The Popular (But Risky) Habit Of Cherry-Picking Economic Data For Business-Cycle Analysis

Fed Chairwoman Janet Yellen’s Senate testimony yesterday suggested that the central bank is still on track to hike rates later this year, courtesy of the stronger economic trend. But the big day may arrive later than previously assumed. It’s all about the data. “Provided that labor market conditions continue to improve,” the Fed will raise rates when it’s “reasonably confident that inflation will move back over the medium term toward our 2% objective,” she advised. But while the official line tells us to prepare for policy tightening at some point, there’s more talk in some corners that the business cycle is looking shaky again. Huh?

Recent housing data is the leading example at the moment, including yesterday’s monthly update on housing prices via the December report for the S&P/Case-Shiller Home Price Indices. Although prices for the widely followed 20-city index rose through the end of 2014 at a moderate pace and beat expectations, David Blitzer, chairman of the index committee at S&P Dow Jones that publishes the figures, says that “the housing recovery is faltering.” He explains that

while prices and sales of existing homes are close to normal, construction and new home sales remain weak. Before the current business cycle, any time housing starts were at their current level of about one million at annual rates, the economy was in a recession.

Blitzer adds that the soft housing data has been accompanied by “favorable conditions elsewhere in the economy: strong job growth, a declining unemployment rate, continued low interest rates and positive consumer confidence.”

But there are other dark clouds that, for some, look ominous. A Bloomberg story this week, for instance, noted:

A gauge of growth rates for raw materials including cattle hides, tallow, plywood and burlap has been signaling economic contraction since September. The last time the growth rate of the JoC-ECRI Industrial Materials Price Index was falling to these levels, the world was mired in recession. At the same time, the price-tracking Bloomberg Commodity Index is near a 12-year low, with bear markets for more than half of the 22 items it measures.

Once again the message seems to be that if you torture the data long enough, you can force it to say anything you want. The danger, of course, is that we’re cherry-picking the numbers to prove (or disprove) a particular outlook. Tell me what you want to say and I’ll provide the relevant mix of supporting numbers.

Yes, wobbly real estate and falling commodity prices have been associated with recessions in the past. Some economists even go so far as to conclude that “Housing IS the Business Cycle.” Meanwhile, soaring or tumbling oil prices have been linked with major turning points in the economic trend, thanks to lessons learned after the economic crisis that followed Opec’s engineered shortage in the early 1970s. Another popular quip is that copper has a Ph.D. in economics.

These and many other relationships with the broad economy are worthy of attention, but not in isolation. It’s tempting to think that we can summarize economic risk in one or two indicators. But that’s assuming too much. Even the yield spread between long and short Treasury yields (or the Fed funds rate), as valuable as it’s been in the past for predicting recessions, shouldn’t be elevated to the status of infallible. Everything is destined to be wrong at some point, and in fact that’s what history generally shows. The key point, however, is that the list of what works, and what doesn’t, will evolve through time. Alas, how and when these shifts unfold and render an indicator useless–or germane–is unknown in advance.

The solution (or at least the lesser evil) is to monitor a broad, diversified set of indicators for estimating recession risk. Why? For the same reason that you wouldn’t use four stocks to measure the broad trend in overall market. Fortunately, there are a number of robust possibilities for evaluating the macro picture, starting with the Chicago Fed National Activity Index. Another choice that’s updated on these pages: the US Economic Profile, which monitors business cycle risk across a spectrum of indicators.

The case for a diversified basis for assessing the state of the economy’s forward momentum at this point should be a no-brainer. The last several years have witnessed several high-profile predictions of recession come to naught, largely due to embracing misleading signals born of cherry-picking the numbers in some degree. Another (and related) pitfall is rushing to judgment. There’s great drama to be mined with breathless analysts telling us that the lastest data point from 20 minutes ago is a game-changer. But a sober review of who’s made reasonably good calls on the business cycle–and the underlying methodologies–over the years isn’t kind to attention-grabbing headlines and feverish pronouncements on macro trends.

Some analysts like to make a case for picking out one or two indicators that, for historical reasons, seem to tell us all that we need to know. But every recession is different because every expansion is unique. The business cycle through history rhymes to a degree, but it evolves as well, sometimes radically so. This isn’t opinion; it’s fact, as I detail in my book on monitoring the business cycle for timely clues about recession risk.

The incentive to dismiss this advice is linked to the hope that we’ll see major shifts in the macro trend earlier and/or more clearly by looking at fewer variables. An intriguing idea, but it’s misguided. The truth is that no one really knows what will trigger a recession the next time. Looking to the past is flawed, but it’s the only game in town. How we decide to review the past matters, though.

If we know which indicators will be relevant, and when, the solution for real-time analysis would be obvious. Unfortunately, it’s never really clear what stress points will be crucial the next time, or when. The next best thing is keeping an eye on a proxy that draws on a broad set of cyclical signals. The goal is developing timely signals that are relatively reliable. That’s tougher than it sounds, in part because there’s a direct tradeoff between the two. It’s easy to develop timely signals, but without the reliability factor it’s all rather meaningless–and vice versa.

The only way to maximize the relevance on each front, at the same time, is to analyze a carefully selected mix of indicators. That’s a yawn if you’re looking for snappy headlines, but it’s the only solution if you’re trying to minimize the potential for surprises in the art/science of business-cycle analysis. We’re all going to recognize the arrival of the next recession after the fact–that’s fate. The question is how long is long?

On that note, recession risk is still quite low, based on a broad review of the published numbers. Yes, the housing market may be an early sign of trouble; ditto the weak trend in commodity prices. But there are many positives one could point to as well. But it’s ridiculous to get into a tit-for-tat debate–my indicators are better than yours! The emphasis should be on developing superior multi-factor business cycle benchmarks. Fortunately, there’s no shortage of efforts on this crucial work. The bad news is that you’re not likely to read about it unless you’re looking beyond the usual suspects.