Market Overview

7 Ways Professional Asset Managers Evaluate Portfolios

7 Ways Professional Asset Managers Evaluate Portfolios

There’s no shortage of competing investment strategies pitched at both institutions and individuals. The challenge for investors is to, of course, figure out which are worth it and which aren't.

In some industries, like Moody's and Standard & Poor's for credit and Kelley Blue Buck for cars, neutral third-parties have filled such a need.

But the unfortunate truth of investment management is there is no single credible judge of expected performance. Virtually all money managers supply their own performance metrics under their own set of assumptions. What remains are services such as Morningstar that take the non-controversial approach of evaluating investments in isolation and with a backward-looking analysis.

What is needed is a method of evaluation that is forward-looking and can be applied across all investment types and time horizons. While none are perfect, they are all industry standards for investment analysis.

Historical Backtesting

In many ways, this feels like the most intuitive approach. If you are proposing an investment strategy to me, tell me how it would have performed had I invested in it already. There are two things that make this method problematic, however. First, it is susceptible to what is known as in-sample bias.

Virtually anyone can construct an investment strategy that looks good in the rearview mirror. Had you put all your money in Apple Inc (NASDAQ: AAPL) in 2003 you’d likely be a millionaire many times over, but is it necessarily the best investment over the next 20 years? Who can say?

The second problem is that while history often repeats itself, it does not do so verbatim. To put it more concretely, the risks and trends we’ve seen in the past are unlikely to replay in the same way. Will there be another market crash? Most likely, yes. Will it involve subprime mortgages or overhyped internet stocks? Again, who can say? The world changes. Constantly evaluating your investments based on past risks is a bit like steering around potholes that you’ve already passed. What you need to do is look for the next pothole.

Forward-Looking Estimates

Estimating forward performance involves predicting future performance based on some sort of model. While this tries to address the obvious problems of historical backtesting, there are problems with this methodology as well. For one, accurately predicting the future return and variance of just about any investment is next to impossible, especially when the window of prediction is very short.

For broad asset classes there may be a little bit of predictability in long-term averages and indeed, many large banks publish these estimates as “Capital Markets Assumptions” on a yearly basis. It is important to note that these estimates are not necessarily attempting to predict the following year’s performance, but instead, say something about the expected average return over many years. Attempting to include an individual security with idiosyncratic risk is almost out of the question.

The other problem with static estimates is that they ignore all path dependency. If equities are expected to return 6 percent per year, it is unlikely that this return will proceed consistently at the same rate each year. Rather, there will be up years and down years. It will correlate with other asset classes. With a dynamic investment strategy, even one as simple as a so-called “glide path,” the timing of returns matters. Losses early in an investor’s lifecycle may have greater impact on their future wealth than losses that come later.

Forward-Looking Simulation

The issue of path dependency can be addressed with a slightly more complex methodology known as a Monte Carlo simulation. Instead of a single path imagined into the future, computers are used to generate many random potential paths into the future. This allows us to see a “distribution” of wealth at any point in the future, which may simply be reported as some average. The Monte Carlo framework, however, does not specify how these paths are to be generated.

This is a major misconception and we often see investment managers touting their Monte Carlo simulations without any further detail. This would be like a politician saying their plan for closing the deficit is to “create more jobs.” Well sure, that’s a great strategy, how exactly does one go about doing that?

The most widely used engine to drive these simulations is known as Geometric Brownian Motion (GBM) which is a mathematical way of saying that prices evolve randomly, but within very well-defined bounds. Succinctly, it assumes that asset returns follow a normal distribution.

The problem is, it is well known that asset returns do not move in such a well-behaved manner. Real investments, like stocks, are more likely to crash or skyrocket than predicted by this model. Do these market crashes matter? Without going into detail, the answer is yes. Day-to-day volatility does not have the same effect on your wealth as the huge moves that we saw during the housing crisis. It’s these extreme events that drive long-term wealth, and to assume them away is to lose sight of why we are simulating to begin with.

Resampling Simulation

The primary issue with the previous method was that the prices for securities were oversimplified. However, a technique called statistical bootstrapping, or resampling, can move to correct this by allowing price paths that are non-parametric and mimic some of the more complex behavior observed in real markets.

Particularly, these price paths can exhibit similar levels of skew and kurtosis (“fat tails”) that we care about. Imagine each day of historical returns is a marble. We put then mix all these marbles in a bag and start pulling out individual ones at random. Each time we pull a marble, we record its value stringing it into a complex necklace so to speak.

We can repeat this procedure as often as we like, and what we find are very realistic paths of security prices. But unfortunately, it is still not a silver bullet. Something is still missing.

Efficient market hypotheses claim (and with solid evidence), that there is little to no serial correlation in security prices day today. A stock going up yesterday has no effect on whether it will go up or down on the following day. Markets do, however, show autocorrelation in volatility. A big move today creates a greater chance of a big move tomorrow — we just don’t know what direction.

Resampling With Volatility Structure

Instead of sampling each day one at a time as in the previous example, we can sample little consecutive strings of data, maybe a week at a time or even random intervals. Within these chunks, the market will be as real as real can be, but we may see jumps between high and low volatility periods when we “stitch” them together.

To correct for this, we can actually impose a volatility structure across these two discrete chunks. This requires a little more sophisticated math, but the general idea is that if we stitch a patch of data after a high volatility chunk, we can give a bit of a boost to the volatility in the second chunk. And likewise with a low volatility chunk we can dampen the volatility in the subsequent chunk.

Adding Taxes

So far this simulation has just been limited to how we can simulate a portfolio into the future along a realistic trajectory. But price movement is only part of the story. You have an ETF that exceeds all performance expectations for the time you’ve held it; you are up $10,000. Are you $10,000 richer? No.

If you have a tax rate of 35 percent you are only $6,500 richer — not a negligible difference. So along each of these Monte Carlo paths you can expect to pay taxes. You can offset gains with losses. You may have to sell securities to pay during an up year. To neglect some semblance of a tax model in the simulation is to live in a fantasy world. To truly estimate and compare investment strategies, a tax model is absolutely critical.

Behavioral Models

What do people do when the market crashes? They panic and sell. Perhaps that is a bit too simplistic, but empirical data does support the notion that inflow and outflows are affected by the overall market level and in particular, that outflows tend to increase during market downturns. So if we take an investor’s portfolio and simulate a path into the future, one of those paths could conceivably experience a major crash, maybe dropping far more than the investor's stated risk tolerance.

With the models discussed so far, there is an implicit assumption that the investor has nerves of steel and will continue to hold and rebalance through any market condition. If we admit the empirical evidence, we know that at least some investors will react to extreme market conditions, changing their allocation to a safer set of investments. Since strong market upswings tend to happen proximate to these crashes, there is a good chance the investor will miss the market rebound, a loss that will be magnified across the remaining duration of his or her investment horizon.

What is needed is a behavioral model, some component that reflects the reality that investors and their portfolios do not exist in two separate worlds, but interact with each other: an investor becoming more or less aggressive as he or she becomes richer or poorer. Given data on inflows and outflows we can calibrate such a model and we can calibrate it according to exogenous variables such as the investor’s age, location or even spending habits. Firms like Decipher offer services like this. 

The Kitchen Sink

With all the complexity included so far, this simulation is starting to resemble a game of The Sims rather than the 40-page pdf full of pie charts that one usually has tossed at them by a financial advisor. But this is only the tip of the iceberg. We can begin to think about other interactions, with interest rates and inflation. We can begin to include future cashflow payments from social security or annuities. We can include expected liabilities such as health care costs and how that interacts with insurance premiums and payouts. We have a framework where a near-arbitrary level of complexity can be incorporated to create a truly holistic view into anyone’s financial future.

Posted-In: contributorEntrepreneurship Hedge Funds General Best of Benzinga


Related Articles (AAPL)

View Comments and Join the Discussion!