US Services - Two Poor Indicators

So who are we to believe? October's ISM non-manufacturing composite index rose 2.1pts to 54.7, suggesting a slight improvement in the service-sector economy; but Markit's services PMI fell 0.3pts to 50.6, which was the worst result since February 2016 and suggested the service-sector economy is almost at a standstill.

Both cannot be right, and since services account for almost exactly 60% of US GDP, confidence that we understand the current direction would be useful.

Fair disclosure: I distrust Markit's PMIs because of the way its PMIs, particularly for Eurozone, Germany and France, so regularly evade published consensus, thus generating lucrative headlines. No other data-series that I track escapes published consensus so regularly, and for someone who works with statistics all the time, the track record is, shall we say, sufficiently unusual to demand explanation.

First, are the two surveys results usually consistent with each other? Since 2012, the ISM non-manufacturing composite has averaged 56.1, it's average monthly movement has been minus 0.01, with a standard deviation of 1.94. During the same period, Markit's services PMI has average 54.6, its average monthly movement is minus 0.08, with a standard deviation of 1.86. Similar metrics so far, then. However, when we look at those monthly movements, we find there has been no significant correlation in the way they move (a correlation of 0.07 over 102 observations). The null hypothesis holds: what the ISM reports bears no interesting relation to what Markit finds.

So maybe one is right and the other wrong? We can test this by looking at the annualized changes in service-sector GDP and comparing it with quarterly changes in the two indexes. Which is the more reliable directional indicator?

The answer is neither. The ISM non-manufacturing scores a correlation of just 0.002 since 2012, whilst Markit scores just 0.06. In other words: they are not signals, they are noise.

A final point. Though both are noise, from an eyeball point of view, Markit's services PMI looks slightly more plausible than the ISM. But who am I going to trust: the data or my own lying eyes?