As the financial markets enter what I expect to be a rather disruptive completion to the recent speculative half-cycle, it will be helpful for investors to consider certain propositions that are readily available from history, rather than insisting on re-learning them the hard way.

Employment data is the last to know

Proposition: Risk-sensitive assets and confidence measures generally precede economic shifts; production, consumption and income measures are coincident with broad economic activity; and labor market measures lag the economy. The unemployment rate is the single most lagging economic indicator available.

I’ll begin by noting that while some of our measures are already warning of recession risk, I currently characterize those risks as increasing but not decisive. The more sensitive the measure, the more vulnerable it is to false signals. Given that market conditions already place us in a defensive investment position, I’d prefer to sacrifice a bit of timeliness on a recession signal in return for a bit more confidence. There’s little to be gained from rushing to a recession warning on the basis of less reliable measures. The conservative version of our Recession Warning Composite wasn’t quite triggered in September data, but suffice it to say we’re quite close.

Notice that our Recession Warning Composite isn’t based on any single indicator, but on joint deterioration in financial measures (stocks, credit, yield spreads), confidence (ISM), and either a slowing of employment growth or a modest 0.4% increase in the unemployment rate. In previous cycles, I used these employment measures as confirming indicators, but I added them to increase our confidence in initial recession warning signals, despite the fact that they can add a slight delay.

As I noted in my October 2000 recession warning, also based on this composite:

It is quite true that consensus economic forecasts remain relatively upbeat here. Unfortunately, most economists have never fully internalized the ‘rational expectations’ view that market prices convey information. Of course, accepting this view does not require one to believe that prices convey information perfectly (which is what the ‘efficient markets hypothesis’ assumes). But where finance economists take this information concept too far, economic forecasters don’t take it far enough. As a result, economic forecasts are generally based on coincident indicators such as GDP growth and industrial production, or pathetically lagging indicators such as consumer confidence and the unemployment rate. Even the ‘leading indicators’ fail to live up to their name. This tendency to gauge economic prospects by looking backward is why economists failed to foresee the Great Depression and every recession since.
– John P. Hussman, Ph.D., October 2000

The reason our Recession Warning Composite focuses on psychologically-driven variables like stock prices, yield spreads, credit risk, and purchasing manager surveys is that few measures of “hard” economic activity reliably lead recessions. For example, while recessions tend to be preceded by sub-par employment growth over the preceding 6-12 month period, job growth tends to turn negative only after entering a recession, not before. Similarly, the rate of unemployment tends to tick up slightly before a recession starts, but the strongest increases in the unemployment rate typically lag the start of a recession by about 8 months. New claims for unemployment also have only slight short-leading usefulness.

When we examine broader economic measures that are well-correlated with overall economic activity, we observe clear deterioration, but these measures are not yet at levels that one would decisively associate with recessions. The chart below shows the average standardized value of 25 economic measures we track, including employment growth, income, production, capacity use, and similar variables. Broadly speaking, this particular set is more useful in gauging the magnitude of economic fluctuations than identifying turning points.