Trend following indicators on economic time series are useful risk management tools minimizing maximum drawdowns in your portfolio. Based on the simple model described thoroughly in this article, the worst drawdown experienced for the 1999-2017 period is -17.0% vs -55.2% for the S&P 500 TR.
Market timing is an intriguing yet controversial topic that’s been attempted many times with mixed results. Per its textbook definition, the strategy consists of buying and selling financial assets by trying to anticipate the direction and / or the amplitude of price movements using various techniques.
Out of all methods available to time markets, trend following is a major one used by many market participants over the last decades. The most popular indicator is the so-called “death cross” which represents the price crossover of the 50-Day moving average and the 200-Day moving average.
Two problems arise with trend following indicators: 1) decisions are taken neither with economic nor fundamental considerations, 2) they are prone to whipsaws when financial assets are range-bound. A solution to (at least) the first stated issue is to use them in conjunction with economic indicators.
First, we must select a relevant economic indicator reflective of the business cycle. Second, we need a timely indicator so we must discard all those updated on a quarterly / infrequent basis. In the end, very few meet these two requirements. One manages to capture our attention: the initial jobless claims.
Initial jobless claims data is released every Thursday at 8:30 ET. Its weekly frequency makes it an indicator of choice to assess economic conditions. However, the time series is known to be volatile from one week to another. To remove noise, we use a 4-week moving average on the raw data.
As seen on the chart above, the data is still rough despite our 4-Week smoothing moving average. We smooth it even more but this time with exponential moving averages for responsiveness. After some iterations, we find the 39-Week EMA and the 52-Week EMA to be quite optimal when combined.
In this research, we invest either 100% US equities in the S&P 500 TR when initial claims are falling or 100% US treasuries in ITTROV TR when initial claims are rising. These indices can be replaced with their investable ETF equivalent, SPY as a proxy for S&P 500 TR and SHY as a proxy for ITTROV TR.
The rules are extremely simple: you switch from SPY to SHY when both:
- The current 4-Week moving average > the previous 4-Week moving average.
- The 39-Week (3 quarters) EMA > the 52-Week (4 quarters) EMA.
You switch back from SHY to SPY when any of these conditions is not fulfilled.
This simple model is backtested on a weekly basis and achieves the results shown above. Without any leverage, the Model Portfolio generates a 13.7% compounded annual rate of return, 800bps above the S&P 500 TR. The maximum drawdown is -17.0% vs -55.2% for only 50 trades total since 1999.
The goal of this market timing model is not to beat the market when market conditions are favourable but to adopt a defensive stance when economic conditions deteriorate. There are many ways to become defensive: 1) lift cash, 2) change your asset allocation or 3) hedge with the use of derivatives.
In conclusion, the use of trend following indicators on initial jobless claims improves risk management of your portfolio. As always, quantitative models are made ex-ante but that’s normal. Financial theories are created from empirical evidence found in the past. Only time will tell if it works ex-post.
