Asset allocation answers the question of how much to put in each bucket.
This is where things can get complicated, and if you listen to the experts, it will. One of the oldest and seminal approaches to asset allocation is Modern Portfolio Theory or MPT. (MPT is also known as mean-variance theory, which will not be discussed here. There is a wealth of information on this subject on line.) But like most financial models, it relies on a fairly complex array of indicators and, most importantly, assumptions about the future. It is not a settled argument, and it never will be, because you can’t prove an unknown future. It all assumes a future that is similar to the past in some ways. I spent some effort exploring MPT and decided against using it to construct an optimal portfolio. I knew that my goal was to first build a learning machine that would in turn build an optimal portfolio. So this begins with some simple stuff.
In the simplest world there are 2 buckets to consider: A stock bucket and a bond bucket. (Five buckets were selected as a sweet spot in the previous article. Three of those were stock buckets and two were bond buckets, and I will return to that later, as it still is a statistical sweet spot.) These 2 buckets represent the 2 asset classes that most often tend to move in opposite directions, thus reducing portfolio volatility while providing reasonable long term returns. Usually.
A very popular rule of thumb suggests, 60% should be in the stock bucket and 40% in the bond bucket. And a popular variant of that rule, states when you are young you can put more in the stock bucket and as you age you should move more to the bond bucket. This rule has become popular because it has worked quite well for 30 years or more. It is a good place to start and use as a bench mark against other ideas. Ultimately, it is your personal situation and market conditions that should dictate asset allocation.
Your personal situation is a function of your age, income, spending, assets, and risk tolerance. You can probably do a good job of figuring out most of those things except risk tolerance, which I will get into later. Market conditions are largely a function of current valuations and interest rates, which tend to drive future expectations. It's not intuitively obvious, but it's not as arcane as the experts would have you believe either.
Getting back to the simple example of a portfolio of 2 assets: a stock fund and a bond fund; How did the simple 60/40 rule work in the past. (As every financial person will tell you, the past does not mean it will happen in the future and its true, but it’s a place to start. Much more on that later.) I’ll start with 2 funds from the “bucket list”: VTSMX for stocks and VUSTX for bonds. (I chose these Vanguard funds because they are real things you can buy and sell and have a history that dates back to 1993, but any other equivalent choices will yield similar results over the same time periods).
How would that portfolio have performed over their hisotry if you maintained a re-balanced 60% VTSMX and 40% VUSTX blend? It would have returned an average of about 8% per year. (Inflation and taxes would have taken a bite out of that number, but we won’t consider that for now). The five bucket portfolio, (VEIEX, VTSMX, VTRIX, VBMFX, VUSTX) which covers a broader range of assets, performed about the same over the same period. Improving the portfolio's granularity with 15 asset classes and the overall performance (returns and volatility) would still remain roughly the same over the same time frames. So to keep things equal and simple, I will use those 5 Vanguard funds as a benchmark for further study. Those mutual funds have a long track record for training the learning machines, but using other ETFs from the same categories listed in the previous article would perform the same. (By example SCHE, SCHB, SCHF, SCHZ, TLT would perform the same over the same period.)
There is one more common and simple asset allocation model that should be considered . It is putting the same amount in each bucket, also known as the naive asset allocation model. If that 5 bucket portfolio has an equal amount of those 5 Vanguard funds in each bucket, it winds up as a 60% equity and 40% bond portfolio. That turns into a good benchmark and has been a very nicely performing and broadly diversified portfolio for many years.
But no portfolio will always deliver its average return in the future. The variability of any asset or portfolio is its volatility, and the measure is called its standard deviation.
Standard deviation is a common measure of volatility, or risk, but it may not be intuitive for most. A more intuitive measure for you to consider as an initial guide to your risk tolerance, is how much would a given portfolio or asset drop in value over a single 12 month period. A nicely diversified portfolios all had an average annualized standard deviation of about 12%. And all suffered a fairly painful drop of more than 30% in a single 12 month period during the last 20 plus years. Most experts will quote you the volatility, (viz. standard deviation), as the risk of an asset or portfolio as the annualized standard deviation. So multiply that by 2 or 3 to get how much that investment may drop in a bad 12 month period. That should give you a gut level expectation of your personal sense of risk tolerance.
Past performance of an asset or portfolio relative to its future expectations are critical to understand and the relationship is neither obvious nor constant. But there are macro-economic forces that can have a major influence on long term future expectations: interest rates and current valuations. Interest rates have been mostly falling for over twenty years and bonds do especially well when interest rates fall. As of this writing, in 2019, interest rates are historically low and rising. It is unlikely that the next ten or twenty years will give bonds the same tail wind. And stock valuations, by some metrics are now near all time highs. Neither benefit is likely to be sustained ten years into the future.
Without getting into the numerical weeds now, (I will later) a total return of 8% per year for that benchmark portfolio for the decade of the 2020's is not very probable (more on that later in the Future Expectations discussion). And that is before you take out inflation, taxes and any other "expert" fees you may pay. And it won't be a smooth ride, returns will likely gyrate from year to year. No fixed combination of the 5 buckets would have returned significantly more over the past 20 years. US equities returned the most and putting more weight into bonds would have reduced volatility.
Well, this isn't trivial enough to use a simple old rule of thumb and be done with it. You really must understand how much you need relative to how much you are likely to earn (or lose) given any investment strategy. I would suggest you have the necessary tools in your investing tool bag: Something to evaluate your plans relative to future possibilities (think Monte Carlo Simulator, not as exotic as the name, and I built a simple but powerful one you can have, just contact me), and a tool to estimate long term future expectations. If you aren't satisfied that a conventional static approach will achieve your goals, then a quantitative tool to explore a more active investing strategy is something to consider; the learning machines of InvestEngines.
The next step is to understand what the future may look like and if it will meet your needs.