Looking For Macro Signals (And Ignoring The Noise)

|
 |  Includes: DIA, SPY
by: James Picerno

The Big Picture's Barry Ritholtz reminds us that most of the media frenzy that surrounds the monthly payrolls report is misplaced. "The jobs report is overrated ... it's meaningless," he tells Yahoo Finance. "All we care about are the long-term trends: Is the economy creating jobs? Are wages going up or are wages flat? What does this mean relative to inflation?"

Agreed, although expecting that the crowd will get religion on this point is on par with assuming that you'll wake up on Mars tomorrow. Behavioral issues aside, the numbers speak loud and clear, which is why I focus on year-over-year comparisons for payrolls and other macro indicators in search of robust signals. As an example, let's consider private non-farm payrolls and how the numbers change for monthly vs. annual comparisons using vintage (initial release) and revised data.

In the first chart below, it's clear that there's a big difference between the initially reported payrolls estimate (blue line) relative to the revised numbers (orange line) over the past 40 years. The lesson, of course, is that the real-time release of the latest data point isn't all that valuable if we're comparing it to the previous month. In fact, it's best to assume that the month-to-month volatility is noise.

A more useful perspective is reviewing the year-over-year changes. As the next chart shows, the noise factor is substantially reduced for annual percentage changes in terms of vintage vs. revised data. This is clear when we compute correlations between vintage and revised numbers for the monthly and annual data sets. The correlation between the annual series of vintage and revised numbers is quite high: 0.96 (a perfect positive correlation reading would be 1.0). That's a sign that the vintage annual changes are a pretty good estimate of what the revised numbers will eventually show. By comparison, the equivalent correlation reading for the monthly data is quite a bit lower at 0.69. In other words, it's best to ignore the monthly volatility in the quest to develop useful intuition about the business cycle.

To be sure, you still can't assume that the latest annual change is written in stone. Revision risk can never be completely removed. As a result, the potential for being misled is always lurking in the shadows. But here too there's a way to minimize the noise risk a bit more: analyze a broad set of economic and financial indicators, which is the basis for our Economic Trend and Momentum indices.

The practical value is that you have a much better chance of estimating the macro trend -- the signal -- by focusing on year-over-year changes across a spectrum of data sets. To that point, consider that private payrolls have consistently told us that the labor market has been growing at roughly 2% a year. That's been a steady trend in the vintage and revised estimates over the past two years. In fact, the annual pace picked up slightly to 2.1% in the July report, as I discussed a few weeks back.

None of this makes for exciting headlines or provides the necessary drama that's highly prized among the talking heads in TV-land. But if you're interested in boosting the odds of finding the signal in the data, there's plenty of room for progress relative to the crowd's usual obsessions.