Building a Meta Strategy – The Foundation

After having looked at the objectives for this strategy in the previous post, let´s take a step back to lay out some foundations before diving into the details. 

Are technical or fundamental indicators actually of practical use to tactically adapt a portfolio to market conditions?

Useful to us as an investor would be to create a strategy that is able to do either one of two things after costs and taxes: 

  1. it produces higher returns and / or
  2. it takes less risk with lower volatility and drawdowns than a buy-and-hold investment in the S&P 500 

When we combine these two factors, we are looking at enhancing risk-adjusted return. Even if a strategy doesn’t manage to increase returns, decreasing risk would enable us to sleep better at night (and avoid behavioral mistakes) or to use leverage to increase returns for the same level of risk as a buy-and-hold index investment.

This is exactly what a very simple technical or fundamental strategy has been able to do historically:

  • Holding the S&P 500 only when it trades above its long term moving average (e.g. the 200 day or 10 month simple moving average) has led to a 50% decrease in volatility by avoiding the biggest market drawdowns without considerably diminishing returns.

This chart shows the strategy backtest of a generic moving average crossover strategy as an example for a technical long-term trend strategy, all of which show very similar results. 
The strategy has returned 6,92% annually vs the S&P´s 7,7% since 1950 before dividends.

This Leading Economic Indicators strategy (long S&P when LEI 6-month rate of change is positive, otherwise in cash) backtest was run on monthly closing prices – intra-month drawdowns were higher at times.
The strategy has returned 6,37% annually vs the S&P´s 6,61% since 1959 before dividends.

For comparison the massive drawdowns of the S&P as a buy-and-hold investment:

It is easy to find many different examples, supporting this simple base case, used by practitioners, described in white papers and in academic studies. 

Running backtests across different basic methods (e.g. moving average cross-overs, absolute momentum or different leading economic indicators like the unemployment rate or the treasury yield curve), I found uncannily similar results over the long-term (across 35 to 68 years depending on the availability of the data):

  • Both fundamental and technical long-term timing returns are usually in the range of buy-and-hold returns (+/- 0,75% p.a.), mostly a little lower.
  • Volatility is cut by 1/3rd to half, with drawdowns down about as much. Maximum drawdown is lowered by 1/3rd, but the 1987 maximum usually stands out as a unique event that is missed by virtually all longer-term measures – all other drawdowns top out at half the maximum losses buy-and-hold had to endure.

These general results are the basis for our model. All methods differ in their year-to-year performance, but the essential advantage is the same and I’m quite confident it will prove robust in the future. 

I always make sure to include outlier events, like 1987, in backtests as a worst case scenario. It is easy to find parameters that exclude the most negative events, but doing so would overfit the data in my opinion and lead to overly optimistic results. Who knows the exact nature of the extreme events the future holds in store for us?


Why do we find these results?

The U.S. stock market and the U.S. economy move in the same general direction, because the economy drives corporate earnings, which drives stock prices over the long run. Over shorter time periods stock prices strongly oscillate around this general trend, sometimes running ahead of economic growth, sometimes falling back despite of the economy growing – this is mainly caused by changing expectations. 

But over the long term the really severe bear market declines (when stocks fall more than 40%) always happen in the context of economic recessions, when growth actually turns negative. Usually a higher degree of overvaluation in equity prices going into a recession leads to deeper declines.

This chart shows the S&P 500 with all declines around 30% or greater marked in yellow and recession periods in grey. Several smaller recessions at lower equity valuations led to a stock market decline of less than 30%, while only one 30%+ decline happened outside of a recession: the crash of 1987, an outlier event that was foreshadowed by preciously little solid data. (The 1962 decline stopped at -28%).

As GDP numbers themselves are lagging they will show a recession too late (after stock prices have already gone down a lot), but several leading economic indicators often signal declining economic growth around the same time stock prices top. 

Economic developments are constantly incorporated into stock prices and these, by definition, will break below their long term trend and trigger technical signals in the beginning of strong bear markets. 

Overall the cycles of the economy tend to be long enough, that it makes sense to adjust a portfolio to economic conditions or to the price trend reflecting these conditions – doing so has resulted in reduced portfolio volatility historically.

Are recessions actually a big issue?

A majority of market participants and economist spends their time trying to predict recessions. Their rate of success is not very high – usually we are faced with a stream of news vastly over-predicting the likelihood of catastrophic events. It is difficult to do and the basic probabilities are against an economic downturn happening at any given time.
Since 1929 the U.S. went through 14 recession – only once every 6,4 years on average. Each contraction was also much shorter than the usual expansion – 9 months on average – which puts the recession base rate at only 11,66%.  

Our default position should therefore be long equities or risky strategies unless the model signals otherwise.

When looking at the inflation adjusted long-term development of U.S. equities, there were really only three very long periods with devastatingly negative returns (severe drawdowns lasting for years before new highs were reached): 

  • 1929 – 1942, with new highs reached in 1956
  • 1968 – 1982, with new highs reached in 1992
  • 2000 – 2009, with new highs reached in 2015

Inflation adjusted long-term chart by macrotrends

On a more granular level, market drops larger than 30% from fresh highs happened 6 times during this period – smaller recessions with moderate equity valuations quite often lead to smaller declines around 30%.

It is good practice to visualize this very long-term market trajectory, if only to realize that after major market bottoms all upheavals were relatively temporary in nature for a period of 20 to 30 years. This puts the doomsayers, who are always predicting the worst market drop ever about to rear its ugly head, into perspective – after all we are less than 10 years away from one of the major market bottoms of the last 100 years.

On the other hand, because these major cycles last so long, it is very difficult in practice for an investors to stick to a buy-and-hold approach – a long-term down cycle at the wrong time could prove to be ruinous. Even phases like 1998 or 2015-2016, not to mention 1987, are very hard to tolerate as each might just be the start of a long, devastating period for us personally – we can never know for certain. 

As evidence points to the robustness and reliability of methods, that enable us to move to the sideline during many large corrections and bear markets without hurting our bottom line, it would be vastly preferable to use them.

5 comments

Submit a comment

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s