All about using portfolio optimiers for investment management.

Portfolio optimization is asset allocation. The investment software used is called a portfolio optimizer. It generates a chart called the efficient frontier.

The efficient frontier is a line made up of dots (around 100 on average. See the image below). The bottom X axis of the chart is the historical rate of return of the portfolios (returns increase as you move to the right). The vertical Y axis is risk as measured by standard deviation (risk increases as you go up). So the efficient frontier usually slopes upward, because returns increase as one assumes more risk.

Each dot is an portfolio comprised of a unique mix of the investments one allowed the optimizer to use. When you go up the efficient frontier to the next dot (toward the top right corner), it will have slightly more risky assets and slightly less conservative assets. Vice versa, when you go to the next dot down, you get less return and less risk. Each dot is a plot data point telling what the portfolio's risk and rate of return was over the selected time frame.

Asset allocation is the art and science of spreading money around between different types of investments to lower risk through diversification. Portfolio optimization just quantifies how much risk and return an asset allocation has had over one time horizon. So in a nutshell, all portfolio optimization does is refine and quantifies the risk and return characteristics of a certain mix of assets (or asset classes) over a certain time frame.

Without portfolio optimization, you're just slapping a few asset classes together and saying that you've reduced risk via diversification. Portfolio optimization tells you how much in numbers. It tells you how much risk you've reduced, and how much return you've gained (or failed to lose), by using asset allocation.

The goal is typically to find the portfolio with the asset allocation mix that is the furthest "northwest." The dot that's closest to the top left corner is the portfolio that is the most efficient. In other words, of all of the assets input into the optimizer, it's the mix that has had the most return per unit of risk over the selected time frame. This is why some investment firms are called names like, "Northwest Quadrant." They're trying to tell people that they get good returns with low risk for their clients.

The first section below goes into the details of why it's so cool, and the last section tells why it's all not as cool as all that.

In summary, it doesn't work well because as soon as one slightly changes the time frame, you get a whole different mix of assets. This huge amount of difference makes it obvious that even the most efficient portfolio possible should be taken with a huge grain of salt. When this portfolio is implemented in the Real World, the results going forward are usually nowhere as efficient at the optimizer said it would be. This is why hardly anyone uses optimizers.

Now the Details:

Portfolio optimization is the heart of what’s called "Modern Portfolio Theory," or MPT. For decades, MPT has guided investment managers responsible for trillions of dollars of pension funds, endowment funds, and other institutional portfolios around the world. MPT helps enable investors to maximize their returns and achieve their financial objectives - all while minimizing both risk and investment expenses.

In a nutshell, portfolio optimization picks up where the asset allocation process leaves off. It basically refines an asset allocation once it’s established.

The purpose, and the whole gist of MPT, is to include assets in portfolios that have good returns while providing various diversification benefits. This reduces risk without sacrificing returns.

Diversification is a good thing, and prevents problems associated with "having all of your eggs in one basket." Too much diversification can dilute performance because you have less of what’s going up than you would like, so there is a trade-off.

The combination of asset allocation and using an optimizer to optimize that asset allocation is a very effective way to get the balance of diversification that fits clients’ lives. It is the state of the art, and nobody has come up with anything that works better over the last thirty years.

MPT is not really very modern at all. Mathematician Harry Markowitz, currently semi-retired, won the Nobel Prize for this seminal work in the 1950’s. William Sharpe, currently a Stanford professor, took the complex math of Markowitz’s work and distilled it into a practical and useable form that runs on PCs in the 1980s. He then won the 1990 Nobel Memorial Prize for Economic Science for this work.

So What is Investment Portfolio Optimization Anyway?

The following is a summary on the workings of optimizers that work at the actual asset level. In other words, it works with actual stocks, mutual funds, and other assets that real people can actually own. Most optimizers work only at the asset class level, such as cash, bonds, growth stocks, and international stocks. The following discussion does not apply to lower-end optimizers that only work at the asset class level.

The optimization process is the "rocket science" of the investment management business. Basically the optimizer’s job is to find the combination of assets (in the whole portfolio) that have demonstrated the least risk given a rate of return, or the maximum rate of return given an amount of risk. This is not an exact science and there is no crystal ball, so optimization just gets you as close as computers and historical data can take you.

An optimizer is just a computer program that uses three sets of monthly historical data collected on each individual asset. An individual asset would be an actual mutual fund or stock. These three sets of data are:

· Risk as measured by standard deviation(s). Even after decades of investment and computer revolutions, standard deviation is still just about the only way a computer can understand investment risk.

· Rate of return.

· The covariance that an asset has relative to other assets. The covariance numbers tell the computer how much an asset’s market value increases or decreases relative to the other assets over the same time frame.

For example, if oil prices rise substantially over a given time frame, one would expect that airline stocks (which have fuel made from oil as their main expense) would decline in value. If airline stocks actually fell and oil stocks actually rose during this time frame, this would be called negative covariance between the two. It’s also called negative correlation but the differences are minor.

Negative covariance is rare, and it’s not needed to reduce risk. As long as adding an asset to other assets lowers overall portfolio risk, without noticeable lowering the overall rate of return too much, there is a diversification benefit.

The whole point in optimizing is to find the mix of assets that when combined has had the most return for the least amount of risk to the portfolio as a whole. This happens by finding and combining assets without high covariance so when one asset class goes down, others don’t follow it down as much as they might have without optimizing. This reduces risk because at any point in time, some asset classes may be performing badly, while others may be doing well. The opposite condition could take place at any time.

Managing risk is important because nobody can predict with certainty when assets will need to be sold to raise money to spend. Since nobody can predict these things, the risk is that something may need to be sold while it is (temporarily) down in value. Once an asset is sold, and the money spent, the loss cannot be made up again (in contrast to reinvesting the sale proceeds into another asset).

Managing risk is also important for the basic fact that clients don’t like it when they get their monthly statements and they have less money than they had the month before. They tend to call their advisors and ask them why, even when the press has been consistently reporting a down market. Nobody like this, so it works to everyone’s advantage to minimize it as much as possible. Minimizing risk also works well in model portfolios because it’s usually what advisors show prospects when they ask to see some historical track record. Showing prospects that a model is up for the year, while the market is down, is a great bottom line for a long boring story.

If this has been around so long, why haven’t I heard of it before?

You haven’t heard about this before because advisers, the press, and the commission-based industry doesn’t want you to know. They don’t want you to know because this is an area that they know they can’t compete in. Everything is as good as it gets for them right now, so why spoil it?

It’s also a source of controversy from rookie brokers to Ivy League college professors. The media doesn’t report on asset allocation because it’s boring, they focus on people with stock picks and market timers, because that gets the ratings up.

The main controversy stems from the fact that the input data the optimizer uses is “historical data.” In other words, the only thing the computer has to work with are price changes from investments that happened in the past.

So the obvious question is, how well does this process predict rates of return, risk, and covariance into future? We will summarize many books, articles, and years of study and experience here:

The optimization process is completely meaningless in predicting future rates of return. The rates of return shown for investments in the reports are based on historical data and statistical confidence levels.

Absolutely nobody knows, and no firm, university, or computer program can predict what the rate of return will be on any non-fixed asset. An example of a “fixed asset” would be a CD because you know what interest rate you will get and you always sort of know what you will get back if you cash it out early or wait until it matures.

Everything that doesn’t have rate of return set in stone when you buy it is a non-fixed, or “variable” asset. Predicting rates of return on non-fixed/variable assets just can’t be done, and anyone who says it can is either a liar or a fool!

The optimization process is good at predicting risk into the future. Risk is being measured by standard deviation of monthly returns. Standard deviation is a statistical summary that illustrates the chance of an investment’s price moving up or down by so much compared to the investment’s average price in a one-year period. This is still the only way computers understand investment risk.

Empirical evidence shows that most assets exhibit similar risk characteristics over time. In other words, a stock’s price volatility (how much it goes up or down relative to it’s average price) is, on average, about the same year after year. Being able to predict the risk on assets that you can actually own is very important- even if it’s just in the ballpark.

The optimization process is very good at predicting future covariance from past covariance, which is the main point and benefit from optimizing. In other words, oil stocks and airline stocks have moved contrary to each other for decades, and until a new source of jet fuel is found, they more than likely will continue to behave as they have in the past.

Covariance between assets that you can actually own is the key to the whole concept of MPT. It is this stability over time that makes the optimizer useful.

An Illustration

Assume you are living on an island with tropical beaches and a high mountain with good snow in the winter.

You own a mountain ski resort that returns on average 15% ± 10%, every year during the winter (the ± number is the standard deviation). But in the summer it returns -5% ± 10%.

To you, this is a high-risk deal because if you don't do really well in the winter, you won't be able to save enough, and then you'll starve in the summer when nobody skis. On a bad year you may only see +5% in the winter and -15% in the summer, for an average loss of around 10% for the year.

Now assume you want to buy another business because you don’t like the risk of only being in the ski business. Assume your only two choices are:

1) An umbrella business that returns 20% ± 15%, during the winter, and -10% ± 10% in the summer (because it rains all winter, and it never rains in the summer); or

2) A suntan lotion/sunglasses/surf shop that returns 7% ± 10% during the summer, and 5% ± 2% during the winter.

Without optimizing, most people would probably choose the umbrella business because it has a higher rate of return. And it does, but only in the winter. In the summer, you'll still be taking an even greater risk of starving if things don’t work out.

But without the optimizer, there's no way to know this. The optimizer will choose the surf business even though the return is lower. Why? Because the risk as measured by how much your rate of return varies month to month, when both businesses are combined (the ski-umbrella vs. the ski-suntan lotion combination), went from intolerably high with the ski-umbrella combination, to tolerable low with the ski-suntan business combination.

You give up more potential overall return on your money by opting for the suntan business over the umbrella business. But your risk of starving if one of your businesses has a bad year has now been mostly eliminated. This is what the optimizer does. It finds the combination of assets that will produce the highest rate of return, with the lowest amount of risk.

This portfolio, or combination of assets, is called an "efficient portfolio." This portfolio is just a certain mix of the assets used, represented by one dot on a line of dots. This line of many portfolios is called the efficient frontier. When you move up to the next dot, the amount of riskier assets increases, and safer assets decreases, and vice versa.

The investment portfolio selected in the image below is represented by the blue dot on the efficient frontier. The list of assets and their percentages are shown in the table at the bottom. The blue dot represents the optimized portfolio that met the client’s guideline asset allocation. It has less risk and more return than the index portfolio over the same time horizon. The red triangle represents the risk and return of the index portfolio over the ten-year time horizon.

Image of an efficient frontier from a portfolio optimizer.

Keep in mind that the portfolio is only as efficient as it can be given the assets the user gave it to work with. In this example, we only gave the computer three assets (businesses) to work with. In the real world, we could give it over 100 to work with. Efficient portfolios are a very good thing, are worth pursuing, and come in handy - especially when the markets are very volatile, or go down and then stays flat for a long time.

Basically the job of the optimizer is to find the combination of assets that have demonstrated the least risk given a set rate of return, or the maximum rate of return given a set amount of risk. It does this by using three sets of data all over the same historical time frame: risk as measured by standard deviation (s), the rate of return, and the covariance that asset (e.g., mutual fund) has relative to other assets. Standard deviation is a common statistical term used to measure how much something varies around its average value. For example, if something has an average value of 100, and this value varied between 95 and 105 about 67% of the time, the standard deviation would be five.

The correlation numbers tell the computer how much the asset’s market value moves relative to the other assets over the same time frame. (Correlation is the mathematical representation where covariance is scaled between 1 and -1 so computers can understand and work with it.)

For example, if oil prices rise over a given time frame, then one would expect that airline stocks (which use fuel made from oil - their second largest expense) would decline in value. If airline stocks actually fell and oil stocks actually rose during this time frame, this would be called negative covariance. Negative covariance is rare, but it’s not needed to do the job of reducing risk.

As long as adding that asset lowers overall portfolio risk, there is a diversification benefit. If the optimizer thinks adding a particular asset provides a diversification benefit, which could potentially lead to lower overall portfolio risk, without lowering the return significantly, the program will choose to use this asset at the exclusion of others. For example, if you had a portfolio with a lot of oil stocks, the optimizer would choose to add airline stocks instead of adding more oil stocks, even if the historic return of oil stocks is higher than airline stocks.

The whole point of optimizing is to find the combination of assets that when combined have had the most return for the least amount of risk to the portfolio as a whole. This happens by finding and combining assets without high covariance so when one asset class goes down, others don’t follow it down as much as they might have if you didn’t optimize. This reduces risk because at any point in time, some asset classes will be performing badly, while others are doing well. Next year, the opposite condition could take place.

Nobody can predict, and almost nothing can protect from, the risks to asset classes or actual assets (such as individual stocks). Most of this desirable diversification effect happens during the asset allocation process, but the optimizer serves as the main refining tool to lower overall portfolio risk. Optimization takes diversification another step past asset allocation.

Important Optimizer Things to Keep in Mind

The main point is to find the best combination of assets that, when combined, have shown good risk reduction while maintaining a reasonable rate of return, all while keeping your constraints in mind. In other words, the end result of the optimization process has to conform to your calculated guideline asset allocation.

This is very important, and is the most overlooked point by novice users of optimization software. Most novice optimizer users let the results of the optimizer determine the asset allocation - which is even more inappropriate and adds more risk than using model portfolios.

How well does this process predict rates of return, risk, and covariance into the future? This optimization process is meaningless in predicting future rates of return. The rates of return shown in the reports are based on historical data and confidence levels. Nobody knows, and no computer program can predict, what the rate of return will be on most assets with any degree of reliability. The process is okay at predicting risk (as measured by standard deviation). Empirical evidence shows that most assets exhibit similar risk characteristics over time. The process is good at predicting future covariance between assets. This is the main point of optimizing.

Here is something very important to keep in mind: The particular optimization software that we used to use works with actual assets. In other words, it works with the actual mutual funds and stocks that you may currently own, or will own. You may have seen this type of study done before, but using asset classes such as growth stocks, bonds, cash, etc.

In those studies, the asset class optimizer uses only groups of assets, such as bonds, to represent an actual bond mutual fund. That more general methodology does provide some value, but in order to do this work correctly, you need to use the actual assets and not asset classes. A real-life growth fund has a substantially different data set (rates of returns, risk s, and covariances) than the asset class as a whole, or other growth funds. This is why optimizing with generic asset classes, and then substituting an actual mutual fund to represent the asset class after the work is done, is defeating the whole purpose.

Why Optimizing at the Asset Level is Better than the Asset Class Level

Optimizing at the asset class level means the optimizer can only work with broad asset classes, like U.S. growth stocks, international stocks, etc. Optimizing at the asset level means the program works with investments people actually can own, like Fidelity Magellan mutual fund, Microsoft stock, etc.

Optimizing at the asset class level is by far the most popular way to go from an advisor’s point of view. This is because it’s easy, cheap, and (relatively speaking) it takes hardly any time to do the work. There are numerous inexpensive software programs that make optimizing at the asset class level easy, while looking good from a sales perspective.

After optimizing at the asset class level is done, the optimal mix of asset classes is chosen, and then the advisor will select actual investments that represent those asset classes.

For example, if the optimal portfolio chosen off the efficient frontier calls for 25% U.S. bonds, then the advisor will usually recommend 25% of the portfolio be placed into their favorite bond mutual fund. Keep in mind the software program has no idea of what the risk, return, and correlation properties of this exact mutual fund were. It only knows what the broad bond index, or asset class, used by the program looks like. The whole point is to use the asset class as a surrogate to represent this actual bond mutual fund. In other words, the advisor is pretending the U.S. bond asset class is the actual bond mutual fund the advisor wants to use in the portfolio.

That’s why we call asset class optimization, "pretend optimization." If only no-load index mutual funds that closely represent the generic asset classes were used, then this problem wouldn’t be significant enough to fuss over. If each asset class is represented by an index fund similar to that asset class, then this pretend optimal mix is pretty much the same as a real optimization. The problem with that is that you can only find index funds for about half of the asset classes out in the real world.

The worst end-result is that the optimizer has picked the asset allocation mix for the client based on asset class historical data, while completely ignoring all personal characteristics of the client’s life. In other words, one of the things asset class optimization does is determine the asset allocation mix. This is done a lot in real life because it’s quick, cheap, and easy. This is the biggest reason why optimizers get bad press in the industry.

For example, a novice advisor may give a moderately conservative investor a portfolio with way too much in equities because over some arbitrary time frame, the optimizer found a low-risk portfolio using several equity indices, and very little in bonds and cash. When this portfolio is brought out into the real world, the investor may lose money in a short period of time, get mad, and close their account. Then the novice advisor tells everyone "optimizers are bad, I’ll never use them ever again!"

Optimizing at the asset level, and keeping the calculated allocation as a constraint, solves all of these problems. It does this mostly by using actual assets that real life investors can own. Most all stocks, mutual funds, and variable annuity/life sub-accounts can be used. The program can then use each actual asset’s historic rate of return, risk, and how it goes up and down in relation to any other asset.

We only optimize at the asset level, and we always use an asset allocation mix that was calculated specifically for each client as an optimizer constraint. This way the end result is an asset allocation mix that we feel best fits the client’s life - which is more important than just finding portfolios with the highest amount of return for a given amount of risk (or vice-versa). The drawback is that it has a very long learning curve, is expensive, and it takes many times more time and work than pretend optimization.

That’s the difference in doing the work. Here are the differences in the results:

· Optimizing using asset classes overstates diversification. Asset classes include many more securities than mutual funds. For example, the S&P 500 is an index made up of 500 stocks. Most growth stock mutual funds hold less than 75 stocks. Optimizing using the S&P 500 index to represent U.S. stocks, and then actually using a growth mutual fund (that owns 75 stocks) would result in much less diversification than the report stated. This is because the computer thought you were using an asset with 500 securities, when in real life, you bought an asset in its place with only 75 stocks. You have 6 - 7 times less diversification in real life than the computer thought you had. Overstating diversification increases risk, and that’s bad.

· Mutual funds can be selected that have a long-term history of outperforming their appropriate benchmark index or asset class. In other words, a good professional advisor can usually pick better investments to work with than generic asset classes or indices.

· If a fund has had risk and return that’s way out of line with its asset class, you could be taking on a lot more risk than you thought you were taking. In every asset class, there is a wide range of assets to choose from. Some have a higher rate of return than the asset class, but also take on much more risk. This is a dangerous thing to do in an asset class like bonds. The opposite is also a problem. Some small-cap stock funds have a lot less return and a lot less risk that the small-cap asset class/index. This is the asset class that’s most likely to go through the roof in a bull market.

You’ll miss this boat by having a low-risk (low beta) fund in this asset class. The only way to know for sure what you’re getting is to optimize at the asset level because the optimizer will tend to choose the small-cap fund with the highest beta.

· There are thousands more actual investments than there are asset classes. Most low-end optimizers only have about half the asset classes needed to manage portfolio risk.

· Asset classes are just averages of all of the actual assets in that asset class. The actual risk, return, and correlation properties of a real asset is usually much different than this average.

These three sets of data are what the program uses to do the work. If you own a real asset that has a completely different set of data than the asset class, then you get little benefit from optimizing. You’re doing the simulation with an orange, and then using an apple in real life.

· Some advisors use the results of an optimization to establish the asset allocation. When an advisor optimizes at the asset class level, the result is a mix of asset classes that have shown efficient characteristics over some time frame. This would then be the asset allocation you would get in portfolios. As you’ve learned above, we feel asset allocation mixes should be determined by the client’s life situation, not by which combination of asset classes had the highest return over some arbitrary time horizon. We feel you should create a personalized asset allocation and then optimize it. Optimizing at the asset class level is the exact opposite of this.

Why You Don't Need Portfolio Optimization Software

Why is it not worth spending a few grand annually on portfolio optimizers? It's not because it's very expensive, because investment people can usually afford investment software tools.

It's not because of all of the bugs, software problems, lack of historical data, and program limitations that prevent you from doing what you need to do (these are ongoing frustrations even with the world's top investment software vendors that have been doing this since the late 80s).

It's not because it's almost impossible for "normal people" to learn how to use them correctly. Only "strange people" who are both computer and investment experts at the same time can do this. Most investment people are older salespeople with marginal computer skills. They excel in people skills, which we don't have. They understand investment management, but can barely check their e-mail. Younger people with great computer skills, have just enough information in their heads about portfolio management to be dangerous. So it's rare to have both skills. If you're not an expert at both, then optimizers are nothing but trouble.

Nor is it because it takes hundreds of hours to learn how to control it.

Nor is it because most Broker Dealer Compliance people won't let you use them because its improper use will just end up getting everyone into huge trouble.

Nor is it because it's hard to explain the concepts to clients and prospects whose brains fill up and quit listening after a few minutes of boring investment management presentations.

Nor is it because it takes from five to twenty hours to build a custom optimized investment portfolio per client.

It's because asset class correlation coefficient numbers are too random. Two assets that have moved in opposite directions over the last ten years, could have moved in sync with each other over the last five years. This problem is compounded by optimizers that work at the asset level (e.g., mutual funds), because a mutual fund may change the way it does things quarterly (which instantly negates all of the past return data which the correlation coefficient numbers were based on).

Investment vehicle returns, and the resulting correlation coefficients, are event driven - so they change daily based on millions of people making investment trades based on how they react to the daily news. This means they can't be predicted, they change as the world economy changes, and they change dramatically when you change the benchmark slightly.

If they were stable over time, then all of this would be the "Holy Grail of Investing," and something about correlation coefficients would be in the financial news every day. You never hear about it because it's just not practical to use these days for managing money for people in the Real World. That, and it's just too boring.

The bottom line is that one can't really use correlation coefficients to forecast what's going to happen over any future time horizon. This is because they are long-term averages, which means you'll most always invest money right when the there is the highest correlation, and then you'll sell right before the low point in correlations.

It's the same thing as saying that the S&P500 has had an average of +10% annual return for the last 100 years, and then watching it go down 20% as soon as you invest money, and then up 20% as soon as you get out. Even if one holds on for the long-term, the correlations will not be the same as when you first invested.

All of this empirical research is cool, and it's fascinating to see the numbers change over time, but whenever one tries to use them in the Real World, they rarely have the expected outcome. So if you're thinking that you have to be an expert at deciphering all of this calculus to get good investment portfolio returns, don't waste your time. You're just going to waste a lot of work, time, and money making a fuss about an investment strategy that works great as a concept, but fails when the details are applied.

You'll get all excited when you find something like Emerging Market Bonds having a correlation coefficient to the S&P 500 of -0.9 (near perfect negative correlation) over a recent time frame. Then as soon as you make the trades, both of these asset classes will go down 10% over the same week. It just doesn't work like it's supposed to in the Real World because the correlation coefficients change at random.

If you just grasp these basic asset allocation concepts, and then apply them using a money management system that actually works well in the Real World of real investors investing their real money, then you'll do just fine.

If you try to construct an investment portfolio based on historical asset class correlation coefficients (using any time horizon), then you're just going to fail and end up losing lots of money. This is after spending lots of time, money, and work just trying to get this extremely complex software to do what you want it to do.

You can calculate correlation coefficients between an asset and four benchmark indices by using the Portfolio Statistics sheet of the Asset Allocation Calculator.

When looking at the correlation coefficients between asset classes, they will vary depending on the time frame you select. If one were to determine a standard deviation for all of these numbers over a moving average of time horizons, one would get a range in which these numbers are stable.

For example, if you took just three years worth of correlation coefficients between large-cap growth and real estate, you'd see a huge negative correlation in the first five years of the 21st century. Everyone was hiding their money in real estate because of perceived perpetual down-to-flat equity markets. So as the S&P 500 has remained flat for going on almost seven years, real estate has had one of its biggest Bull Markets ever.

So let's say the number over these five years was -0.5. If something were to happen today to make interest rates double, then both the S&P 500 and Real Estate would go down. Let's say the correlation coefficient over this time frame is +0.5.

Then after that, let's say interest rates went down below where they are now, making both the S&P 500 and real estate go up, but the S&P 500 went up a lot, and real estate didn't near as much, so the number in this time frame is -0.25.

Then add to that, the late 90's when stocks rallied and real estate fell. The sigma of the range of values in these three time frames would be ~0.75. That's too much, and is an example of how you can get into big trouble by letting an optimizer run unconstrained.

If you make a purely scientific bet based on the correlation coefficients between these two asset classes over the last five years, then the randomness of economic science (the art form part) will step in as soon as you make the trades, and you'll lose a ton of money - which is exactly what the portfolio optimizer is supposed to prevent.

You just can't get an average correlation coefficient number over a long time horizon and expect things to move like that in the future. Any little thing will come along and will instantly negate all of the previous numbers - with the result being losing a lot of money.

The point, and all one can do, is to know (from tinkering with a portfolio optimizer for many years) out of which of the dozens of asset classes that are available to invest in, which have correlation coefficients that are stable enough to use to manage people's money in the Real World. This is what we've done here in our asset allocation software.

The short version is that to be able to perform asset allocation effectively (using correlation coefficients between asset classes as the main driving force), one must be able to distinguish the acceptable ranges in which the correlation coefficients shift.

Since like all other facets of economic science, where most everything is event driven, and therefore random and unpredictable, this part of investment management is more of an art than an exact science. So if one allows the cold science of asset allocation to let a portfolio optimizer run unconstrained, then you're just begging to lose a lot of money.

For example, in 2001, the most efficient unconstrained portfolio was 50% cash and 50% Microsoft stock. Nobody in their right mind would or should do that in the Real World, but that's what it recommended as the portfolio that will have both the least risk and highest return (Microsoft soon ended its period of historic double-digit annual gains and has been flat for years).

Portfolio optimizers are like wild beasts that need to be tamed. And the only way to do that is look at it more like an art than a science, and the only way to get a "feel for it" is to use it correctly in the Real World when your butt is on the line because you're using its recommendations to make trades with people's life savings in the Real World. After a decade of that, you develop an instinct, which is most always contrary to the science. After you develop this instinct, then managing the ever-changing details becomes so overwhelming, that it's just not worth the trouble.

Out of the dozens of asset classes that are available to invest in, most have unacceptable correlation coefficients ranges. So we limited the number of asset classes we work with to those that have 1) acceptable correlation coefficients ranges, 2) will make money, and 3) will reduce risk.

We whittled them down to about 21 - so we manage 22 asset classes, and the last one is cash, which sort of is a default that doesn't count. This way we can watch to see over time, and various market environments, how to best take advantage of the random ranges of correlation coefficients between assets classes.

This allows us to both maximize the good affects and minimize the bad affects of this randomness. This is where the art of portfolio management constrains the logical science. Over the years of tinkering with it, we have it down, and it hasn't changed much since. The results are in the returns, which is the goal, and the bottom line that speaks for itself.

Information

Menu of Retirement Planning Software

Asset Allocation Overview with Our Historical Returns

Asset Allocation Software

Asset Allocation Models with Our Historical Returns

Our Monthly Updated Mutual Fund Picks

Financial Plan Tools

Family Budgeting Calculator

Financial Planning Fact Finders

Investment Policy Statement Software

Bond Calculators

Investment Software for Comparing 22 Methods of Investing

Buy Term Life Insurance and Invest the Difference in Mutual Funds vs. Whole Life Insurance Calculator

Real Estate Software