Tuesday, April 22, 2008

Academic Imprimatur for Our Investment Strategy

Our investment strategy is based on the idea that including non correlated assets in a portfolio can raise returns while reducing volatility. The difference between our portfolios and traditional portfolios is that we use assets whose returns have low or negative correlation to stocks. Traditional portfolios pay lip service to this concept by including bonds and token allocations to REITs and Commodities. Our portfolios include much larger allocations to REITs and Commodities due to their ability to deliver non correlated returns. Tim at The Mess That Greenspan Made has a post that highlights recent academic research which vindicates our view:

There's a new study out by Professor Craig Israelsen at BYU that provides more good news about commodities as an asset class.

As a number of other reports have also done in recent years, it dismisses the notion that commodities should be viewed as some sort of a "No Go Zone" for individual investors. Instead, in this report, commodities are shown to provide good diversification for typical investment portfolios while boosting overall returns.

Professor Israelson's study can be found here and an interview can be found here.

Here's the key part of the study that supports our strategy:

HardAssetsInvestor.com (HAI): What did your study on correlations show?

Craig Israelsen (Israelsen): Basically that diversification really works. That's a real stunner, isn't it?

HAI: Shocking. But seriously, how did it work?

Israelsen: I built equal-weighted portfolios out of up to seven different asset classes: large-cap U.S. equities, small-cap U.S. equities, non-U.S. equities, U.S. intermediate-term bonds, cash, REITs and commodities.

For commodities, I used the S&P GSCI commodities index going back to 1970. I'd note that before 2001 or 2000, the GSCI was not investable, so I'm making the assumption that there could have been an actual portfolio tracking that index back in 1970.

In my original analysis, I started with an equally weighted two-asset class portfolio composed of large-cap and small-cap U.S. stocks, and I looked at the returns. Then I started adding more asset classes: non-U.S. stocks, bonds, cash, REITs and commodities. I found that as you added the additional asset classes, you improved the returns and limited the worst one-year drawdown of the total portfolio. But importantly, it was not a linear relationship.

HAI: How so?

Israelsen: There's a major change when you get to commodities and REITs.

With the five-asset portfolio - large-cap U.S. stocks, small-cap U.S. stocks, non-U.S. stocks, bonds and cash - which is about what the typical target date portfolio held three years ago, you get a 10% internal rate of return while sustaining retirement withdrawals. The worst one-year drawdown since 1970 is 17%.

When you add REITs and commodities, the internal rate of return rises to 11.3%, which is nice. But the worst one-year drawdown falls to 10%. That's a 40% reduction!

Most people wouldn't immediately notice a 1.3% increase in the annual return. But they would notice a 40% reduction in the worst one-year drawdown. You can feel that.

HAI: Why does that happen?

Israelsen: Commodities and real estate have fairly low correlations to the core assets of large-cap U.S. stocks, small-cap U.S stocks and developed markets international equities.

When people buy foreign stocks, they think they are diversified. But the three main equity asset classes have correlations of 0.7 to 0.9. You don't get a lot of correlation benefit from adding more equities to an equity portfolio.

Cash is a good diversifier, and so are bonds. But they don't have a very attractive long-term return. The real benefit comes when you add REITs and commodities. They have equity-like returns, but low correlations ... and in one important way, they have lower risk than equities.

We don't equal weight the asset classes in our portfolis so our results are somewhat different that what is mentioned here - our returns are higher and our one year drawdown is smaller - but the basic concept is the same as what we have been using for some time. I developed the models we use way back in 2001 when I was working at Oppenheimer. It is nice to finally see specific academic research that verfies what we have experienced in the real world.

No comments: