# Adaptive Asset Allocation

### Adaptive Asset Allocation:

Dynamic Global Portfolios to Profit in Good Times – and Bad

#### By Adam Butler, Michael Philbrick and Rodrigo Gordilla

Book Review by Jacob Rothman

Investment Philosophy for managing money for others:

Better to “lose half your clients during a raging bull market than to lose half your clients’ money during a vicious bear market.”

The authors challenge the standard way of looking at risk – volatility, and stress that risk the probability of missing one’s financial goals, and that investing should be done to minimize this risk. Since most investors are unable to stick with a strategy for 35 years, even if that is their true time horizon, a portfolio should be built to allow the investor to stay the course. “Applying intelligent strategies that meet both financial and emotional needs is how the best advisers execute on the most important concepts in wealth management.” (p. 9)

A lot of portfolio analysis provides a lot of useless precision, but misses the whole point of whether a portfolio is 1) something an investor can stick with and 2) will enable the investor to reach the final goal. P. 16

While many advisors and investors focus on risk and return on an asset by asset or fund by fund basis, the correct way is to examine how each affects the whole portfolio. The authors compare this to creating a meal – having good ingredients is important, but the interplay among the ingredients and getting them in the right proportion is very valuable. (p.21) They assert that a great recipe with average ingredients beats a weak recipe with great ingredients based on a minimum variance portfolio vs. a revenue or dividend weighted portfolio of value stocks. This is portfolio optimization – using correlation along with volatility and expected return in portfolio construction. “The most important lesson from the Cass study… is that it is significantly more important to understand how the different parts of the portfolio work together than it is to understand the individual parts themselves. (p. 28)

Portfolio volatility matters because it reduces the compounding effect for savers and creates shortfalls for retirees (the opposite of dollar cost averaging). (p. 33-34) The portfolio returns over the first few years of retirement are very important – if the portfolio loses heavily, withdrawals will come at lower levels and the portfolio will not recover. Historically, there has been a wide range of returns over a 20 year period. Annualized stock market returns have ranged from 2% to 18%. (p. 39). Financial plans based on average returns have roughly a 50% failure rate. (p.42) As a case in point, the authors looked at DJIA from 1966 to 1996 and assumed a saver made regular monthly contributions which totaled $868k over that period. The portfolio would have been worth $3mm at the end of 30 years. The compounded annual growth rate of the Dow was 8% over the full time period. As a contra-example, the annual returns were reversed to show what would have happened if the bull market from ’81 to ’96 would have occurred first. The DJIA return would be the same, but the investor would have only accumulated $1.1mm, as the strong market returns would have happened when accumulated savings were small, followed by a flat market when the balance was higher. We are subject to luck, as we can’t control when we are born. (pp. 53-56) Similarly, a person retiring in 1966 with $3mm and withdrawing 8% of the original portfolio value annually would have run out of money by 1978, but if the time sequence were reversed, he would have had almost $6mm leftover. (pp. 59-61) Sequence of returns are very important and outside of our control. Even if we know we can earn 8% annualized over a long period of time, our actual results can vary greatly based on the return pattern. Michael Kitces found that the safe withdrawal rate (in retrospect) for a thirty year time horizon is 91% correlated to the returns over the first fifteen years. (p. 73)

After establishing the importance of sequence of returns and demonstrate the wide divergence of experiences based on when people are saving and when they retire, the book looks at how to create realistic expectations for the future. Since a return is calculated as the future value over the purchase price, the higher price one pays now, the lower the expected return will be. The book uses four different valuations to get a thorough look at current valuations. 1. Price to earnings (using a longer than one year earnings average) focuses on the earnings statement, 2. Q ratio (the ratio of market valuation to replacement cost) focuses on the balance sheet, 3. Market cap to GNP focuses on corporate value relative to the size of the economy, and 4. Deviation from long-term price trend focuses on the price series. All four metrics have solid long-term track records of predicting forward returns and all four use a different approach from the others. (p. 69) Using these metrics can both help calibrate expectations as well as inform how a portfolio should be constructed. It makes sense to take more market risk when expected returns are higher and risk is lower. This is not how most advisors construct portfolios, however. The authors note that “a ‘traditional’ advisor” would have recommended the same allocation at the peak of the technology bubble as at the depth of the Great Recession. (p. 70) They also cite Jeremy Grantham in noting that profit margins mean-revert in any properly functioning capitalist economy. Currently, profit margins are very high. If both profit margins and earnings-based valuations revert to normal levels, investors will take a double hit. (p. 71) (Note that only the first of the four valuation metrics suggested use profits, so the other three are not skewed by anomalies in margins.) As of October 2014, the model used by the authors predicted real returns for the S&P 500 of –1.35% for fifteen years, and 0% for twenty years. (p.74) (Ostensibly, since the market return has been strong in the five year since this prediction, the current outlook is even worse.)

The authors explain their methodology for predicting long-term returns and show that their model has historically explained about 80% of 15-20 year returns, which is 80% more accurate than always assuming long-term averages. Useful data sources for tracking valuation metrics and learning about expected returns:

Gmo.com

Aqr.com

Hussmanfunds.com

Alphaarchitect.com

Pimco.com

Researchaffiliates.com

Strategic asset allocation assumes long-term averages for expected volatility, expected returns and expected correlations, yet large deviations from the averages occur regularly. (p. 95)

Diversification works – an all stock portfolio has returned much more volatility and greater drawdowns than a balanced portfolio. The authors look at a simplified permanent portfolio with equal weights to stocks, bonds, gold and cash, from 1995 to 2014. This returned only slightly less than a 60/40 stocks/bonds portfolio (9.1% vs. 9.5%) with much lower volatility and drawdown and slightly higher Sharpe ratio. In deflationary Japan from 1991-2014, the simplified portfolio nearly doubled the value of simply buying and holding the Nikkei, with a max DD of almost 1/3 of the Nikkei. (pp 115-118)

The authors look at risk parity, but rather than using long-term volatility averages, they use trailing 60 day volatility as a predictor of short-term volatility. Using a strategy of scaling exposure to the S&P 500 based on trailing 60 day vol and setting the portfolio to 15% volatility. Over the 1990-2014 time period, this nearly doubled the return vs. buy and hold, while controlling risk. This does require leverage in low-volatility markets. (p. 123-124)

Finally, the authors get to the meat of their system – momentum. “Smart investors recognize that one of the most powerful forces in human decision making is social influence, whereby a person’s decisions are influenced by the actions of those around them, and vice versa. Research shows that, when faced with a choice in the absence of trusted information, humans will usually choose to follow the crowd rather than act against it. Momentum is simply the manifestation of this phenomenon in the market.” (p. 130) Using a ten-asset universe that includes global stocks, US treasuries, real estate and commodities (I don’t know why they exclude corporates) they test portfolios over a twenty-year period. Including only the two assets with the highest momentum returned 16.3% CAGR vs 10.4% for the S&P, with the max drawdown of 14.6% vs. 55.5%. The top five also significantly outperformed the S&P 500. (p. 131)

Putting it all together, the book lays out how all of this might work in constructing a portfolio. The S&P 500 is used as a benchmark. (10.3% CAGR, 15.5% vol, 77% positive rolling 1 yr, 0.67 Sharpe, -50.8% Max DD). Naïve risk parity using their ten assets (15% target vol, no leverage) reduces return, but reduces risk even more (8.5% CAGR, 8.6% vol, 89% positive rolling 1 yr, .99 Sharpe, -24.2% max DD). This alone makes a portfolio that investors can stick with, and equaled a 60/40 return with less risk. A minimum variance version gives up a little return, but reduces the max DD to 6.6%. Next, the book starts with a momentum portfolio using the top five of the ten asset classes. (13% CAGR, 11% vol, 93% rolling positive 1 yr, 1.17 Sharpe, -21.7% max DD). (This is similar to the Mama Bear portfolio from Muscular Portfolios, but uses five rather than 3 assets.) This portfolio is then optimized, using naïve risk parity as a first step, and then implementing minimum variance. This is the portfolio that incorporates momentum, volatility and correlation. The results are excellent (15% CAGR, 9.4% vol, 99% positive rolling 1 yr, 1.6 Sharpe and -8.8% Max DD). (pp. 133-141)

The book ends with a more technical defense of the research (not overly fitted, large sample size, theoretically sound, multiple discovery, structural impediments to implementation) and the reprinted of a couple of white papers formerly published by the authors, which argue for Adaptive Asset Allocation as a neglected area that both provides significantly more value than security selection as well as a better information ratio due to more decisions being made over time (thus reducing the probability that apparent skill is merely luck.)

Conclusion: stock picking is sexy and a 60/40 equity/bond portfolio allocation is very common, but there both are subject to long and intense periods of underperformance which are intolerable for most investors. High correlation within asset classes means simply choosing the right asset classes can be very valuable. Momentum works and can be applied to asset classes. Recent Volatility is also a predictor of short-term risk and reward. Correlation is important to proper diversification. By combining these concepts into one portfolio, an investment manager can likely increase return over time while decreases volatility – both by significant margins.