There's Another Recession Out There Somewhere... Now & Forever

Lakshman Achuthan of Economic Cycle Research Institute toured the TV circuit again yesterday to revive and defend his firm's long-standing forecast that recession risk is high (see interviews on Bloomberg, and Yahoo Finance). He asserted that the recession started several months ago, noting that this past July marks the peak in the current business cycle for the U.S. The supporting evidence for this analysis, he explained, is patently clear in the behavior of three indicators. It all sounds plausible, but there's plenty of room for doubt too. The main problem is the ambiguity of the model, as it was outlined. Transparency and clarity regarding the underlying process are essential in business cycle analysis, particularly if you're arguing in no uncertain terms that the forecast is a virtual certainty. Essential, that is, if you're trying to evaluate the legitimacy of the prediction. Unfortunately, those features were in short supply in yesterday warnings.

It doesn't help Achuthan's position that ECRI's recession forecast has been kicking around since September 2011, when the firm warned that "the U.S. economy is indeed tipping into a new recession" and "if you think this is a bad economy, you haven't seen anything yet." By most accounts, the recession has yet to arrive. But Achuthan begs to differ and says the proof is now visible in three indicators: industrial production, personal income, and sales (manufacturing, wholesale and retail). As an accompanying ECRI essay published yesterday explains:

Reviewing the indicators used to officially decide U.S. recession dates, it looks like the recession began around July 2012. This is because, in retrospect, three of those four coincident indicators – the broad measures of production, income, employment and sales – saw their high points in July (vertical red line in chart), with only employment still rising.

The "tell-tale chart," according to ECRI, is as plain as the nose on your face. Judge for yourself:

Note that employment isn't signaling a recession, as Achuthan readily concedes. But the other three indicators have peaked, he says, and so the writing is on the wall. July 2012 was the turning point.

The problem is that it's not obvious how ECRI came up with this analysis. Yes, it appears that industrial production, income and sales have peaked. But if you spend any time looking at the historical records of economic indicators you know that peaks in the data come and go fairly regularly without the start of recessions. How can we filter the numbers to elicit clearer signals? There are several methodologies that can enhance clarity, even though nothing is flawless. We can also remain skeptical of what's likely to lead us astray, starting with the shaky notion that looking at indicators in recent history, without benefit of crunching numbers for relative changes, is a short cut to clarity.

Note that the data presented above in ECRI's chart is the actual values of the indicators and not changes. Evaluating data series this way in search of business cycle insight is problematic for several reasons. Unfiltered data of this nature often looks menacing when it isn't. For example, consider industrial production's 10-year history. As you can see, there were several times when it looked like this indicator was rolling over only to resume its upward bias. To further complicate matters, industrial production was trending higher in late-2007, just as the last recession commenced.

This challenge is endemic to most, if not all economic and financial indicators. One way to filter out some of the noise is by looking at year-over-year changes. Let's take another look at industrial production on a rolling one-year-percentage-change basis. In the next graph below, industrial production seems to show a cleaner signal around and below the zero mark in connection with recessions.

The signal is a bit clearer, but there's still the issue of timeliness. What to do? Plugging a carefully selected array of indicators into a diffusion index and tracking the broad trend—primarily on a year-over-year basis, albeit with some exceptions—enhances the aggregate signal a bit more in terms of clarity and timing. It's still a lagging signal, but only marginally so if you do the analysis correctly. The tradeoff is that you have a higher level of confidence in the data analysis. In fact, this approach is a regular feature on these pages via the updates of The Capital Spectator Economic Trend Index (CS-ETI), with a recent update published here.

CS-ETI is no silver bullet, although it offers a solid starting point for deeper analysis—a baseline, if you will, for business cycle research and evaluating the risk of recession. Even so, bit of humility helps.

One of the big problems in the grand scheme of economic predictions is the assumption that recessions are always predictable events well in advance. If we're looking out a few months, there's a reasonable argument that risk can be assessed with some degree of confidence. For example, I run a number of econometric applications that analyze the economic trend from several angles for clues about where the data's probably headed in the next one to three months. This is useful as an exercise for assessing risk, particularly if you track and analyze the inevitable errors.

But what you can't do is look ahead six months, 12 months, 18 months down the road and reasonably argue that the economy will peak, or not. There are simply too many unknowns when you're gazing that far into the future.

In fact, it's really, really hard to call business cycle peaks with a high level of confidence before there's confirming data to back up your claim. Many have tried, and failed. The record on this front is quite clear, as many studies remind. The next best thing is looking for high confidence signals as early as possible that the economy has recently peaked. This type of analysis can be supplemented to a degree with econometric applications that assess the near-term future.

The dirty little secret in business cycle analysis is that the risk of missing the start of a recession is no less threatening than crying wolf far in advance. Both of these extremes tend to come with a high price in opportunity costs. The good news is that a compromise between these two extremes is practical and, in many respects, the only game in town. It takes quite a lot of work, however, and the analysis doesn't lend itself to dramatic headlines. But if you're looking for genuine perspective on how the big picture's unfolding, that's a small price to pay.