The essential idea behind evaluating current and projected economic conditions for the meta strategy, is to map different allocation decisions to the signals of a core tactical asset allocation strategy. The weaker the economic development is judged to be, the stronger we react to technical market timing signals.
Are there fundamental indicators pinpointing probable recessions or the weakening of economic conditions?
To be useful we would need these indicators to be timely and exact, meaning they signal a recession ahead of time, without missing any and giving relatively few false positives.
Falsely predicted recessions or very early warnings are not actually such a big problem in practice, if we simply fall back on our technical model for the exact timing of allocation decisions. We aim to reduce the amount of whipsaws we go through as they act as a frustrating „death by a thousand cuts“, but each single false warning won’t do us much harm as long as we get back into the market fairly quickly.
So, if you hear the argument: „this indicator predicted 9 out of the last 5 recessions“, it does not mean it is useless – you can still use it as a filter to improve a technical model that warned of 15 out of the last 5 bear markets (and even such a low success rate can still protect you from the big declines). What we want to avoid is a model that predicted only 4 of the 5 last recessions and we need to add an emergency stop loss in case the model turns out to be wrong in the future.
This is the sequence of events we are looking for primarily:
- Leading economic indicators deteriorate and trigger recession warnings one by one around a major stock market top.
- Technical signals of different sensitivity are in place: the more fundamental warnings are triggered, the stronger we react to sensitive signals.
- The stock market tops at some point and technical signals will gradually get us out of the stock market after the top: more sensitive signals will get us out closer to the top.
- If a bear market or very strong correction start without fundamental early warnings the least sensitive technical signals serve as a stop loss.
Several economic measures fit the bill – you can easily access all of the data (and much more) through the St. Louis FED´s FRED website.
From several reliable resources verified by my own research and tests, I compiled a handful of indicators that add value and make the most common sense to me – but the list is by no means exhaustive. I want to avoid unnecessary complexity and conflicting signals, but not be completely binary (several models with historically successful backtests, that I found, rely on only one economic measure with one signal point).
I will go through them, including some very widely used measures, that do not seem to be particularly useful for our purpose.
Leading Economic Indicators
There are dozens of economic indicators which can be combined into a model and I have whittled them down to a handful, that I use. The key is, that they need to moderately lead the actual economic growth numbers to be timely enough to be useful – neither too early nor too late.
Not very useful indicators
Let’s first look at some widely used measures, that didn’t improve outcomes, when I tested concrete timing signals for the purposes of the meta strategy:
- Yield Curve – 10 year UST minus 2 year UST: the most widely quoted part of the US Treasury Yield Curve is usually very early and has inconsistent lead times in signaling a recession when it inverts. Being much too early equals being wrong in practice – but it can serve as an alert to play closer attention to more timely indications including other parts of the yield curve.
- Valuation: CAPE Ratio or Tobin´s Q are useful for very long term outlooks above 5 years, but I couldn’t find consistent target areas one can use to quantify turning points in the market – over-valued markets can always become even more expensive. If you have identified a turning point in the market, the valuation at the top can give a hint at the severity of the bear market we can expect: high valuations lead to larger market drops in recessions. By rotating into equities with lower valuations and avoiding the most expensive areas, the CAPE works well in weighting the equity allocation in a global buy and hold portfolio.
- Housing market: it is often used as one of the earliest economic warning signs, when it starts to decline. I found it too early to be useful and a very noisy indicator raising many false red flags – it is probably popular because it flagged the 2008 financial crisis (which was caused by real estate problems after all) so clearly.
The best leading indicators
Fundamental indicators look at different areas of the economy and are differentiated by sensitivity for our purpose: Less sensitive indicators flag recessions and ignore most other temporary market dislocations. As „minor, temporary“ drops include crashes as in 1987, and several 20% to 30% declines, that I prefer to include in my market timing model, I also use several sensitive fundamental measures.
Essentially I want to generate three different outcomes from my fundamental analysis to be able to adjust my portfolio exposure gradually: Green, Yellow (a correction warning is signaled by at least 2 indicators) and Red (a recession warning is signaled by at least 2 indicators + several correction indicators).
For my analysis I combined on a single chart: the FRED fundamental data (blue throughout the charts) with the S&P 500 (orange, log scale) and U.S. recessions (grey shaded areas). Because I want to see the interplay of fundamental signals (red arrows) with my technical strategy, I included entry (green vertical lines) and exit (red vertical lines) signals from a generic, long-term trend system – a 60-day, 275-day moving average crossover. In general the golden cross is used as the final re-entry point in the market, because it showed a timely entry with a high success rate throughout my backtests – I consider being in the market as the „normal“ state as we are in a bull market with a high probability at any given point in time due to the equity market´s upward bias. For the meta strategy the entry doesn’t carry nearly as much importance as the protective exit.
As a visual person, I find this chart based overview more nuanced than the return numbers of the backtests – it enables me to understand the way each indicator has worked in the past and how it fits together with the price development of the market.
My final goal is not only to allocate to the S&P 500 according to the model, but rather to choose the most suitable strategies in my arsenal to use at any given time and to determine how high my exposure to each strategy should be.
Recession Warning: Red
Composite Leading Economic Index – LEI: this monthly composite of leading indicators filtered out most false signals and flagged all recessions since 1960 close to major market tops while ignoring most larger corrections. Signals are generated when LEI´s 6-month rate of change turns negative and are marked as blue vertical lines on the chart (click to enlarge):
Today LEI gives a green light.
Yield Curve: 10 year UST minus 3 month US T-bill: this most timely part of the yield curve signals recessions around market tops, sometimes showing problems appearing 1 – 2 years before (1989, 1998 or 2006). Signals are generated when the yield curve inverts:
Today the 10 year UST minus 3 month US T-bill yield curve still gives a green light, but nears an inversion (as it did in 95 and 98 without long-term consequences).
Unemployment Rate: data available since 1950 gives more nuanced indications before all recessions with few false positives. Often signals cluster before severe bear markets, indicating gradually worsening conditions coincident with the formation of a market top – very useful in practice. Recent deep bear markets started from very low levels of unemployment starting to trend up significantly. Signals are generated when the unemployment rate starts to trend up year over year (the rate crosses above its 12 month average):
Today the unemployment rate has just turned red for the first time in several years.
Correction Warning: Yellow
Initial Claims: they are closely related to the unemployment rate, but show earlier, weekly tendencies for employment. The noisy data needs to be smoothed to usefully supplement the unemployment rate indication. It gives more false alarms, but also points to smaller significant market events in time (e.g. 1987, 1989, 2018). Signals are generated when initial claims start to trend up and the 12 week moving average crosses above the 52 week average:
Today initial claims are just starting an up-trend, flashing a yellow warning in tandem with the unemployment rate.
Real Retail Sales: long term data since 1950 shows that significant market declines, not only recessions, were usually preceded by retail sales stopping to rise (exceptions are 1987, 1998 & 2011). Signals are generated when retail sales flatten or top (6 month average is flat or lower over a period of 3 months or more):
Today retail sales give a green light – it is extremely rare for markets to go into a tailspin while retail sales are rising.
National Financial Conditions– leverage subindex: financial conditions tighten more and more before a market decline – the gradual increase above neutral (0) shows economic conditions getting tougher and tougher. This often happens early, but with few false signals, and many large corrections (e.g. 1987, 1998 or 2011) are indicated before they actually play out. I find this particularly useful to switch to more sensitive technical indicators to time an early exit from my most vulnerable strategies. Signals are generated when financial conditions cross into positive (= tight) territory:
Today financial conditions give a green light, but may have bottomed – 2018´s market drop coincided with a rise in financial tightness from very loose conditions.
The combination of fundamental indicators gives a yellow warning today – some economic deterioration is starting to appear.
For the meta strategy this implies, that today we should react more sensitively to our technical signals – yellow, red and stop loss signals, all of which were triggered in the market decline in December 2018. As we are still in a long term downtrend, but economic conditions are largely positive, we should be in safe assets right now, but not explicitly short.
Re-entry into risky strategies will begin when the S&P 500 reclaims the 275-day moving average, going to full exposure after we see a golden cross. I don’t use the VIX Term Structure to time my re-entry, because its changes tend to be very volatile below the long term trend – strong rallies often turn around very suddenly to be revealed as bear market rallies.