Dynamic Asset Allocation

|
Includes: FKRCX, VFINX, VWESX
by: Abhishek Kansal

Summary

Identifying the right asset classes and proportions to diversify is difficult for an investor.

The scientific methods for diversification, namely Markowitz’s Mean Variance Optimization have not been practically applicable.

Investing in all asset classes evenly at all times will reduce risk but lower returns too.

A diversification strategy that reduces exposure to asset classes trending down long term has historically outperformed the stock market both in terms of overall return and volatility.

Diversification is widely accepted as the most important aspect in building a portfolio. For investors looking to accomplish their long term financial goals, diversification helps reduce risks and volatility as market and economy go through various expansion, contraction cycles. However the specifications on how much to diversify and in what asset classes are often vague and left to the judgment of an individual investor. There aren't many established or prevalent public tools that would take investor characteristics as an input (for example risk tolerance, time horizon etc.) and output a recommended model portfolio. A recommended portfolio that provides a list of specific asset classes (mutual funds, ETFs or stocks) and propose percentage weights for investor to review and consider as a starting point. Further, the primary goal for diversification is looked at as risk minimization or reduced volatility in your portfolio. That comes at a cost since lower risk leads to lower return. Could diversification lead to lower risk and yet outperform the market in terms of returns? This article proposes a diversification strategy that has historically outperformed the market, with lower drawdowns and can be used by investors to build a long term asset allocation strategy.

Background: Let's start with understanding the history and state of financial theory on diversification. Harry Markowitz's Mean Variance Optimization (MVO) method developed in 1952 forms the core backbone of financial theory on diversification. The core insight of Markowitz's work was that by combining assets that are negatively correlated (i.e. they typically move in different directions) one can reduce the overall volatility of a portfolio without impacting the expected return. Markowitz provided a mathematical algorithm that can use this insight to generate the ideal portfolio (named as Markowitz Efficient Portfolio) with lowest risk/volatility possible. This was a powerful algorithm and Markowitz rightfully won a Nobel Prize in 1990 for it. Unfortunately even though this was a powerful algorithm, it has not turned out to be practically applicable (Reference papers: 1, 2, 3). It entails complicated mathematics sensitive to minor changes in the input and requires accurate future forecast on potential assets. Historical returns are very poor forecasts. Variations of Markowitz's algorithm like Black Litterman model have been proposed to overcome these limitations, however even these require sophisticated inputs (like asset market weightings, volatilities and correlations) that may not be easy to provide for by an average investor.

Diversification Strategy Options: To build a model that is simple to understand, compute and specific in terms of output recommendations, we start with Markowitz's key insight: incorporate assets that are negatively correlated in a portfolio. However correlation between two assets can change over time and rather quickly so we don't want to assume future correlation will be same as past. Instead we incorporate asset classes that have the potential to have negative future correlation. Thus we include assets in the portfolio that are fundamentally or significantly different from each other. To illustrate this with an example, let's start with Stocks, Gold and Bonds as three available asset classes that are fairly different from each other. Let's pick a mutual fund or index from each of these to start with diversification in asset class itself and not be exposed to individual stock risk. I picked the Vanguard 500 Index Fund (MUTF:VFINX), the Vanguard Long Term Investment Grade Fund (MUTF:VWESX) and the Franklin Gold and Precious Metals Fund (MUTF:FKRCX) to represent stocks, bond and gold in this test portfolio. We could have picked ETFs like the SPDR S&P 500 Trust ETF (NYSEARCA:SPY), the SPDR Gold Trust ETF (NYSEARCA:GLD) and the iShares 20+ Year Treasury Bond ETF (NYSEARCA:TLT) but those have historical data only since 2002. Using VFINX, VWESX and FKRCX as proxies for stock, bond and gold allowed me to back test on historical data going all the way back to 1985 from Yahoo Finance.

The simplest diversification without making any future assumptions on expected returns would be to allocate equal one third percentage to each asset class. How would this constant equally diversified portfolio would have worked as compared to staying 100% invested in stocks? Overall, stocks would have generated better returns but they'd have also seen larger volatility as seen in the higher drawdown in table below. The graph below shows how the two portfolios would have grown and the table shows annualized return and drawdown numbers for the duration.

Looking at the above numbers, a simple strategy of equal breakdown across multiple asset classes provided a good start for reasonable growth and yet lower drawdowns. However, could we have generated better returns than being in stocks alone? We can take advantage of being in an asset class rather than an individual stock. Individual stocks can go through wild up and down swings, but asset classes do show longer bull - bear trend. For example, the graph below shows that "Gold - Precious metal equities" have been a 4 year long bear market since 2011. Similarly U.S. stocks went through 2-3 year bear market in 2000 and 2008.

One improvement that we can make in our diversification strategy is to exclude any asset class that is in its longer term bear market and equally invest in all other asset classes. An asset class can be marked in bear market if its 52 week return is less than -2%. We could use any other indicator too like simple moving average or 52 week minima drop. They will all work. The important thing is to classify it as a bear and exit or reduce your sizing in that asset class. Any heuristic that improves the accuracy of classifying an asset class is in bear market will improve the strategy further.

In our proposed dynamic allocation strategy we simply reduce allocation to zero on an asset class which has lost more than 2% over the last one year. All other assets are held in equal proportions to make up 100% of portfolio and balanced weekly. For simplicity we have assumed balancing weekly has zero costs, in reality transaction costs may necessitate balancing over a longer time period like 1 or 3 months. Back testing this strategy on historical data since 1984 returns an annual return of 11.87% with an average drawdown of 3.73%. The worst case drop from a 52 week high was 31.35%. So an outperformance both in terms of returns as well as lower volatility.

Conclusion: Investors who manage their portfolio on their own, can use the learning above to build their own long term portfolio management strategy. They can extend the above proposed strategy to cover a comprehensive set of asset classes to include all major sectors like real estate, commodities etc. as well as international economies. Including more asset classes should help reduce risk but too many asset classes will decrease the overall return. Investors can try a variations where instead of equal allocation across all asset classes, sectors that are booming have higher weighted allocation versus sectors that are underperforming. Catching a long term bull market in an asset class and over indexing on those asset classes is likely to help improve returns. They can adjust the maximum level of weighting in a single asset class based on their risk tolerance to limit over exposure in a single asset class. Investor can thus build their own diversified portfolio, test its historical performance on returns and drawdown and thus be equipped to make smarter investing decisions for the long term.

Disclaimer: The author does not have any holdings in the mutual funds (VFINX, VWESX and FKRCX) used to test described diversification strategy. These funds have been used only for illustrative purpose and the author is not making any recommendations to buy them. Past results are not necessarily indicative of future performance. Results discussed above are based on hypothetical trades. Actual results may vary due to certain market factors like lack of liquidity.

Disclosure: I am/we are long SPY.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.