reports the goodness of fit (R2) is not a disaster and the coefficients (betas) are statistically significant (I have also refrained from using an intercept due to using standardized data). When eye-balling the data somewhat closer it looks as though– if anything –real data is indeed leading its respective PMI constituent.
In the second and third step, we use the betas to express real data in PMI-terms and also weigh them together with the static PMI-weights to get a real-world equivalent to PMI.
Graph 2: ISM/PMI and its real-world equivalent
Again, real ISM/PMI is demonstrating a slight lead to official PMI data but currently the data seems quite aligned as the official PMI-numbers have come down over the past few months. Anecdotally, this rhymes well with our prejudicial view that PMI data might also harbor a pronounced expectations component and should therefore be treated with some caution as not all shifts in stock markets (expectations) are fundamentally driven. Remember, Paul Samuelson’s famous quip that stock markets “predicted nine of the last five recessions”. That is also why the recent uptick in PMI-data should also be treated with some caution as “real PMI” has yet to improve.
Handle with care
Looking at the residual – the difference between actual PMI and real PMI – is quite revealing.
Graph 3: The “residual” is pure expectation
Lo and behold, the difference between official PMI and real PMI seems to be well-correlated with stock market developments (% YoY changes). When stock markets are growing on an annual basis confidence (“the residual”) seem to improve, and vice versa.
How could we use this? – Well, at least as long as you haven’t rigorously explored how the causality runs, I would be careful to put too much into this. It’s meant to be an illustration and reminder of how difficult it can be to interpret (among others) survey data, even the widely cited and used ISM-index. In all honesty, it should be every bit as useful to try and exploit the real, hard data sets that are used here or try to excavate whatever information content financial market prices can reveal. At best, the above analysis might suggest a way to improve on your ISM-forecast…
– Then again, since previous colleagues and clients obviously managed to make something out of this bland analysis a few years ago, I cannot wait to see what our talented user base will be able to make out of it this time around. Happy hunting and until next week!
 The constituents (and weights) of the PMI are New orders (with a 30% weight), Output (with a 25% weight), Employment (with a 20% weight), Delivery times (with a 15% weight) and, finally, Stocks of purchased goods (with a 10% weight).
Disclaimer: We don’t usually have views and opinions about economic and financial states of affairs, (not ones that we express publicly as a company, anyway). We do believe, however, that people can and do appreciate a variety of perspectives. What you’ve just read is the perspective of the author. While we think our writers are very smart, Macrobond Financial does not expressly endorse the views presented here. And, as the old adage goes, you shouldn’t believe everything you read (not without finding the data, performing a few analyses and presenting it in a nice chart). We want to make it clear that we are not offering this information as investment advice. That being said, if you have Macrobond, you can easily check everything that’s mentioned here, and decide for yourself. If you don’t have Macrobond, now you have a great reason to get it.