“The economy is showing signs of life,” advises David Malpass in today's Wall Street Journal. "The U.S. economy may finally be in a position to accelerate above the so-called new normal—the painfully slow 2% average growth rate that has persisted since 2009," the economist writes. His outlook is hardly unusual these days, but just 10 months ago in the same newspaper—in September 2012—he warned that "economic signals point to a 2013 recession." He reasoned at the time that the then-current monthly declines in new durable goods orders and real personal income "waved bright red recession flags." As we now know, the recession never came. In fact, as I’ve been discussing for quite a while now (including last week’s macro update), recession risk remains minimal. But how did Malpass (and, to be fair, a number of other economists) misread the macro signals late last year? Looking for an answer can help us decide what works, and what doesn't, in the always-crucial task of estimating the odds for recession risk in real time.
Like a number of analysts, I think that Malpass made a couple of errors in the dark art/science of estimating business cycle risk in real time, or so his 2012 op-ed piece implies. First, he cherry picked the data to make a point, suggesting that a couple of numbers told us all that we need to know for assessing the business cycle. Second, he looked at recent comparisons, which are vulnerable to statistical noise and revisions.
The antidote is quite simple, even if the computational work is somewhat technical and doesn't lend itself to short bits of drama on editorial pages. In any case, looking at a broad and diversified set of economic and financial indicators is the foundation for analyzing the business cycle in a statistically meaningful way. Second, year-over-year comparisons are far more valuable because it sidesteps the seasonal distortions and other complications that often infect shorter-term views.
In fact, on the same day that Malpass published his recession warning on September 28, 2012, I wrote "that August  wasn’t the start of a recession, despite what you might have heard elsewhere." What was the basis for that opinion? The numbers, namely, a broad, diversified set of economic and financial indicators.
There’s really no substitute for looking at a wide array of carefully selected data sets on a regular basis for developing a relatively high level of confidence for estimating recession risk odds. It's not perfect--nothing is in this realm. But as a reasonable model for limiting false signals, it's the only game in town. This lesson should be crystal clear at this point. Yet for a number of reasons that probably have little to do with economics per se, the world is inclined to look at the data du jour sans proper context.
The reality is that analyzing the broad economic trend is devilishly difficult (assuming you want to avoid getting whipsawed by the numbers). There’s so much data, not to mention opinion, at our fingertips that it's easy to become misled and overlook what’s relevant. It doesn’t help that the financial media is too often interested in publishing compelling stories, every day, about the perceived macro drama that, we’re told, conveniently arrives in each morning’s economic news updates.
But you can’t tell much, if anything, from focusing on a handful of data points, much less the four that arrived this morning. If you’re truly interested in a robust assessment of recession risk—the mother of all known risk factors—it’s crucial to focus on a process. Insight doesn't come all at once, but through time, assuming you're doing the necessary work of monitoring and evaluating a representative proxy of the overall economic trend. Keep in mind, too, that the most important information in this task may be revealed in how your recession risk estimates change through time.
Meantime, Mr. Economy (Mr. Market’s nefarious twin brother) is forever trying to confuse us on multiple fronts. There are a number of basic defensive measures you can and should take. In addition to looking at a broad mix of data sets, it’s useful to monitor key financial indicators separately for additional perspective, which is the goal behind my Macro-Markets Risk Index.
The first goal, however, should be one of estimating what the numbers tell us about the current state of the economy, based on the latest numbers. Granted, it’s easy to be misled with economic data, in part because of revisions. But this is one of the reasons for building a diversified proxy index of the economy: some, perhaps even most of the revisions, will cancel each other out through time in a broad, representative array of numbers.
In fact, that’s what my research shows. Indeed, the vintage data for my Economic Trend & Momentum indices tracks the revised version of these benchmarks rather closely through time. That’s a clue for thinking that the real-time data is telling us something fairly close to what the revised numbers will show. But you can't assume as much if you're looking a handful of data sets, which are highly vulnerable to revision risk in isolation.
With some solid intelligence about how the business cycle has evolved recently, you can take the next step by carefully looking ahead. This moves into the realm of guesswork, although the risk is relatively low if we're looking into the short-term future. In any case, the future is still uncertain, but applying robust econometric techniques for looking forward by a few months can be productive, albeit with all the usual caveats. For example, I use an autoregressive integrated moving average (ARIMA) model that estimates the missing numbers for each indicator (see my latest economic profile update for more details).
It’s really not a big surprise to find out that when you make projections for a broad set of indicators, the tracking error between the estimates and the actual data in aggregated form is minimal. In short, it’s possible to limit the surprise factor quite a lot (relative to looking at a few data sets) for the next two or three months.
We're always at risk of being wrong, of course, no matter how clever our analytical models. When the economy one day falls into recession, as it almost certainly will at some point, the volatility of revisions will probably rise and uncertainty overall will increase. The true acid test of any system is how it fares at major turning points for the economy, which can only be identified with hindsight. Meanwhile, the best we can hope for (assuming that we’re looking for high-confidence signals) is comparatively timely insight for deciding that a recession has recently started.
That’s deeply unsatisfying to most folks, which is why the world is continually flooded with business cycle forecasts that, collectively, are all over the map. But reliability is important for estimating recession risk, for reasons that are (hopefully) obvious. Still, perfection doesn’t exist. The good news is that there’s plenty of room for improvement. Every business cycle model is flawed, but not all flaws are created equal.