Predicting the unpredictable
Recession forecasting, long thought to be one the “Holy Grails” of Economics, is well known to be incredibly difficult, some say impossible. Intuitively, it’s not hard to see why. Firstly, recession triggers vary widely in nature, and for the most part appear to be random. Second, unlike the physical sciences, initial forecasts can be nullified by an “observer effect”. Suppose that some types of recession triggers are indeed forecastable. If policymakers foresee a looming recession and adjust their policies accordingly to offset it, then the only recessions that actually take place will, by definition, be the ones that are unforecastable.
That being said, these difficulties haven’t prevented the profession from trying. While we may never reach a satisfactory level of forecasting accuracy, the ability to determine the likelihood of a recession occurring in, say, the next 6 months, is valuable in itself. To that end, these efforts have borne some fruit. Over the years, some indicators have been found to provide a decent (though not infallible) warning of future slowdowns. Of these indicators, the most widely known and studied is the yield curve.
The yield curve is simply a plot of bond yields across different maturities. It is usually upward sloping, because investors generally demand a higher yield to compensate for the extra risk from holding longer maturity bonds. This extra risk is often referred to as the “term premium”. Occasionally though, the US Treasury yield curve has inverted, meaning that short-term Treasuries yield more than their long-term counterparts. Historically, an inversion in the US Treasury curve has preceded every US recession since the 1950s, save for one false signal in 1967 when an inversion was followed by an economic slowdown, but not a recession. Indeed, our statistical tests show that when 10 year Treasury yields are higher than 1 year yields, there is a spike in the probability of a recession occurring over the next year.
Accuracy in simplicity
The yield curve’s accuracy in predicting recessions is surprising, given its simplicity. The complex relationship between yields and the economy makes it difficult to pinpoint the exact mechanism underlying the link between yield-curve inversions and economic slowdowns. But the prevailing theory is that yield-curve inversions are caused by the market pricing in a recession. Long-term yields reflect market expectations of future short-term rates set by the central bank. Therefore, an inversion usually means that markets are expecting the central bank to lower interest rates in the future (usually in response to weakening economic activity).
However, just because the yield curve has been a good predictor of recessions, doesn't mean that it will continue to be so. Economics and finance are littered with indicators that appear to work well… until they don’t. With data so easily available within these fields, it becomes all too easy for researchers to torture vast troves of data in search of apparently predictive relationships that actually turn out to be illusory, a vice known as “data mining”.
The fact that there is an intuitive explanation behind the yield curve’s predictive power, as well as evidence to support this intuition, suggests that it isn’t just a result of data mining, but we can never be 100% certain. At the very least, the fact that the relationship hasn't been as strong for other developed economies should tell us that this relationship isn’t a universal rule – some degree of subjective interpretation is required.
There’s also the question of why the yield curve should dominate other market based indicators (eg: the stock market, forecast surveys, etc…) in terms of its predictive power. After all, why should fixed income investors be any better than equity investors or professional economists in forecasting future economic activity?
Besides that, the current bond market environment may also reduce the yield curve’s predictive power. The term premium used to be much higher in the past. Therefore, an inversion meant that historically, the markets are predicting a large number of rate cuts in response to a very sharp downturn, since you need a very sharp fall in rate expectations to cancel out the term premium. With the term premium much lower today, a modest inversion wouldn’t have signaled a dire outlook as it used to in the past, thus making it a less valuable signal.
Despite its limitations, we still think that the yield curve provides useful information about future economic activity. But we also use it alongside a host of other indicators to form our outlook on the US business cycle. All indicators have their own limitations, and by looking at them together, we intend for their individual errors to cancel out, thus giving us a more accurate view of the economy.
So far, those indicators continue to signal low recession risk. There are still scant signs of overheating within the economy, the private sector is still running a financial surplus, and the ISM manufacturing index is still looking healthy. Commentators have recently highlighted the flattening of the yield curve, but evidence shows that it is the inversion itself that acts as the signal, not the degree of flattening. Given that the yield curve hasn't inverted, it is still signaling low odds of a recession over the next year. Combined with our base outlook that the ongoing trade spat doesn't escalate into a full blown trade war, all this suggests to us that the US economic backdrop remains broadly supportive of risk assets, and remains a key factor in our decision to maintain a moderate risk-on stance within our portfolios.
These are our current opinions but the future, as ever, is uncertain and past performance of investments is not a reliable indicator of future performance. The value of investments can fall as well as rise and you could get back less than you invest. If you’re not sure about investing, seek independent advice.