This post was co-authored with Pierre Mitchell, Spend Matters’ Chief Research Officer. This is the second post in a two-part series. You can read the first post here.
For example, you can see in the above-referenced graph that month-to-month variances are “noisier” than year-to-year comparisons. Smaller sample sizes increase volatility and show inherent noise in these numbers. This appears to be especially true as sample sizes drop by varying degrees in a given month and sector. Sector-level variances then become amplified, more so when factoring in seasonal effects. Therefore, another supporting explanation for the positive September PMI numbers might stem from industries such as the automotive sector, which has also performed very well in our own monthly commodity reports (let’s remember that September showed record auto sales). This trend also appears in Fed data, suggesting the manufacturing sector received its horsepower from the “motor vehicles and parts” sectors:
Does ISM data rely too heavily on the automotive sector? We can’t say for sure, but that may be a factor. We are looking to get some more insights from ISM and other providers in the future, so stay tuned on that.
Look at the Trend, Not the Blip
As we have stated many times on these digital pages: one occurrence is a data point, two times is a line, and three times makes a trend. And just because we have anecdotal evidence that something doesn’t “feel as strong as the data suggests” doesn’t necessarily mean our gut should override the data. In fact, it is tempting for us to bring our expertise, and biases, to bear when we review the metals pricing forecasts that go against what we (and others) might expect to see. But, we use these moments to dig deeper into the causal factors and try to improve our models, rather then just overwrite them. This is a slippery slope in forecasting, and we’ll tackle this issue in future posts.
For now, let’s watch to see if ISM data supports a trend as opposed to a blip.