As of last week, the combination of evidence we observe continues to be associated with strong recession risk and the likelihood of a "whipsaw trap" in the stock market. We'll respond to new data as it changes, but I expect that the primary window of interest here is about 6-8 weeks. In the event that economic data can produce fairly upbeat readings over that horizon, and the S&P 500 can remain at or about present levels, our estimate of oncoming recession risk would back off fairly quickly. Presently, that outcome would be outside of the norm based on the leading economic measures we track, as well as the overvalued, overbought, overbullish condition of the stock market.
I want to emphasize again that I am neither a cheerleader for recession, nor a table-pounder for recession. It's just that given the data that we presently observe, an oncoming recession remains the most probable outcome. When unseen states of the world have to be inferred from imperfect and noisy observable data, there are a few choices when the evidence isn't 100%. You can either choose a side and pound the table, or you can become comfortable dwelling in uncertainty, and take a position in proportion to the evidence, and the extent to which each possible outcome would affect you.
With most analysts dismissing the likelihood of recession, I have been vocal about ongoing recession concerns not because I want to align myself with one side, but because the investment implications are very asymmetric. A slow but steady stream of modestly good economic news is largely priced in by investors, but a recession and the accompanying earnings disappointments would destroy some critical pillars of hope that investors are relying on to support already rich valuations. We're always open to shifting our investment stance and outlook in response to new evidence, but the "optimistic" evidence that many observers are using to discard recession concerns is generally based on coincident or lagging data.
In the past couple of weeks, we've seen few articles presenting various opposing views to our (and ECRI's) recession concerns. A few of these are about as valuable as what you might fish out of a Cracker Jack box, focusing on the Conference Board LEI (despite its heavy reliance on real M2, which even the Conference Board has decided to discard beginning with the next report), or framing the ECRI's view as if it is driven only by the weekly leading index. Another article presents an index that does in fact turn down during recessions, but with varying lead and lag times, and inconsistent thresholds, so there's no way to define a useful signal except with hindsight.
While I am convinced that the data weigh heavily toward recession risk, there is a well-done and more optimistic piece by Dwaine van Vuuren out of South Africa that is worth reading, particularly for methodological reasons. I appreciate van Vuuren's approach because the investment world desperately needs more analysts who thoughtfully examine long-term data rather than using toy models or tossing opinions off the top of their heads. He constructs a super-composite of nine economic indices, noting that one can only "safely proclaim" recession when the majority of those has turned negative. I agree with that observation, but that's mainly because the particular measures included in the composite (such as the ISM, Chicago Fed and Philly Fed indices) are largely coincident, very short-leading, or short-lagging. Even if one alters the weights on these, it is very difficult to create leading indicators out of coincident ones. My impression is that the implications of that super-composite are likely to shift in fairly short order, possibly as soon as the Conference Board introduces its new version of the LEI, so it will be interesting to see how these measures evolve in the next couple of months.
On a statistical note, the fact that only a few of the indices in the super-composite are currently in recession territory doesn't translate well into the conclusion that recession risk is low*, because of that distinction between leading and coincident indicators. For investors, this is a particularly important issue. The stock market itself is a short-leading indicator of recession, so from an investor's standpoint, coincident or short-lagging recession indicators are not as useful as one would wish. By the time it's safe to proclaim a recession and close the barn door, the horses are already out. This is why investors have to be very sensitive to early measures of recession risk.
[*Geek's note: There's limited information from running a logit model using "recession 4 months from now" as the dependent variable, if the independent variables generally don't lead at that horizon. Also, there's risk in fitting a multivariate logit or probit model with ordinary variables rather than binary flags, because your output will generally be very sensitive to the exact covariance structure of the data, so a model that fits almost perfectly in training data will typically deteriorate rather quickly in out-of-sample data. The other problem with continuous values in probability models is that when you use them in univariate logit/probit estimations, the model amounts to a threshold filter that is highly nonlinear around a single value, producing probability estimates that cluster either at zero or 100%, with abrupt leaps in-between. The practical difficulty is that this can send your signal from about zero chance of recession to near-certainty of recession with the addition of just one or two slightly weaker data points.]
Capturing a Syndrome
Recession evidence is best measured by capturing a syndrome of conditions that reflects broad deterioration in both real activity and financial indicators. What's perplexing to me is that the recession concerns we're seeing are evident even in composites of very widely tracked economically-sensitive indicators. For example, the chart below is simply the average of standardized values (mean zero, unit variance) of the following variables: 6 month change in S&P 500, 6 month change in nonfarm payrolls, 12 month change in nonfarm payrolls, 6 month change in average weekly hours worked, ISM Purchasing Managers Index, ISM New Orders Index, OECD Leading Indicator - total world, OECD Leading Indicator - US, ECRI Weekly Leading Index growth, Chicago Fed National Activity Index - 3 month average, credit spreads (Baa vs 10-year Treasury), Industrial commodity prices - 12 month and 6 month change, and New building permits 6 month change.
