By Jeffrey P. Snider
The Federal Reserve carefully notes that its Labor Market Conditions Index (LMCI) is subject to revisions in as much as its entire history almost every month. That is the nature of the project trying to tie together 19 separate data points into a coherent yet comprehensive whole. Part of the inherent recurrence of revisions, however, is simply the pace of all the data among those 19 individual series. Some of them, including those within the JOLTS release, are unavailable by the time the Fed puts together its monthly LMCI estimate.
The index thus predicts what the lagging data will be using a Kalman smoother. Kalman filters have been around for decades and are a recursive solution to discrete filtering problems. It belongs to the class of linear prediction methods, assuming that variables are random and Gaussian distributed. While originally meant for digital control systems in engineering settings, such as the Apollo program, Kalman-Bucy filters have been adapted in a wide range of situations because of the efficiency in extracting useful signal from among very noisy data sets.
Economics being what it is, however, on a high frequency time frame such as monthly data the best of the techniques is still noisy. Two months ago, for the month of May 2016, the initial LMCI was down 4.8 points but has since been revised to -3.6 and now -3.2. Last month's -1.9 was revised this month to nearly zero, just -0.1. The latest reading is the first positive since December, +1.0. Given the big jump in the headline payroll figure for July, following the bigger, upward revised gain in June, this isn't surprising. The question is whether it is meaningful.
In the last two "cycles," including the Great Recession, the LMCI turned negative in its 6-month average just as the BLS estimated the cyclical switch from job gains to job losses. This time, however, the BLS figures only a (second) slowing from 2014's "best job market in decades."
In other words, it is more corroboration that the economy did, in fact, slow even to the point of likely affecting the labor market. We just cannot, by these statistical creations, determine just how much labor growth has slowed or from what starting point it did. Of even less accuracy is what that might mean in July 2016 as it relates to the trend or, as economists would have it, change in trend. For now, the 6-month average is still negative at -2.
If there is one consistency to this slowdown economy, it is certainly that there is no linearity to the overall deceleration or contraction. Uneven remains the dominant feature. Despite that, weakness above and beyond the "typical" slower and smaller is still evident especially since last summer. The economy got much worse last year, and there is scant evidence, even in labor figures, that it has gotten any better despite that dropoff. Whether or not that is cyclical is unclear, but it does at least appear, for once, consistent with a broad survey of other economic accounts.